Você está na página 1de 217

The Pennsylvania State University

The Graduate School


College of Earth and Mineral Science

VULNERABILITY OF THAI RICE PRODUCTION TO SIMULTANEOUS


CLIMATE AND SOCIOECONOMIC CHANGE:
A DOUBLE EXPOSURE ANALYSIS

A Dissertation in
Geography
by
Ratchanok Sangpenchan

2011 Ratchanok Sangpenchan

Submitted in Partial Fulfillment


of the Requirements
for the Degree of

Doctor of Philosophy

December 2011

The dissertation of Ratchanok Sangpenchan was reviewed and approved* by the


following:

Brent Yarnal
Professor of Geography
Associate Head of the Department of Geography
Dissertation Advisor
Chair of Committee

William Easterling
Professor of Geography

John Kelmelis
Professor of International Affairs

James S. Shortle
Distinguished Professor of Agricultural and Environmental Economics

Karl Zimmerer
Professor of Geography
Head of the Department of Geography

*Signatures are on file in the Graduate School

iii
ABSTRACT
This dissertation explores the vulnerability of Thai rice production to
simultaneous exposure by climate and socioeconomic change so-called double
exposure. Both processes influence Thailands rice production system, but the
vulnerabilities associated with their interactions are unknown. To understand this double
exposure, the research adopts a mixed-method, qualitative-quantitative analytical
approach consisting of three phases of analysis involving (in order) a Vulnerability
Scoping Diagram, a Principal Component Analysis, and the EPIC crop model. Using
proxy datasets collected from secondary data sources at the provincial level, the first and
second phases together identify the key variables representing each of the three
dimensions of vulnerability exposure, sensitivity, and adaptive capacity. Results show
that the greatest vulnerability in the rice production system occurs in households and
areas with high exposure to climate change, high sensitivity to climate and
socioeconomic stress, and low adaptive capacity. The results also show the geographical
distribution of vulnerability across the country and locate four provinces with low
vulnerability to the double exposure. In the third phase, for each of these four provinces,
the EPIC crop model simulates rice yields associated with future climate change as
projected by two downscaled global climate models. Climate change-only scenarios
demonstrate that yields are expected to decrease 10% from the current productivity
during 2016-2025 and 30% during 2045-2054 under projected changes in climate and
rising CO2 levels. Scenarios applying both climate change and improved technology and
management practices show that a 50% increase in rice production is possible, but

iv
requires strong collaboration between sectors to advance agricultural research and
technology. Moreover, disseminating these advancements requires the strong adaptive
capacity in the rice production system characterized by well-developed social capital,
social networks, financial capacity, and infrastructure and household mobility at the local
scale. The vulnerability assessment and climate and crop adaptation simulations used
here provide useful information to decision makers developing vulnerability reduction
plans in the face of concurrent climate and socioeconomic change.

v
TABLE OF CONTENTS
LIST OF FIGURES ..................................................................................................... viii
LIST OF TABLES ....................................................................................................... x
ACKNOWLEDGEMENTS ......................................................................................... xi
CHAPTER 1 INTRODUCTION ................................................................................ 1
1.1 Impact factors: Climate change, the socioeconomic system, and their
interaction ...................................................................................................... 3
1.1.1 Climate factors...................................................................................... 4
1.1.2 Socioeconomic factors.......................................................................... 6
1.1.3 Interactions between climate and socioeconomic factors: Double
exposure ................................................................................................. 8
1.1.3.1 Demand ...................................................................................... 9
1.1.3.2 Supply......................................................................................... 10
1.2 Agriculture in Thailand................................................................................... 13
1.2.1 Overview...................................................................................................... 13
1.2.2 Climate and socioeconomic impacts ........................................................... 15
1.3 Research Framework: Vulnerability and Scale .............................................. 18
1.4 Research Goal, Questions, and Objectives ..................................................... 24
1.4.1 Research questions ............................................................................... 24
1.4.2 Research objectives .............................................................................. 25
1.5 Study Area ...................................................................................................... 25
1.6 Scope of the Study .......................................................................................... 30
1.7 Thesis Overview ............................................................................................. 31
CHAPTER 2 METHODS ........................................................................................... 32
2.1 Phase 1: The Vulnerability Scoping Diagram ................................................ 33
2.1.1.1 Physical vulnerability ................................................................. 35
2.1.2.1 Socioeconomic vulnerability ...................................................... 36
2.1.2.1.1 Human capital .................................................................. 36
2.1.2.1.2 Financial capital ............................................................... 37
2.1.2.1.3 Social capital .................................................................... 39
2.1.2.1.4 Physical capital................................................................. 40
2.1.2.1.5 Natural capital .................................................................. 41
2.1.3 Developing the VSD............................................................................. 42
2.1.4 VSD input data ..................................................................................... 46
2.1.4.1 Climatic variables: temperature, rainfall and SPI ...................... 46
2.1.4.2 Other biophysical data ................................................................ 50
2.1.4.3 Socioeconomic proxies .............................................................. 51
2.2 Phase 2: The Principal Component Analysis .................................................. 53

vi
2.3 Phase 3: Crop Model ...................................................................................... 55
2.3.1 Description of EPIC and the applications ............................................ 58
2.3.2 EPIC input data..................................................................................... 62
2.3.2.1 Climate change projection data .................................................. 62
2.3.2.2 Soil data ...................................................................................... 66
2.3.2.3 Crop growth and crop management data .................................... 67
2.3.2.4 Adaptation options ..................................................................... 68
2.4 The justification for this research and its focus .............................................. 69
CHAPTER 3 PRINCIPAL COMPONENT ANALYSIS OF VULNERABILITY .... 71
3.1 Overview of PCA .......................................................................................... 71
3.2 Exposure ......................................................................................................... 76
3.2.1 Preliminary analysis ............................................................................. 76
3.2.2 PCA results ........................................................................................... 78
3.2.3 Interpretation of PCA results ................................................................ 82
3.3 Sensitivity ....................................................................................................... 85
3.3.1 Preliminary analysis ............................................................................. 85
3.3.2 PCA results ........................................................................................... 86
3.3.3 Interpretation of PCA results ................................................................ 90
3.4 Adaptive Capacity .......................................................................................... 96
3.4.1 Preliminary analysis ............................................................................. 96
3.4.2 PCA results ........................................................................................... 97
3.4.3 Interpretation of PCA results ................................................................ 100
3.5 Conclusions..................................................................................................... 104
CHAPTER 4 VULNERABILITY MAPPING ........................................................... 106
4.1 Physical vulnerability ..................................................................................... 107
4.2 Social vulnerability ......................................................................................... 112
4.2.1 Sensitivity ............................................................................................. 112
4.2.2 Adaptive capacity ................................................................................. 119
4.3 Calculating overall vulnerability .................................................................... 123
CHAPTER 5 CLIMATE AND CROP YIELD SCENARIOS ................................... 126
5.1 Climate projection scenarios........................................................................... 127
5.2 Crop and crop management scenarios ............................................................ 135
5.2.1 Crop and crop management baseline parameterization ........................ 135
5.2.1.1 Crop parameter ........................................................................... 136
5.2.1.2 Soil data ...................................................................................... 138
5.3 Simulation results for scenarios 1 and 2 ......................................................... 139
5.3.1 Crop yields............................................................................................ 140
5.3.2 Evapotranspiration and water use efficiency ........................................ 142
5.3.3 Discussion: impacts of combined crop-climate relationships on
yields ...................................................................................................... 145

vii
5.3.3.1 Scenario 1 ................................................................................... 145
5.3.3.2 Scenario 2 ................................................................................... 147
CHAPTER 6 ADAPTIVE STRATEGY SCENARIOS ............................................. 149
6.2 Simulation results for Scenarios 3 and 4 ........................................................ 154
6.2.1 Crop yields............................................................................................ 154
6.2.1.1 Option 1: No Sc + Min ............................................................... 154
6.2.1.2 Option 2: No Sc + Max .............................................................. 155
6.2.1.3 Option 3: Sc + Min ..................................................................... 158
6.2.1.4 Option 4: Sc + Max .................................................................... 158
6.2.2 Water use efficiency and evapotranspiration ....................................... 159
6.2.3 Discussion: impacts of combined crop-climate and elevated CO2
relationships on yield ............................................................................. 165
6.2.4 Integrating the results from all phases .................................................. 166
CHAPTER 7 DISCUSSION AND CONCLUSIONS ................................................ 175
7.1 Thai rice production and double exposure...................................................... 175
7.2 Is Thai rice production moving towards resilience? ....................................... 178
Appendix A Socioeconomic variables and proxies for sensitivity and adaptive
capacity ................................................................................................................. 184
Bibliography ................................................................................................................ 188

viii
LIST OF FIGURES
Figure 1.1: Political and topographic map of Thailand ............................................... 27
Figure 2.1: Vulnerability Scoping Diagram................................................................ 34
Figure 2.2: Vulnerability Scoping Diagram for the indicators of rice farm
household vulnerability ........................................................................................ 43
Figure 2.3: Methods applied to this study................................................................... 63
Figure 3.1: Eigenvector-based classification framework............................................. 73
Figure 3.2: Scree plot for the PCA of exposure variables .......................................... 80
Figure 3.3: PCA scree plot of the sensitivity components .......................................... 87
Figure 3.4: PCA scree plot of the adaptive capacity components .............................. 98
Figure 3.5: Final VSD with key vulnerability indicators ............................................ 105
Figure 4.1A: Exposure component 1: Minimum temperature .................................... 110
Figure 4.1B: Exposure component 2: Agro-climate ................................................... 110
Figure 4.1C: Exposure component 3 Maximum temperature ..................................... 111
Figure 4.2A: Sensitivity component 1: Household economy ..................................... 116
Figure 4.2B: Sensitivity component 2: Land scale ..................................................... 116
Figure 4.2C: Sensitivity component 3: Human capital ............................................... 117
Figure 4.2D: Sensitivity component 4: Production capacity ..................................... 117
Figure 4.2E: Sensitivity component 5: Land tenure and security of land
ownership .............................................................................................................. 118
Figure 4.3A: Adaptive capacity component 1: Social capital and social network ..... 121
Figure 4.3B: Adaptive capacity component 2: Financial capacity .............................. 121
Figure 4.3C: Adaptive capacity component 3: Infrastructure and household
mobility ................................................................................................................. 122
Figure 4.4: Location of the four case study provinces ................................................ 125

ix
Figure 5.1: Maximum temperature compared between CSIRO (left) and MIROC
(right).. .................................................................................................................. 130
Figure 5.2: As in Figure 5.1, but for minimum temperature....................................... 131
Figure 5.3: As in Figure 5.1, but for diurnal temperature range ................................. 132
Figure 5.4: As in Figure 5.1, but for average rainfall ................................................. 133
Figure 5.5: Phenology of KDML 105 (expressed in number of days); sowing
starts in May and transplanting in June ................................................................ 137
Figure 5.6: Simulated yields from EPIC under Scenario 1 and 2 ............................... 141
Figure 5.7: Crop water use efficiency (WUEF) simulated by EPIC........................... 143
Figure 5.8: Evapotranspiration (ET) simulated by EPIC ............................................ 144
Figure 6.1: Simulated yields from EPIC under Scenario 3 during STF (upper) and
LTF (upper) .......................................................................................................... 156
Figure 6.2: Simulated yields from EPIC under Scenario 4 during STF (upper) and
LTF (lower) .......................................................................................................... 157
Figure 6.3: Water use efficiency simulated by EPIC under Scenario 3 for STF
(upper) and LTF (lower). ...................................................................................... 160
Figure 6.4: Water use efficiency simulated by EPIC under Scenario 4 for STF
(upper) and LTF (lower) ....................................................................................... 161
Figure 6.5: Evapotranspiration simulated by EPIC under Scenario 3 for STF
(upper) and LTF (lower) ....................................................................................... 163
Figure 6.6: Evapotranspiration simulated by EPIC under Scenario 4 for STF
(upper) and LTF (lower) ....................................................................................... 164
Figure 7.1: Interactions of global and local/national scales in determining the
resilience of Thai rice production ......................................................................... 180

x
LIST OF TABLES
Table 2.1: List of proxy variables for vulnerability indicators .................................... 45
Table 2.2: Description of climate models ................................................................... 65
Table 2.3: Grid points for each study areas ................................................................. 65
Table 3.1: The KMO and Barletts test results for exposure variables ....................... 80
Table 3.2: Three-component solution with temperature and moisture items
identified as important on the exposure dimension of vulnerability .................... 81
Table 3.3: Percentage of variance explained by the three components retained in
the exposure PCA ................................................................................................. 81
Table 3.4: The KMO and Bartletts test results for sensitivity variables.................... 85
Table 3.4: Five-component solution with items identified as important on the
sensitivity dimension of vulnerability .................................................................. 88
Table 3.5: Percentage of variance explained by the five components retained in
the sensitivity PCA ............................................................................................... 89
Table 3.6: The KMO and Bartletts test results for adaptive capacity variables ........ 97
Table 3.7: Three-component solution with items identified as important on the
adaptive capacity dimension of vulnerability ....................................................... 99
Table 3.8: Percentage of variance explained by the three components retained in
the adaptive capacity PCA .................................................................................... 100
Table 5.1: Scenarios established for the four case studies .......................................... 126
Table 5.2: Agronomic and management parameter input data of KDML105 for
Scenarios 1 and 2 .................................................................................................. 138
Table 5.3: Characteristics of selected soils used for Scenarios 1 and 2 ...................... 139
Table 6.1: Four options in adaptive strategies designed for Scenarios 3 and 4 ........... 151
Table 6.2: Summary of results .................................................................................... 169
Table A.1: Socioeconomic variables and proxies for sensitivity and adaptive
capacity ................................................................................................................. 185

xi
ACKNOWLEDGEMENTS
First of all, I wish to thank the Agricultural Research Development Agency,
Thailand for providing the financial support for my graduate studies. To my family and
friends, I deeply appreciate and would like to thank for their support in helping me
overcome the hard time in my academic life. To the faculties and community of the
Department of Geography, I would like to extend my appreciation to them for creating a
positive academic environment to all of us in the department.
I wish to convey my deep gratitude to the dissertation committee, Drs. William
Easterling, John Kelmelis, and James Shortle for their intellectual support in developing
my research. It has been a valuable experience to be able to work with them. I would like
to extend my gratitude to Dr. Jimmy R. Williams at Blackland, Texas Agrilife Center for
his guidance in creating parameters for EPIC crop model analysis. Without this help, the
analysis would not have been completed. A number of officials from various institutes in
Thailand, such as the Meteorological Department, Land and Development Department,
Office of Agricultural Economic, and National Statistical Office, have provided valuable
data for developing my dataset. Without all of these help, this research would not have
been possible. To them, I would like to express my deep appreciation for their support.
Most of all, I am deeply grateful to my dissertation advisor, Dr. Brent Yarnal for
his intellectual guidance, encouragement, and dedication in building my intellectual and
academic success. His guidance and attitude have made me believe that I could become a
good scholar and it has been an honor to have known and worked with him.

CHAPTER 1
INTRODUCTION
This research will examine the agricultural impacts of and vulnerabilities to
integrated global climate change and socioeconomic change. More specifically, the
investigation will examine the interacting effects of climate change and socioeconomic
conditions on rice production in Thailand. Agricultural systems are vulnerable systems
due to a high dependency on temperature and precipitation. Variations and long-term
changes in these variables pose challenges to farmers and to a society that relies on the
output of the agricultural system. Although there are recent findings that the CO2
fertilization associated with rising temperature may offset the loss of crop yield by
enhancing crop water use efficiency (e.g. Kimball et al. 2002), severe impacts could
occur if that benefit does not materialize. Individual farmers are inevitably vulnerable to
the negative impacts of climate change, and particular adaptation strategies, such as
adopting new seed varieties, relocating the farm, or installing irrigation systems, are
usually required (Easterling et al. 1993). The adaptation strategies implemented,
however, must cater not only to direct climate manifestations but also to non-climatic
factors, such as socioeconomic change in the agricultural system (Parry et al. 2004). I
will employ the double exposure framework (OBrien et al. 2000) in this study to
assess the joint impact of climate change and socioeconomic change.
The double exposure framework recognizes that socioeconomic change is an ongoing process that can pose a direct or an indirect effect on an agricultural sector through

2
economic policy, market price, and crop yield (OBrien et al. 2000). This process can
help mitigate the loss of or exacerbate the impacts on an existing agricultural system in
addition to the impacts from climate change. Therefore, this study assesses the
vulnerability of agricultural production by taking into consideration the processes of
climate change and socioeconomic change rather than focusing on a single process,
which has been the trend in previous research (Bachelet et al. 1992; Matthews et al. 1997;
Adejuwon 2006). Even though agricultural effects are mostly discussed at larger scales,
individual farmers are likely to confront and respond to the impacts resulting from this
double exposure, and they are likely to be the most sensitive group in the agricultural
production system. Therefore, the gains/losses from double exposure at the national level
should not be extrapolated as the gain/loss at a lower level (e.g., an individual farmer)
(Reilly et al. 1994). Hence, in addition to addressing larger-scale relationships, the study
will assess the vulnerability of farmers.
This research uses a case study approach to assess the impacts and vulnerabilities
associated with double exposure in Thailand. Thailand is currently experiencing
economic prosperity and is ranked as a top global exporter of rice. However, the country
faces challenges due to both biophysical and socioeconomic constraints, especially in the
major rice production area of central Thailand. Given these constraints, the long-term
competitive position of the Thai rice economy is uncertain. This research questions
whether Thai rice production can overcome the current and future impacts from double
exposure and remain competitive in the global rice market. What are the ideal

characteristics and adaptive strategies required to mitigate the negative impacts that may
occur in the future?
The outline and format of this chapter is described by section. Section 1.1
addresses the two major impact factors (climate change and the socioeconomic system)
and their interaction with rice production. Section 1.2 gives some brief information about
agriculture, including specific details about rice production in Thailand. Section 1.3
explains the vulnerability framework that will be used in this research, and presents a
discussion of scale considerations as well. Sections 1.4 and 1.5, respectively, describe the
objectives and the study area Thailand. Section 1.6 notes the scope of the study, and
Section 1.7 concludes the chapter with an overview of the dissertation.

1.1 Impact factors: Climate change, the socioeconomic system, and their interaction
There are increasing numbers of studies focusing on assessing vulnerability to
multiple stressors rather than to a single factor. This research focuses on the interaction of
two processesclimate change and socioeconomic changethat result in positive and
negative impacts on Thai agricultural production. I will first address the basic ideas
behind both factors beginning with climate and moving to the socioeconomic system.
Then I will review the research on agricultural production as an exposure unit influenced
by the interconnection between these two factors. Double exposure will frame the
research idea, the literature review, and the methods used in this research.

1.1.1 Climate factors


Many regional studies have identified how climate plays a major role in
influencing biophysical sectors and that the impacts are not uniformity distributed. The
change in climate refers to the change in the parameters of the distribution (Kate et al.
1985). As widely recognized, an increase of greenhouse gas (GHG) emissions has
contributed to global climate changes including rising sea levels, elevated temperatures,
higher variability in seasonal rainfall, and changes in the frequency and intensity of
weather- and climate-related natural hazards. The projected changes in climate identified
by the Intergovernmental Panel on Climate Change (IPCC) reflect spatial differences in
magnitude and direction of climate trends for multiple regions of the world.
Based on the Fourth Assessment Report (AR4) of IPCC, atmosphere-ocean
general circulation models (AOGCMs) project an increase in global temperature from
2011-2030 compared to the historical baseline 1961-1990 of about 0.64-0.69 C. Greater
increases in temperature of 1.3-1.8 C are projected for mid-century, 2046-2065 (Meehl
et al. 2007). Different magnitudes of warming are reported for various regions. For
example, most areas of Northern America, Europe, Africa, the Mediterranean, and
continental areas of Australia are expected to be warmer than the global annual mean
temperature. The projected temperatures in South Asia, East Asia, and most areas of
Southeast Asia are similar to the global annual mean temperature (Christensen et al.
2007).

Projected precipitation changes have different patterns than temperature changes.


Increases in average rainfall are projected for northern Europe, Canada, the northeastern
US, northern Asia, and most areas of Southeast Asia. Decreases in average rainfall are
projected to occur in North Africa, southern Australia, Central America, the southwestern
US, Central Asia, and Central Europe. The Mediterranean and southwestern Australia are
projected to be at high risk from drought conditions (Christensen et al. 2007). On the one
hand, monsoonal precipitation is likely to increase in Asia and the southern part of the
West Africa; on the other hand, decreases are expected in the Sahel, Mexico, and Central
America in association with increasing precipitation over the eastern equatorial Pacific
through changes in the Walker Circulation and local Hadley circulation (Meehl et al.
2007).
Data produced by AOGCMs, however, has coarse resolution and cannot
sufficiently capture the finer resolution needed to assess climate impacts at the regional
scale. Therefore, multiple regional climate models and statistical techniques have been
developed to downscale regional-scale climate variables from the coarse-resolution data
of the AOGCMs (Mearns et al. 2003; Christensen et al. 2007). Nowadays, the climate
information simulated from regional climate models, such as CCSM3, CSIRO-Mk3,
UKMO-HadCM3, and ECHAM5, has been widely used in the study of climate change
impacts (Polsky et al. 2000; Parry et al. 2004). Despite claims that downscaling
techniques have successfully simulated future regional climates (Reilly 2002; AIACC
2006; Christensen et al. 2007), the accuracy of simulated climate variations is still poor

for some regions, such as Southeast Asia, which requires a finer-scale analysis to capture
its physical diversity (Boer and Faqih 2004). Additionally, models still have a low ability
to represent the ENSO variability crucial to defining accurate interannual monsoonal
rainfall (Christensen et al. 2007). Analysis of the performance of several regional climate
models has also shown significant differences from one climate model to another, thereby
requiring further regional model development (Mearns 2003; Boer and Faqih 2004;
Wang et al. 2005).
Because deficiencies of the models in projecting future regional climates remain,
it is preferable to use multiple regional climate models to cover a range of potential
impacts from future climate changes (Brown and Rosenberg 1999; Reilly 2002). The
climate variables used in this research come from two regional climate models: the
Australian CSIRO model from the Division of Atmospheric Research and the Japanese
MIROC (hires) model from the Center for Climate System Research Institute. Two
climate datasets will establish a climate envelope indicating a range of possible climate
conditions and impact scenarios for the study area.

1.1.2 Socioeconomic factors


Early impacts studies generally considered regional economic effects of climate
change or of economic change (Kates et al. 1998), rather than the interactive process of
simultaneous changes in the climate and economy (OBrien and Leichenko 2000). These

early studies paid attention to cause-effect relationships between climate and


socioeconomic factors for example, how climate will potentially affect regional
livelihood issues, such as food security, farm income, or market price (Kumar and Parikh
2001). This suggested possible vulnerability of individuals or sectors to future changes in
climate.
However, this view assumes the socioeconomic system is static rather than
dynamic, changing through space and time (Fssel 2007; OBrien et al. 2007). The
assumption of a static socioeconomic system leads to mismatches caused by
extrapolating the societal conditions associated with future climate change from present
societal conditions. This approach therefore overlooks the ability of individuals and
social systems to adjust to the constant changes across a range of spatial and temporal
scales (Fssel 2007). Adaptation strategies responding to such results can also be
misleading (Dockerty et al. 2006). For these reasons, there is a need foCr more
integrative approach that links the two dynamic factors, climate change and economic
globalization (OBrien and Leichenko 2000; Cutter 2003; Dockerty et al. 2006; Fssel
2007). One important way to address impacts based on the interactions between these two
stressors is the double exposure framework of Leichenko and OBrien (2008).

1.1.3 Interactions between climate and socioeconomic factors: Double exposure


The key idea behind the double exposure framework recognizes that responses
and decision-making of individuals, groups, or societies are influenced by the interactions
of at least two simultaneously operating systems in their case, climate and economics
(OBrien and Leichenko 2000). As suggested above, this framework argues that
traditional impacts research, which considers multiple impact-driven factors in
separation, overlooks the cumulative effects of both climate and economics, which
simultaneously interact with each other (Belliveau et al. 2006). The result of the
interactions can both mitigate the losses of and exacerbate the impacts on an existing
system from climate change alone (OBrien and Leichenko 2000). To date, many
researchers have shown interest in addressing the dynamic role of various factors (e.g.
political, cultural, technological, and economic) integrated with an impact study of the
changes in climate conditions (e.g. OBrien and Leichenko 2000; Belliveau et al. 2006;
Acosta-Michlik 2008). For example, double exposure studies of agriculture generally pay
attention to the linkage between processes of climate and of economic globalization
(Belliveau et al. 2006). Variations and changes in climate can pose a threat to agricultural
production, such as decreasing yield and/or lower yield quality. At the same time,
international, regional, and local market price and policy are also constantly adjusting and
changing in response both to climate and to other influences. Therefore, potential impacts
on agriculture at all scales do not simply derive from climate but also from complex
interactions with economics, market policy, and so on. In the next section, the literature

review will discuss demand, supply, and resulting prices as key influences on crop
production

1.1.3.1 Demand
Recent research has focused on the transforming role of interacting driving forces
such as population increase, income growth, and prices as major factors that, in addition
to climate factors, influence the changing demand in food crops (Nelson et al. 2009;
Rosegrant et al. 2001). Driven primarily by developing countries, the world population
increasing from 6 billion people in 2009 to about 7.5 billion people in 2020 and to about
9 billion people in 2050 resulting in an increasing absolute demand for cereals (Rosegrant
et al. 2001; FAO 2009a; Nelson et al. 2009). Moreover, the crop demand is also
determined by the changes in dietary preference due to higher incomes in developing
countries that shift grain crop consumption towards high protein food. This shift may
result in higher demand for animal feedstock, leading to the conversion of land from
grain crops for human consumption either to grassland for feedstock or grains for animal
consumption (Rosegrant et al. 2001). Yet, the demand for human grain consumption
remains high because of low-income countries such as Bangladesh, Nepal, Cambodia,
Myanmar, and Philippines (Nguyen 2002; Nelson et al. 2010). As a consequence, the
overall growth rate of grain crop demand continues to increase.

10

Price is another indicator that influences the impacts of climate change and
socioeconomic change on an agriculture sector. Based on fundamental economic
principles, changes in supply and demand are related to changes in prices, except for
inelastic commodities such as rice. As rice is an essential food for daily consumption in
many countries, consumers continue to buy rice even when the price increases. The
International Food Policy Research Institute (IFPRI 2010) shows that world food prices
for most agricultural crops including rice will continue to increase by 60% between 2000
and 2050 under a no climate change scenario. Price increases are driven by population
and income growth as well as increased demand for biofuel. Under a climate change
scenario, projected lower grain supplies will increase relative demand and then drive the
price higher than the no climate change scenario by 30% with no CO2 fertilization
effect. Price is a bit lower when CO2 fertilization is accounted for.

1.1.3.2 Supply
Besides climate factors, energy prices, urbanization, agricultural investment, and
technology and government trade policies are key factors that affect agricultural output
on the supply side (Lambin et al. 2001; Rosegrant 2001; Von Braun 2008; Thongrattana
2009). Energy prices are fundamental determinants of food crop production and prices.
High energy prices affect agricultural production by directly increasing the costs of
operating machinery and using fuel-based inputs, such as fertilizers, pesticides, irrigation,
and transport (Braun 2008). The intensive use of fuel-based inputs means a significant

11

increase in production cost and decrease in farm income. This economic constraint on
Thai farmers has been reported to reduce the adaptive capacity of farmers by inhibiting
them from adopting farm management techniques that increase yields (Isvilanonda and
Hossain 2000; Mitin 2009). However, in 2007, despite the rising costs of energy and
fertilizers, rice crops in Thailand used over 262 thousand tons of nitrogen fertilizer,
which resulted in the increase in rice production costs up to 50% (Krisner 2008).
Nonetheless, because of high demand and consequent high prices from the international
market, the production of Thai rice remained high despite the very high cost of
production (USDA 2007).
Urbanization associated with emerging economic development places demands on
essential agricultural resources. With economic growth in Thailand, population has
concentrated in cities and metropolitan areas in the nations Central Plain and has
extended into the southern North region. These areas are also major cultivation zones for
Thai rice, making up over 80% of the nations total rice production land area
(Kupkanchanakul 2000). This urban growth reduces the available crop area and
agricultural employment through competition between urban and farm work and attrition
of farm workers from lost land. Urban growth simultaneously increases the competition
for water among household and commercial consumption, electric generation, and crop
production (Shivakoi et al. 2008). Meanwhile, the demand for rice continues to increase
despite increasingly limited resources. Thus, Thai rice production faces the twin
production challenges of shrinking supply of available cropland and water for irrigation.

12

These declines cause a reduction in the food supply and consequently lead to higher food
prices (Lambin et al. 2001; Shivakoti et al. 2008).
The increased demand for biofuel feedstock has contributed to a rise in food
prices and further constrained food supplies. With high concerns over surges in oil prices,
energy security, and climate change, many experts think that the transition from fossil
fuels to biofuels promises to buffer price shocks, improve energy security, and reduce
carbon emissions. However, an increase in demand for biofuel feedstock reduces supplies
of cereals because farmers naturally convert their land to more profitable crops. Low
supplies contribute to rapid price increases for rice and other cereal crops (Rosegrant
2001; Nelson et al. 2009).
With or without climate change and even with the limitations to production noted
above, agricultural research and technology is expected to increase productivity and meet
the continually growing demand for food (Phelinas 2001; Shivakoti et al. 2005; Nelson et
al. 2009). To make sure that these constraints do not overwhelm the agricultural systems
ability to meet this demand, it is crucial for the government and its policy makers to give
priority to investments in rice production technologies such as new high-yield varieties
that meet customer taste and market demand, demand less water, require less intense
inputs, and suit local biophysical conditions.

13

1.2 Agriculture in Thailand

1.2.1 Overview
Agriculture has continuously played an important role in Thailands economy and
society by providing food, commodities, and employment. Despite the significant
decrease of its contribution to Thailands Gross Domestic Product from 25% of GDP
in the mid 1980s to 12.3% of GDP in 2009 (CIA 2010) because of the rise of
industrialization and urbanization in the twentieth century agriculture is still the
largest sector of the Thai economy. The country is a leading exporter of crops such as
rice, corn, soybeans, sugarcane, tapioca, and rubber (USDA 2010). Approximately 50%
of the labor force is employed in the agricultural sector (GAIN 2010).
More than half of the cultivated area in Thailand is used for rice production.
Approximately 70 million ha or 53% of the total cultivated area was used for this purpose
in 2007 (OAE 2010). Geographically, rice can be grown under a wide range of
biophysical and climatic conditions from deep water (>80 cm height of water), to lowland
(50-100 cm height of water), to upland (<50 cm height of water) conditions. The two
major types of cultivation practices generally found in the country are rainfed rice and
irrigated rice. Rice cultivation in Thailand is dominated by rainfed lowland rice, which
occupies over 70% of the total area cultivated for rice and accounts for 80% of total rice
production. Irrigated rice occupies about 20% of the total area cultivated for rice and
accounts for about 20% of total production (OAE 2010). Normally, rainfed rice can be

14

grown only during the rainy season, but with irrigation farmers can grow rice two or three
times a year. Irrigated rice is produced mainly in the Central Plain (Shivakoti et al. 2005).
Similar to other Asian countries, the advent of high-yielding varieties (HYV) of
rice during the Green Revolution has significantly improved the quantity of rice
production; nonetheless, the benefit of these varieties is uneven. The adoption of HYV in
parallel with the construction of irrigation systems allowed farmers to grow multiple
crops and increased the productivity of rice (Ishii 1998; Molle and Keawkulaya 1998).
However, HYV have been criticized for their intensive production inputs, such as
fertilizers and pesticides, their susceptibility to local pests and diseases, their unsuitability
for rainfed areas, and their low quality, which has generated low market prices (Ishii,
1975; Molle and Keawkulaya 1998; Phelinas 2001). Therefore, HYV for Thailand have
had only minimal impact and limited growth in some areas (Molle and Keawkulaya
1998; Phelinas 2001). Moreover, the poor taste of HYV is not favored in the international
market, and production costs of HYV are higher compared to other rice varieties. Under
these circumstances, the Thai rice economy has focused on high-quality aromatic rice,
which receives higher price yet produces lower yields than HYV. Thailands agricultural
sector has decided to assert its comparative advantage in international trade by focusing
on the quality-based rice market rather than the quantity-based market (Phelinas 2001).

15

1.2.2 Climate and socioeconomic impacts


Rice production in Thailand has faced some constraints due to climatic and
socioeconomic factors. Even though the country is located in the tropics, rice production
is affected by variations in rainfall frequency, total rainfall during the growing period,
flooding, and mid-season dry spells (Bachelet 1992; Chinvanno et al. 2008). The direct
and indirect impacts from the changes in climate variation have been reported as the
major concerns on the current rain-fed rice production in Thailand. For example,
biophysical impacts (e.g. soil physical changes or flooding) are classified as the firstorder impacts from climate events. The consequences of the biophysical impacts in the
forms of damages to immature plants and reduction and losses in harvested yields are
classified as the second-order impacts. The human well-beings (e.g. household income,
financial and wealth, migration of household members, and labor force, etc.) are
classified in the higher-order impacts (see Chinvanno et al. 2008). In addition to the
current climate, previous climate change studies analyzing rice suggested that the
projected temperature increase may reduce yields in the region (Buddhaboon et al. 2008;
Felkner et al. 2009) and may shift the potential production areas towards the upper
central region of Thailand (Buddhaboon et al. 2008).
Recent research shows that there are increasing challenges to Thai rice production
because climate impacts occur in parallel with the expansion of urbanization,
industrialization, and population growth, all of which take land, water, and labor from the

16

agricultural sector (Kupkanchanakul 2000; Phelinas 2001). The expansion of urban areas
results in the competition for water among various sectors, and this competition continues
to increase across the nations regions. Historically, water extracted from northern
Thailand was diverted for domestic and agricultural use in the Central Plain (Ishii 1975).
In recent years, however, this flow has decreased significantly because a larger share has
gone to the upper and lower basins (Shivakoti et al. 2005). In addition to water resources,
labor shortages attributable to the diversion of the labor force from agricultural to
industrial sectors have constrained agricultural production. Agricultural sector demand
for wage labor exceeds the availability of the local supply (Ishii 1975; Johnson 1981;
Phelinas 2001).
The scarcity of production resources has increased the cost of rice production.
Due to an intensive demand for production inputs, deprivation of land, competition for
water, and scarcity of wage labor, the cost of production has increased and income is
expected to fluctuate. In a world market, developing countries usually set a low crop
price to compete with opponents (Shivakoti et al. 2005). As a result, profits are marginal
and farmers who depend primarily on the income from rice are highly sensitive to the
changes in market price. However, farmers who have enough capital have more options
and may decide to switch to other cash crops that are more profitable.
Moreover, economic globalization has resulted in more intensive use of fertilizers,
agricultural chemicals, irrigated water, and labor in order to increase rice yields to meet

17

the demands of the market. Additionally, in those areas that adopted HYV rice, Thailand
transitioned from a single-crop agricultural system to multiple cropping, which extracts
massive amounts of water from natural sources as well as irrigation (Shivakoti et al.
2005).
Thus, rice production in Thailand faces substantial challenges because of various
factors. Despite limited natural resources and declining arable land, labor, and water
supplies, Thailand needs to increase yields and lower the costs of production while
maintaining the grain quality widely expected in the world market. More important, tastes
have changed and demand more high-quality rice. Thailand must maintain the ability to
respond to increases in demand (Ishii 1975; Shivakoti et al. 2005). Moreover, in the
future, it is likely that there will be significant changes in agricultural practices, rural
society, the national economy, and the relationship between government and individual
economic sectors. Mechanisms are needed to improve the flexibility and capacity
required to deal with these stresses (Shivokoti et al. 2005). It is important to note that
agricultural development plans and strategies to address the stresses typically operate at
the national scale, but the impacts function at the local, household, and individual scales.
Currently there is varying ability and resources to deal with stress among farmers.
Farmers appear to be the first group affected by the negative impacts of climate and from
the changes of market policy, price, national or international demand, or limited access to
necessary production resources (Kupkanchanakul 2000; Parry et al. 2004). Some places
or persons will gain or lose depending on their capacity to cope with future changes.

18

Recent research has assessed impacts and vulnerabilities of rice production in order to
find agricultural and economic strategies that consider stakeholders at multiple scales,
including farmers at local scales.

1.3 Research Framework: Vulnerability and Scale


The concept of vulnerability has been used in a variety of research contexts to
refer to the degree to which a system is likely to be harmed by climate and other stresses.
The three major dimensions of vulnerability are exposure, sensitivity, and adaptive
capacity (Polsky et al. 2007). Exposure refers to the stresses caused by changes in
frequency, intensity and the nature of climate and non-climate stresses. Sensitivity refers
to the degree to which an individual or group (as the system of interest) is affected by
exposure to climate and other stresses. The ability of the system to respond to the
exposures and the effects in order to adjust to and cope with the impacts is referred to as
adaptive capacity (Kelly and Adger 2000; Fssel and Klein 2006).
The vulnerability framework in previous research recognizes the roles of both
climate and non-climate exposures and stressors that contribute to the vulnerability of a
system (Cutter 1996; Kelly and Adger 2000). Vulnerability research, especially in studies
of climate change, traditionally focused on climate variability and climate-related
exposures such as sea-level rise, flood, drought, and extreme events (e.g., Adger and
Kelly 1999; Rygel et al. 2006). The expansion towards social dimensions has been

19

featured in more recent literature. Social exposures include economic, policy-making,


social, environmental, technological, and other socioeconomic factors. (Moser 2010).
However, climate and social exposures have generally been considered functioning
independently of each other. This separately functioning, factor-driven vulnerability
research has been criticized for overlooking the real-world context in that vulnerability is
driven by the interaction of exposures rather than a single factor. Vulnerability studies
should consider this interaction among multiple stressors as a dynamic rather than a static
process (Adger and Kelly 1999; Turner et al. 2003; OBrien et al. 2004; Belliveau et al.
2006). One of the studies that emphasized the interacting processes between climate and
socioeconomic exposures was the double exposure framework of OBrien and Leichenko
(2000) introduced earlier.
The challenges underlying the examination of two global processes in multiple
exposures research rest on two major concerns the scale used in the analysis and the
definition and framework of vulnerability (Wilbanks and Kates 1999; Fssel and Klein
2006; Eriksen and Kelly 2007). Theoretically, scale matters in the study of global change,
because (1) the changes and consequences of the interaction across scales are
complicated to predict and understand, and (2) the interpretations of the results and the
impacts can mean something different at global and local levels (Wilbanks and Kates
1999; Kelly and Adger 2000; Wilbanks 2006).

20

Wilbanks (2006) argues that there are three reasons for focusing on the local,
detailed scale. First, complex interactions of key processes e.g., environmental,
economic, and social processes moving across time and areal extents and underlying
environmental systems are too complex to unravel at any scale beyond the local. This
perspective is supported by the work of Cutter (1996), Kasperson et al. (1995), Easterling
(1997), Wilbanks and Kates (1999), and Carlo and Tol (2002). The second reason is that
observed processes at a detailed scale contain more variance than observed processes at a
general scale, and the greater variety of observed processes and relationships at a local
scale can provide important knowledge about the substantive questions being asked
(Wilbanks 2006). Third, looking at a particular issue top-down can lead to significantly
different conclusions from researchers looking at that very same issue bottom-up
(Kasperson et al. 1995; Wilbanks 2006). ). However, research should consider the
importance of the linkages between different scales and the research questions being
asked (Easterling 1997; Wilbanks 2006).
This research will consider the three arguments presented by Wilbanks and,
although it will focus mainly on the local scale, it will also take into consideration the
linkages that exist between local and regional conditions. The research will also examine
the regional-to-national linkages that may influence rice production. Research conducted
by Easterling (1997) supports the approach taken in this study and its focus on the local
scale with linkages to regional scale. For example, the knowledge of dynamic processes
embedded in integrated regional assessment is often derived from the understanding

21

gained from location-specific field studies (Easterling 1997; Carlo and Tol 2002). In
another example, the efficacy of adaptation varies from place to place. In the short term,
flexibility in the use of agricultural practices and in the capital investment of individual
farmers and regional marketing systems influences adaptability. In the long term,
however, regional differences in rates of depreciation of capital investment are also
influential (Easterling 1997). Finally, socioeconomic and environmental data sets are
most likely to match best at relatively small spatial scales (Lonergan and Prudham 1994).
Although national and international linkages are important, the understanding of
the processes most often comes from in situ and regional experimentation (Easterling
1997; Carlo and Tol 2002). If a detailed global-scale approach were taken in this study,
the robustness of individual localities would tend to be overestimated because of the lack
of sensitivity to local obstacles and constraints (Wilbanks 2006). The spatial variability of
climate change would also be obscured. Given the fact that marginal systems are best
studied at the local to regional scales (Easterling 1997), such as would be the case with
Thailand, there is justification for the scalar decision taken in this study.
There is a need to clarify the terminology and conceptual framework used in
vulnerability studies because vulnerability means different things to different scholars.
According to Fssel (2007), OBrien et al. (2007) and Nelson et al. (2010a), the
terminology describes the dimensions of vulnerability, while the conceptual framework
defines the methodological approach in assessing the vulnerability. Empirical evidence

22

for hazard assessment and exposure is generally confined to floods, droughts, storms, or
other extreme events. For agricultural vulnerability assessment, exposure often refers to
changes in climate variability, such as temperature and rainfall variations, which
influences crop biophysical sensitivity, annual yield, agricultural land use changes, and
food security (Berry et al. 2006; Nelson et al. 2010b). Agricultural socioeconomic
research can focus on the sensitivity and adaptive capacity of market mechanisms,
international trade and policy, or the well-being of society.
Apart from the above discussion, vulnerability has been viewed in two other ways
end-point and starting-point vulnerability. The end-point vulnerability approach views
climate change as the root problem and initiates the analysis with attempts to establish the
future climate impacts and the potential adaptation options. In contrast, the starting-point
perspective considers social vulnerability the root problem and focuses on uncovering
current social vulnerability to climate before suggesting the effective adaptation options
(Fssel and Klein 2006; Eriksen and Kelly 2007; Fssel 2007).
According to OBrien et al. (2007), conceptual frameworks can be classified into
two major groups contextual (qualitative) vulnerability assessments and outcome
(quantitative) vulnerability assessments. The differences relate to the choice of
appropriate methodological designs. Outcome vulnerability, which is similar to the endpoint approach, focuses on the linear relationship between exposure and the projected
impacts of climate change on a specific exposure unit. Outcome vulnerability then

23

suggests adaptation options to reduce or limit the negative outcomes. Most of the
research studies in this category use impact models (e.g., crop, hydrologic, or economic
models) as analytical tools.
The contextual vulnerability model, which is similar to the starting-point
approach, considers processes of climate-society interactions as a robust exposure factor
that influences vulnerability. Important connotations of contextual vulnerability are: (1)
impacts are unevenly distributed over the exposure unit; (2) the exposure unit has
differential ability to respond to, adapt to and recover from the impacts it will experience;
and (3) the existing vulnerability of the exposure unit will also influence its capacity to
cope with future impacts (Wilbanks and Kates 1999; Kelly and Adger 2000; Eriksen and
Kelly 2007). Hence, identifying key indicators of existing vulnerability will enhance the
ability of investigators to understand the nature and characteristics of future vulnerability
(see Eriksen and Kelly 2007).
The literature agrees that failure to outline a clear definition and conceptual
framework in vulnerability assessment studies will result in a common methodological
fallacy, as described by Nelson et al. (2010a). This fallacy results from the
overwhelming use of biophysical or macroeconomic models in assessing and predicting
impacts over the starting-point research approach, which results in the drivers that cause
the vulnerability to be overlooked (OBrien et al. 2007; Nelson et al. 2010a). Moreover,
there is a need for any future vulnerability research to integrate both quantitative and

24

qualitative analyses to develop insights and results that are meaningful to users (Cutter
2003; Moser 2010; Nelson et al. 2010a). Both, the quantitative and qualitative approach
used in this study will be discussed in subsequent sections.

1.4 Research Goal, Questions, and Objectives


Emerging from the above, the goal of the study is to understand the spatially
distributed impacts and vulnerabilities of local rice production in Thailand resulting from
the double exposure to climate change and socioeconomic change. To reach this goal, the
research seeks to answer three questions and strategic objectives.

1.4.1 Research questions


1) Who will be vulnerable to double exposure and what are key indicators of
vulnerability of Thai rice production?
2) What are the key characteristics of places and agricultural practices that might
reduce the vulnerability of rice production to double exposure?
3) What are the consequences on rice production resulting from double exposure to
climate and socioeconomic change?

25

1.4.2 Research objectives


1) To identify the important climatic and socioeconomic indicators associated with
the vulnerability of Thai rice production, including the dynamic interactions
between climatic and socioeconomic indicators
2) To isolate the most influential indicators and distinguish the four least
vulnerable provinces
3) To model the range of sensitivities of rice crop yields to varying climatic and
socioeconomic scenarios

1.5 Study Area


Thailand is a Southeast Asian tropical country covering approximately 51 million
hectares. It shares borders with Myanmar, Laos, Cambodia, and Malaysia. The country
extends from 5 to 40 north latitude and 97 to 106 east longitude (Figure 1.1). The
monsoon dominates temperature and precipitation over Thailand, with a dry season
associated with the Northeast Monsoon and a wet season associated with the Southwest
Monsoon. From the beginning of November to February, except for the southernmost
portions of the country, the Northeast Monsoon brings cool and dry air from the Siberian
anticyclone to Thailand. The Southwest Monsoon, the main source of precipitation in
Thailand, brings humidity from the Indian Ocean for a rainy season that lasts from May
to October (Ratanopad and Kainz 2006; Chinvanno et al. 2008). The average annual

26

rainfall in most areas ranges from 1100-1500 mm, although rainfall totals up to 4,500 mm
are found along the southeastern coast and in peninsular Thailand. Average temperature
in Thailand varies from 24.429.3 C (7685 F).
Climate characteristics in Thailand fall into three major Kppen classification
groups: Aw, Am, and Cw. Despite overall dominance by the monsoon, the majority of
Thailand has a Tropical Savanna (Aw) climate, with the exception of the southeastern
coast and southern peninsular provinces where Tropical Monsoon (Am) predominate.
The northern mountainous area is categorized as Humid Subtropical (Cw).
Thailand has three main seasons. A rainy season from May to October during
this period the Southwest Monsoon brings a stream of warm moist air from the Indian
Ocean causing abundant rainfall. About 90 percent of the annual rainfall occurs during
this season. A cool dry season occurs from November to February, and warm weather
and variable wind is present in March and April. The warmest and coolest months during
the year are April and January, respectively (Attanandana and Kunaporn 2005).

27

Figure 1.1: Political and topographic map of Thailand (source: wikipedida.org)

Thailand is divided politically into 76 provinces situated in six physiographic


regions northern, central, northeastern, western, southern, and eastern regions divided
by attributions of Thailands physical setting. Only the northern, central, and northeastern
regions (hereafter referred to as the North, Central Plain, and Northeast) are considered as
potentially suitable rice-producing environments (Buddhaboon et al. 2008); these three
regions and their 62 provinces will be the focus of the first phase in this study.
The North is characterized by high mountains with steep river valleys and upland
areas that border the Central Plain. Some upland rice is grown in the high areas and at the

28

lower slopes of the high hills. Lowland rice is grown mainly in the lower valleys in which
water is available. About 22% of Thailands rice area is in the North, which account for
approximately 25% of total rice production. Major rivers in the North, the Ping, Wang,
Yom, and Nan, flow and unite to form the Chao Phraya River and tributary network in
the lowland Central Plain; all rivers drain to the Gulf of Thailand (Shivakoti et al. 2005).
In the Central Plain, the Chao Phraya drainage system occupies about one-third of
the nations territory. This region, known as the rice bowl, contains fertile soil suitable
for paddy rice cultivation. Central Plain wet-season rice occupies about 21% of the
countrys total cultivated rice area and produces 30% of total rice. About one fourth
(450,000 ha) of that cultivated land has irrigated dry-season rice (OAE 2010). Because
the Bangkok Metropolitan Area is situated on the southern portion of the Central Plain,
this region is a national hub for trade, transport and industrial activity, as well as for
major irrigation development projects (Ishii 1975; Shivakoti et al. 2005).
The Northeast consists mainly of the dry Khorat Plateau where some parts are
extremely flat with a few low, rugged, and rocky hills. The Phetchabun, Sankambeng,
and Dong Phaya Yen mountains separate the Northeast from the rest of Thailand. The
Mekong River delineates much of the northern and eastern rim and drains into the South
China Sea. The Northeast is known for its infertile soil with high salinity and poor
drainage and its tendency for drought due to a long dry season; both of these factors do
not favor agricultural activities. However, rice cultivation is possible as the short
monsoon season brings enough rainfall to harvest two-crop cycles per year. Rice

29
occupies 80% of the regions arable land and about 53% of Thailands rice-producing
land is in the Northeast, but the region only accounts for 41% of the nations total rice
production (OAE 2010). The Northeast produces mostly rainfed rice. Rice farmers in this
region are always confronted with the risk of uncertain production due to floods in the
rainy season and water shortages in the dry season (Ishii 1975; Shivakoti et al. 2005).
There is evidence showing that climate change and socioeconomic change will
significantly affect rice production in Thailand. Climate change could influence the
monsoon and subsequently alter the intensity of both temperature and precipitation in
various areas (Kripalani et al. 1995; Mitchell and Hulme 1999; IPCC 2007). Projections
for Thailand show a significant increase in extreme climate events that could occur in the
form of high temperatures, heavy rainfall, and flooding (Chinvanno et al. 2008; Cruz et
al. 2007). The scarcities of land, labor, water, etc. mentioned earlier are the major
challenges resulting from socioeconomic change. Research from many disciplines is
needed to determine how Thailand could increase yields to meet the future demands
while maintaining high grain quality, increasing labor productivity per land area,
increasing farmers incomes, and developing the water-saving and related technologies
that could overcome climatic disturbances (Shivakoti et al. 2005; Bouman et al. 2007)

30

1.6 Scope of the Study


This study will not analyze adaptation strategies because adaptation analysis is
complicated and when combined with the analyses used here would require much
more time than is available for this research. Nonetheless, some adaptation possibilities
will be suggested in this study through the critical adaptive capacities developed for Thai
rice production to cope with future impacts of climatic and socioeconomic changes.
These adaptive capacities are important for developing meaningful adaptation strategies,
policies, and fundamental understanding of place-specific agro-climatic problems.
The study will focus primarily on analysis at the local scale, although the
interactions of climatic and socioeconomic factors that influence Thai rice production
involve four different scales (local, regional, national, and international). As mentioned in
the previous section, the interactions of climatic and socioeconomic factors, especially in
the agricultural context, are too complex to unravel, and the processes and patterns of
relationship may not be well observed beyond the local scale. Moreover, in the
agricultural context, crop producers are usually the first group that experiences or suffers
from climatic and socioeconomic impacts; the local-scale focus will reveal important
information that points to place-specific conditions. However, the linkages existing
between scales (local-regional, and regional-national) will be recognized in the study to
suggest more meaningful and realistic adaptation possibilities.

31

1.7 Thesis Overview


The remainder of this disseration will be structured as follows. Chapter 2
provides the methodology and data for this study and details the Sequential Exploratory
Strategy, a mix-method approach with three phases of analysis. This chapter also
highlights the first phase of analysis, which uses a Vulnerability Scoping Diagram (VSD)
to structure the proxy data. Chapter 3 provides the results of the second phase analysis. In
this phase, I conduct a Principal Component Analysis (PCA) to distinguish the variables
that contribute to the three dimensions of vulnerability: exposure, sensitivity, and
adaptive capacity. Chapter 4 extends the results of the second phase by presenting
vulnerability maps and using them to explore the vulnerability patterns of individual Thai
provinces. Chapter 5 shows results of the third phase of analysis, which uses the EPIC
crop model to explore the impacts of future and projected climate change on Thai rice
production without adaptation, whereas Chapter 6 discusses the plausible impacts of
climate change with adaptation. Chapter 7 discusses the findings of the three-phase
analysis and also draws conclusions on the potential resilience of future Thai rice
production.

32

CHAPTER 2
METHODS
This research adopted a Sequential Exploratory Strategy type of mix-method
approach (Creswell 2009). In a typical Sequential Exploratory Strategy, the research
implements two phases of analysis. The first phase employs a qualitative framework to
explore and inform the selection of data and the second phase uses a quantitative
framework to analyze the selected data. This research differed in that there were three
phases of analysis: an exploratory qualitative analysis, an exploratory quantitative
analysis based on the qualitative analysis, and a quantitative modeling study based on the
exploratory quantitative analysis.
Specifically, the exploratory qualitative analysis structured the proxy data that
represented climatic and socioeconomic-related indicators influencing Thai rice
production. The result of this first phase of research generated the input that allowed the
measuring and comparing of vulnerability components among production areas in
Thailand in the second phase. The second phase employed a Principal Components
Analysis (PCA) to identify key vulnerability indicators, distinguish four provinces (the
case study areas) that are likely to succeed in the face of an evolving climatic and
socioeconomic system, and demonstrate how rice production might be affected by future
projected climate conditions. The four case study areas distinguished by the PCA formed
the basis of the third phase of analysis: crop modeling.

33

The following sections provide details of each phase of analysis. Phases 1 and 2
encompass the qualitative analysis and PCA, respectively, and Phase 3 covers the EPIC
crop model used in this dissertation.

2.1 Phase 1: The Vulnerability Scoping Diagram


The first phase of analysis employed a Vulnerability Scoping Diagram (VSD) to
develop a social vulnerability profile for rice production. Polsky et al. (2007) designed
the VSD with three rings circling around a bullseye (Figure 2.1). The bullseye represents
the concept of vulnerability. The first and nearest ring represents the three dimensions of
vulnerability discussed earlier in this dissertation: exposure, sensitivity, and adaptive
capacity. The middle ring represents the components of these three dimensions. Finally,
the outer ring represents the measurements of the components. The VSD offers two major
functions for a vulnerability assessment, providing a starting point for researchers to
understand the details of vulnerability, and facilitating the comparisons of vulnerability
indicators at different places and times.
Adopting the VSD also facilitated the processes of data collection, the
development of conceptual frameworks, and the evolution of a methodological
framework that suited the collected data. For instance, Pearsall (2009) adopted the VSD
to investigate vulnerabilities of residents and communities to multiple stressesthe
consequences of environmental mitigation projects and regional hazardsat four study
areas in New York City. The study showed that the VSD could practically monitor

34

vulnerabilities to multiple stressors and provide better understanding of the linkages


among vulnerability dimensions in a complex human-environmental system.

Figure 2.1: Vulnerability Scoping Diagram (source: Polsky et al. 2007)

35

2.1.1.1 Physical vulnerability


Indicator 1: Climate variables (temperature and precipitation)
Climate variables are mostly defined as the main factor affecting production
yields and cultivated area (as the first-order impacts) and socioeconomic activities and
well-being (as the second or higher-order impacts) of farm households (Kates et al. 1985;
Parry et al. 1985; Dabi et al. 2008). Some research pays particular attention to rainfall
variations because rice is often cultivated under rainfed conditions. Deviations of rainfall
distribution from normal could change yields from the expected and consequently affect
farm incomes, benefits, practices, and so on. Two types of climate events droughts and
floods are the major concerns of farmers in most developing countries (Dabi et al.
2008). For example, the occurrence of prolonged dry spells during mid-season after
sowing or transplanting rice could delay farm schedules and impose additional costs on
farmers. Flooding that coincides with harvest could cause severe damage to produce at a
time when replanting may be too late. Therefore, farmers who depend on rainfed
cultivation are vulnerable (Chinvanno et al. 2008). Temperature stress, especially during
the growing season, potentially affects crop growth and functioning. Specifically,
temperature stresses can affect crop physiological process by decreasing dry matter
accumulation, influencing productive tillers, reducing grain weight, and increasing floret
sterility (Manju et al. 2010). As a result, crop yield and quality are lower than expected.

36

2.1.2.1 Socioeconomic vulnerability

2.1.2.1.1 Human capital


Indicator 2: Education. The educational attainment of household members can reflect
their vulnerability to climate and economic stresses in two ways. According to Adejuwon
(2008) and Dabi et al. (2008), the households investment in higher education, on the one
hand, can result in good health, labor productivity, and the agricultural information
accessibility because educated household members can understand and participate in the
technological and administrative processes in the modern economy better than members
with little or no formal education.
On the other hand, household members receiving high levels of education
characterize high mobility and flexibility. Huffman (2001) points out that farm household
members who receive higher education often choose off-farm employment because of
higher wage incomes and the perception of less physical work compared to farm work,
which can lead to permanent migration from the farm (Huffman 2001). Although the loss
of labor can hurt the household, the absence of household members due to the off-farm
employment does not necessarily indicate high household vulnerability. Instead,
remittance of wage income from off-farm employment helps secure and diversify
household income. As a result, this alternative source of household incomes helps
decrease reliance on farm production and income driven by the climate variation
(Huffmann 2001; Phelinas 2001; Dabi et al. 2008).

37

Indicator 3: Farm labor. Urbanization and industrialization can cause a shift of farm
labor from rural areas to urban areas for employment or educational opportunities. The
average number of farm household members in Thailand is approximately five people per
household, which is sufficient to supply family laborers for a farm of less than 50 ha.
During the peak season, developing country farmers often complement the absent
household members with the hired laborers to achieve production goals (Morgan and
Munton 1971; Phelinas 2001). However, small farm households may experience hardship
if farm outputs do not generate enough income to meet the labor costs per unit of land
(Morgan and Munton 1971).

2.1.2.1.2 Financial capital


Indicator 4: Household incomes and income sources. Farm household incomes and
income sources can serve as measures of the vulnerability of their production. Dabi et al.
(2008) suggests that households with low incomes, savings, and saleable assets are
generally vulnerable to stresses from climate variability and socioeconomic changes.
With low financial status, the ability of farmers to invest in farm improvements, e.g., the
purchase of farm inputs, farm equipment, and other farm technology, is limited.
Furthermore, low financial status reduces the capacity and ability of farm households
because poor households are likely to focus on the survival and well being of household
members rather than the improvement of farm quality or the production system, which
would decrease their vulnerability to future climate and socioeconomic stresses (Osman-

38

Elasha and Sanjak 2008). Low-income household vulnerability is worst when it relies
exclusively on agricultural production for income and food source (Morgan and Munton
1971).
Indicator 5: Size of farm operation. The size of farm operation (i.e., farmland plus farm
equipment) can measure farm household vulnerability. Eakin et al. (2008) suggest that
both large and small landholdings are sensitive to the variety of climate events; however,
the overall social vulnerability of small landholdings is higher. Large landholdings
represent higher wealth and financial status of the households; they can invest more on
farm production and generate greater yields and incomes than smaller landholdings. With
more access to physical and material resources, large landholdings have greater flexibility
and more stable financial status, which in turn increase their capacity to cope with a
changing economy and environment. Similar to farm size, the size and ownership of
animal units and tractors also indicate the production scale and financial capital of farm
households (Huffman 2001).
Indicator 6: Land ownership and tenure security. Land ownership and land tenure can
determine the ability of farm households to generate food, income, and social and
financial status. According to Deininger and Feder (2001), with land ownership and
tenure, farmers gain the opportunity to obtain financial credits and loans from banks to
invest in the farm. It is the opposite for households that lack ownership and tenure:
farmers have less access to formal banks and rely on non-formal financial institute or do
without investment. With less accessibility to funds, farmers tend to use fewer farm

39

inputs such as fertilizer, pesticide, and insecticide, which may result in relatively lower
yields. Farmers without ownership or tenure are less likely to find room in their tight
budgets to improve the land (Deininger and Feder 2001).

2.1.2.1.3 Social capital


Indicator 7: Governmental support. Social capital and social networks within the
community also determine the vulnerability of farm households. Government support in
the form of production policies and agricultural extension services could reduce this
sensitivity and increase the adaptive capacity of the farm households. For example,
extension units could support adaptive capacity by introducing new farm strategies and
developing necessary skills and knowledge to overcome climate stresses (Dabi et al.
2008). Supporting strategies and policies from government, such as research and
development, education, infrastructure and facilities, and information, could also help
increase adaptive capacity of farm households. Chavas (2001) found that government
policies promoting the use of crop price insurance, farm subsidies, production contracts,
disaster payments help reduce the adverse effects from decreases in crop price and
uncertainties in crop production due to climate and socioeconomic stresses. Farm
households participating in such government programs are more likely to be buffered
against production risks.
Indicator 8: Market channel. Market institutions and their structures within the local area
indicate the strengths and weaknesses of the domestic farm production system. According

40

to Beininger (2001) and Barrett and Mutambatsere (2005), agricultural markets provide
fundamental functions for agricultural input and output distribution, and post-harvest
processing and storage. Basically, farmers purchase farm inputs (e.g., fertilizer, seed, and
machinery), sell their products, and earn incomes back from the agricultural market.
However, the efficiencies of market institutions, physical infrastructure, trading
competition efficiency, and market accessibility to farmers in each local area are unequal,
particularly in developing countries. Agricultural communities with poor communication
and poor transportation systems are less flexible and more sensitive to constraints from
climatic and social stressors. Nonetheless, the formation of local markets and communal
marketing in the form of credit unions, farmer cooperatives, and wholesale-level
cooperatives increases the capabilities of local farmers by facilitating bulk input
procurement, negotiating price, and sharing transportation costs. The cooperation of
farmers also increases their competitiveness and negotiating power relative to
commercial markets.

2.1.2.1.4 Physical capital


Indicator 9: Basic infrastructure and services. The availability and accessibility of basic
production resources can determine the coping capacity of farm households. For rice
cultivation, deep wells, effective water pumps, or well-developed irrigation systems are
vital because these resources can provide water for agricultural and household use when
water becomes scarce during dry periods. In addition, basic infrastructure and facilities

41

such as road, electricity, and telephone located within the accessible distance can improve
and smooth the production process. For example, a well-conditioned road and short
distance between the farm and the market place could reduce delivery time, which could
also minimize yield-quality losses. Good roads also enable large pieces of farm
equipment such as tractors and trucks to move from field to field with ease. Similarly,
electricity improves the quality of farmers lives and creates a foundation for modern
farming.
In addition to basic infrastructure, agricultural facilities and services influence the
overall vulnerability of farm households. Helslop-Thomas et al. (2008) suggest that a
healthcare center located within or conveniently accessible to a community could help
facilitate improve response during an emergency. Inadequate healthcare is associated
with poverty, greater sensitivity in individuals and households, and poor capacity of
households to cope with climate-induced stresses and is related to low farm yields
(Wandiga et al. 2008).

2.1.2.1.5 Natural capital


Indicator 10: Land and water quantity and quality. The expansion of urbanization and
population draws upon the natural resources of the agricultural sector. Existing farmland
is shrinking and demand and prices for that land is skyrocketing, especially in fertile
areas with irrigation systems. Competition for water between domestic and agricultural is
increasing (Phelinas 2001).

42

Consequently, the price of fertile farmland with irrigation systems is too


expensive for small landholders, pushing poorer farmers into more marginal areas. Farm
households with limited financial resources are vulnerable if they are forced onto lands
with low fertility. Improving this lands quality for crop production requires expensive
modifications and inputs. Furthermore, marginal areas are likely to have limited access to
water sources and are far from irrigation systems. With poor land and insufficient water,
yields are low compared to those of fertile production areas with good access to irrigation
systems (Osman-Elasha and Sanjak 2008).

2.1.3 Developing the VSD


Figure 2.2 illustrates the potential indicators of farm household vulnerability to
climate and socioeconomic changes discussed above. The indicators are specified based
on the findings of AIACC research which guidelines some indicators that influence the
vulnerability of the farm households. The exposure unit focuses on the concept of double
exposure discussed in the literature review in this case, changes in the biophysical and
socioeconomic factors that may cause stress to farm households. Among possible
biophysical factors, geographical and environmental variables are useful for assessing
vulnerability (Berry et al. 2006; Leary et al. 2008a; 2008b; Nelson et al. 2010b).
Sensitivity and adaptive capacity, respectively, refer to socioeconomic
characteristics of farm households that determine the response of farm households to the

43

Figure 2.2: Vulnerability Scoping Diagram for the indicators of rice farm household
vulnerability

44

impacts of double exposures and the ability of the households to overcome their exposure
or socioeconomic sensitivities.As mentioned earlier, Osman-Elasha and Sanjak (2008)
present five categories of capitals that determine the livelihood and the coping capacity
of farm households. The indicators can occur repeatedly in the VSD because some
indicators can indicate more than a single dimension of the household vulnerability
(Polsky et al. 2007). For example, financial capital can measure sensitivity of farm
households to climate and socioeconomic stresses; at the same time, it can indicate the
capacity of the farm households to overcome the biophysical and socioeconomic stresses
today and in the future. The specific proxy variables in association with the indicators
listed in the previous section are shown in table 2.1 to clarify the indicators and potential
proxy variables collected for the PCA analysis. Note that the choices of proxies
emphasize contemporary sensitivity and adaptive capacity and do not necessarily reflect
future sensitivity and adaptive capacity concurrent with climate change.
The indicators denote biophysical and socioeconomic factors that influence the
capability and efficacy of farm households to prepare for, respond to, recover from, and
reconstruct their production from climatic and socioeconomic changes and, at the same
time, perhaps improve their livelihoods. In the next section, I provide more detail about
preparation of the proxy dataset based on the ten major indicators derived from the
AIACC research (Berry et al. 2006; Leary et al. 2008a; 2008b; Nelson et al. 2010b). The
list of proxy variables and their sources is provided in Appendix A.

45

Table 2.1: List of proxy variables for vulnerability indicators


Dimensions

Components (Indicators)

Measures (Proxy variables)

Exposure

Temperature

Monthly maximum temperature


Monthly minimum temperature
Experience of farm households production due
to heat stress
Monthly rainfall
Monthly SPI
Experience of farm households production due
to flooding
Soil quality and farmland zone
Quality of soil water
Household composition and household size
Available family labor
Migration of household members
Hired labor
Household employment status

Precipitation

Land and water


Sensitivity

Farm labor

Household incomes &


sources of income

Size of farm operation

Land ownership & tenure

Adaptive
capacity

Basic infrastructure &


services

Government support &


community network

Market channel
Financial status
Education

Annual farm household incomes


Proportion of agricultural income/nonagricultural incomes
Dependency on agricultural incomes
Landholding by size
Frequency of cultivation
Scale of farm equipment
Irrigation system / crop water facility
Type of landownership and land rights
document
Communication means
Accessibility to electricity
Accessibility to road & conditions
Farm facilities and services in the community
Scale of farm facilities
Accessibility and availability to water systems
Accessibility to agricultural knowledge transfer
Farm practices
Health services
Community requests for governmental support
Scales of agricultural market
Marketing institution
Scale of farm equipment
Households financial status
Levels of education
Literacy

46

2.1.4 VSD input data


In association with the indicators mentioned in the previous section, three groups
of data were collected climate, biophysical, and socioeconomic variables that were
available at the provincial scales. The proxy data were collected from various secondary
sources, such as the Meteorological Department, National Statistical Office, Community
Development Department, Office of Agricultural Economics, and Land Development
Department. However, because these data were collected from various sources with
varying details in those data, it is possible that they could appear repeatedly in the
analysis in closely related forms. Moreover, any one proxy could indicate sensitivity and
adaptive capacity of multiple production systems. Nonetheless, all of these potentially
collinear proxies were left for the variable reduction process of PCA to handle (e.g.,
Yarnal 1993).

2.1.4.1 Climatic variables: temperature, rainfall and SPI


The climate data used in this research spanned the years 1971-2009 and were
recorded by the Thai Meteorological Department. The data were comprised of daily
maximum temperature, minimum temperature, and daily rainfall. Monthly total rainfall
and average minimum and maximum temperatures were calculated from the daily records
and used as proxies for exposure. These data were primary inputs required for this
research; however, they were available in only 30 of the 62 provinces of interest. This

47

research, therefore, needed to employ an interpolation technique to create spatially


continuous climate data covering the missing provinces.
Several spatial interpolation methods are available such as kriging (a stochastic
method), Inverse Distance Weighting (IDW), and spline interpolation (a deterministic
method) (OSullivan and Unwin 2003). The interpolation techniques could predict or
estimate the values of variables of the unmonitored provinces from the known values.
Each interpolation method, however, would provide slightly different results due to the
different algorithms used by each method. OSullivan and Unwin (2003) and Childs
(2004) discuss general ways to select a suitable interpolation method. Specific to climate
data, Lanza et al. (2001) found that areal rainfall and temperature could be adequately
predicted using an appropriate linear or nonlinear stochastic technique. Kriging is a
stochastic linear interpolation that has been used widely in meteorology and hydrology.
Kriging assumes that the distribution patterns or directions between observed values are
spatially correlated (OSullivan and Unwin 2003; Child 2004). Ayanlade and Odekunle
(2009) used kriging to estimate rainfall values and found this technique adequate in
capturing spatial variations in climate data for unmonitored weather stations. Moreover,
as Kriging allows users to derive weights, the estimated results become more optimized
and unbiased. Consequently, it is possible to reduce error variances and minimize the
mean of precipitation errors (Chhetri 2007). Therefore, following previous research, this
study used ordinary Kriging to create continuous rainfall and temperature data across
Thailand.

48

To create climate data in provinces without climate stations, monthly rainfall,


maximum temperature, and minimum temperature data were interpolated using observed
climate records from the 30 weather stations in Thailand as the control data points. First,
daily data were converted to monthly data and monthly mean precipitation, maximum
temperature, and minimum temperature were calculated for each month before the
interpolation was computed using the Kriging functions in ArcGIS 9.3. The kriging
output took the form of raster maps showing the distribution of monthly climate data
across the country. The mean monthly climate values were then retrieved for each
province and were used as the input for subsequent analysis.
In addition to rainfall and temperature data, a standardized precipitation index
(SPI) could be used to indicate agroclimatic conditions. SPI expresses the deviation of
precipitation from the long-term mean and is commonly used as an indicator of drought
conditions as well as of anomalous rainfall during wet periods (Mckee et al. 1993;
Wattanakij et al. 2006; Kumar et al. 2009). One empirical study used SPI values to
predict crop water availability and crop yields on a rainfed farm (Yamoah et al. 2000).
The authors demonstrated that negative values of SPI during a growing season indicated
drought stress and showed a correlation with yield reduction and poor crop biophysical
responses. The study also suggested that the SPI values could perform better in
determining crop yields compared to total rainfall variables, especially in regions where
precipitation distributions are abnormally skewed.

49

Calculating SPI values required long-term monthly rainfall time series for all 54
provinces. This research developed seasonal (3-month and 6-month) and annual (12month) total SPI by computing a two-parameter gamma distribution from total
precipitation totals, and then expressing the values of SPI in a standardized normal
distribution.1 Negative SPI values indicate the lower below normal precipitation and
drought conditions), whereas positive values indicate above normal precipitation and
flood conditions (see Mckee et al. 1993; Kumar et al. 2009). The major constraint in
obtaining the SPI values was the availability of continuous records of observed climate
data at the 30 stations. The climate data were discontinuous over time in some provinces,
so this research employed the Kriging interpolation techniques to estimate seasonal SPI
values for those times and provinces where the data were missing. The process for
creating SPI through Kriging interpolation was similar to the process used for rainfall and
temperature interpolation described earlier.
Note that the amount of irrigated water was one of the most desirable factors to
include in the analysis because it captures the minimum availability of crop water during
the growing season as well as changes in water allocation. These data not only can
indicate an alternative source of water for crop production, but also higher demand in
irrigated water can indicate increased pressure for water-use efficiency in crop production
(DPI 2011). However, Thai government agencies do not collect water data at the

Note that precipitation is rarely distributed normally, so climate scientists use the gamma distribution to
transform the data into a pseudo-normal distribution.

50

provincial level and available data are too generalized to specify water use for only
agriculture. Therefore, it was not possible to include the quantity of irrigation water in the
analysis.

2.1.4.2 Other biophysical data


In 2000, Thailands Land and Development Department (LDD) investigated land
suitability for agricultural production especially rice production as characterized by
soil types. LDD classified the suitability of land for growing rice into four levels, going
from most suitable to unsuitable: L1, L2, L3, and Loc (not suitable). According to
Chhetri (2007), the differences in land conditions indicate the different choices of
agricultural practices, technologies, and rice varieties to be considered. For example, an
L1 zone where clay soil is dominant may require less fertilizer and other farm inputs to
maximize crop yields in comparison to the less suitable L2 and L3 zones. Moreover,
these land zones influence farm investment and farm adjustment requirements placed on
landowners. Because they have significant advantages in producing crops, farmers who
occupy land in the L1 zone would be likely to spend less in improving and adjusting the
land condition.

51

2.1.4.3 Socioeconomic proxies


As mentioned earlier, decisions on which socioeconomic proxies to use followed
the AIACC framework in Leary et al. (2008a and 2008b). Proxies representing farm and
social vulnerability and adaptive capacity such as land, labor, capital, production system,
market, etc., were drawn from various secondary sources such as the National Statistical
Office, the Community Development Department, the Office of Agricultural Economics,
and the Land Development Department. This research used proxies at the provincial scale
in order to match the spatial scales of the climate and biophysical data. The
socioeconomic proxies used in this research are listed in Appendix A.
In addition to the concern about spatial scale, this research was also cognizant of
the need for consistency in the temporal scale of the socioeconomic data. Because these
data came from various sources, it was difficult to obtain data consistently for the same
period. Consequently, to produce the greatest possible temporal consistency in the
socioeconomic proxy dataset, this research used the most recent secondary data available
at the provincial scale. For example, most economic and social data were surveyed by for
the Rural Development Information Center, the Community Development Department of
the Ministry of Interior in 2009 (RDIC 2011). CDD collects and updates these data every
two years and includes all of the approximately 71,000 villages across the country. The
data are classified at different levels going from village (the smallest), to district,
province, and region (the largest). Examples of items in the CDD dataset are general
information of household, economic, education, health, labor work, public and

52

community service, transportation and communication, natural resource, community


problem, environment and participation and strengthened community.
Similar to the CDD dataset, the 5th agricultural census from the National
Statistical Office (NSO) was available for 2008. These data are collected by interviewing
agriculturalists across the country every five years. Despite being a bit outdated, the 2008
NSO dataset contains specific data for both rainfed and irrigated rice farms that
supplement the CDD data. Examples of items in the NSO dataset are farm household
characteristics (i.e., gender, educational level, and ownership), farm assets and facilities,
and farm community concerns.
The subdivision of the NSO dataset provides detail of the range of agricultural
household income, the expenditure and the income distribution that could indicate the
sensitivity and adaptive capacity of farmers. These dataset are collected by local
statistical project of the NSO every two years from 1996-2008; however, the depth of
questionnaire in surveying data each time may specifically different based on user
requirement and available survey budget. The desired agricultural income dataset at
provincial scale is not available for the latest year in 2008 but only in 2004 (NSO 2010).
Therefore, the 2004 agricultural economic datasets were integrated based on the
assumption that farm incomes between 2004 and 2008 were not significantly different. It
was also assumed that any gap in incomes would not make significant changes in the
degree of farmers sensitivity and ability to adapt to stresses on production. In short, the

53

NSO datasets were used together with the data from the CDD to represent sensitivity and
adaptive capacity.

2.2 Phase 2: The Principal Component Analysis


The second phase of the research focused on identifying key vulnerability
indicators, interpreting the results, and ranking and selecting potential production areas as
case studies. This phase used principal components analysis (PCA), a mathematical
procedure that has been widely used to seek patterns of common variation among
observed variables, to reduce a large number of observed variables to a smaller number
of components, to provide an operational definition for an underlying process by using
observed variables, and to test a theory about the nature of underlying processes
(Tabachnic and Fidell 2007). Work on vulnerability assessment has employed PCA and
found that this technique provided insightful results to guide decision-making. For
example, Wagner et al. (2006) used PCA to identify key stressors and resources
influencing Mid-Atlantic region ecosystems. Using PCA, associations between key
stressors and resources could be compared quantitatively, highlighting only key
indicators out of the large proxy dataset. Then, the ranked indicators could pinpoint the
areas in the Mid-Atlantic region that were likely to be most vulnerable to multiple
stresses under a 2020 projection.
Here, PCA was used to minimize the number of climate and socioeconomic
variables and to identify a set of variables that underlined key vulnerability concepts of

54
each dimension exposure, sensitivity, and adaptive capacity in the VSD. The PCA
started by arranging the proxy data of each component and each province into a
spreadsheet and then transferring the data to a statistical package, using Minitab 10 and
SPSS 16.0 to perform the analysis. Then the PCA followed the general steps in this kind
of analysis, which included selecting and measuring a set of variables, preparing the
correlation matrix, extracting a set of components from the correlation matrix,
determining the number of components, rotating the components to increase
interpretability, and interpreting the results (see Tabachnic and Fidell 2006; Hair et al.
2009).
It is important to discuss some of the key operator decisions made during the PCA
that were unique to this analysis, specifically regarding number of components to retain,
the loading values used to interpret and name the components, and the use of component
scores to identify the relative vulnerability of each province. Based on a survey of best
practice in PCA, I concluded that I would use scree plots to guide the decision on the
number of components to retain in the analysis. In addition, I would only retain
eigenvalues if they were greater than 1.0 and if communalities (after extraction) were
greater than 0.7. After deciding on the numbers of components to retain, I then named the
components based on the structure of the proxy data that made up the components. Hair
et al. (2009) suggested that, for a sample size of 50, the proxies or variables should have
component loading at least 0.722 to be considered statistically significant; the value of
component loading could be lower if the sample sizes increased. For this research, the

55

unit of sample was the province and the proxies from 62 provinces were of interest. Note
that the data were originally collected at the village scale and included data from more
than 60 thousand villages, suggesting that the provincial data are robust. I therefore
concluded that the sample sizes were large enough to provide meaningful, significant
results with component loadings of 0.6; that is, I used variables equal to or greater than
0.6 to express the nature of the components and ultimately to give them their names.
Component scores were then used to measure the degree of vulnerability of each
province and to rank the most and least vulnerable provinces across the country. First I
computed the standard deviation (Std) values of the component scores to categorize the
different degrees of vulnerability for each component and each province. I then mapped
the Std values of each component and each province in ArcGIS 9.3 to identify patterns of
vulnerability across the country and to highlight the areas judged to be most and least
vulnerable. Finally, I selected four provinces, including one each from Thailands North
region, upper Central Plains, lower Central Plain, and Northeastern region, that had the
highest ranks (least vulnerability) as the case studies for the crop modeling analysis in the
third phase of the study.

2.3 Phase 3: Crop Model


The third phase of analysis aims to answer the research question on the ranges of
rice-yield sensitivities to changing climate and socioeconomic conditions. The analysis of
crop sensitivities is heavily based on process-based modeling, which takes into account

56

the complex interactions of physical/biological processes (e.g., crop, soil, and weather)
and farm management factors. Relevant to this dissertation, crop modeling has been used
to access crop responses to increased carbon dioxide and temperature, reduced crop
water, and adjusted farm practices (Easterling et al. 1992; 2003; Brown and Rosenberg
1997; Mearns et al. 1999). These studies combined process-based crop modeling with
projections of future climate, population, and trade to estimate the overall impact of
future production systems and further inform adaptation strategies for policy decision
making (Brown and Rosenberg 1999; Rtter et al. 2011).
Two process-based approaches, the EPIC (Erosion Productivity Impact
Calculator) and CERES (Crop Estimation through Resource and Environment Synthesis)
models, are well-known for their abilities to simulate cereal crops, particularly rice,
cultivated under varying production environments (Brown and Rosenberg 1997; Mearns
et al. 1999; Timsina and Humphreys 2005; White et al. 2011). Both models employ light
interception of photosynthetically active radiation (PAR) to determine the photosynthesis
of crops; each uses similar multi-layers of soil structure as an input parameter. Both
models use similar methods to calculate basic quantities such as crop biomass (Mearns et
al. 1999). Despite these similarities, each model can generate different outcomes.
What makes these two models substantially different is how they generate the
simulated yields. EPIC is a source-limited model, whereas, CERES is a source and
sink-limited model (Saseendran et al. 1998; Mearns et al. 1999). To simulate yields,
EPIC formulates the potential yields based on the accumulation of actual aboveground

57

biomass and a harvest index. The potential yield is determined by water, temperature,
aeration, or nutrient factors and the decreased yield is triggered by the reduction of a
harvest index parameter to represent the stresses of the production factors mentioned
above. EPIC uses only a single formula to simulate yields for all crops, but sets unique
parameter values for each individual crop: the harvest index, optimal temperature for
crop growth, leaf area index value, LAI development curve points, and more (Easterling
et al. 2003; Niu et al. 2009). Because of this simplicity, the model is flexible and well
suited to simulate the alternative crop rotations, tillage practices, planting dates, irrigation
and fertilizer strategies, and crop growing season lengths that are generally applied in
studies of climate change (Brown and Rosenberg 1997; Easterling et al. 1992), including
this dissertation.
Different from EPIC, the yield formulation of CERES takes into an account
aboveground biomass of crops and is more specific regarding crop phenological
processes. The actual biomass in CERES is a function of the potential biomass
decremented by water moisture or temperature stresses, but the calculation is made
separately for different periods of plant growth stage and plant parts (leaves, stems, grain,
etc.). In contrast, the final biomass from EPIC results from the accumulated heat units for
the whole plant required from the sowing to the maturity stages. Hence, in CERES, the
model requires the number of days needed for accumulation of grain dry matter at each
growing stage, such as emergence, flowering, and grain filling, until harvest (Mearns et
al. 1999, Iglesias 2006). As a consequence, the model requires a great detail of input

58

parameters and is phenologically stage specific. CERES model is considered more


advanced than EPIC and could generate more specific and useful outcomes (Huffman
2001).
Despite the advantages of CERES, I adopted the simpler EPIC model for two
reasons: (1) the availability of input data required for operating the model, and (2) the
model compatibility with my research design. As noted, CERES relies on considerable
input data, and most of these data are based on laboratory tests or fieldwork. Compatible
input data were not available and laboratory testing and fieldwork were not feasible for
this dissertation. In addition, the CERES model could constrain the various sets of cropclimate-management scenarios used to evaluate sensitivity and farm adjustment in this
dissertation and produce fuzzy results (Huffman 2001).

2.3.1 Description of EPIC and the applications


EPIC is a biophysical process-based model developed in the 1980s by USDAARS, SCS, and ERS (U.S. Department of Agricultural Research, Soil Conservation and
Economic Research Services) at the Grassland Soil Water Research Laboratory in
Temple, Texas. This model has been successfully used to evaluate various crop yields,
such as rice, maize, soybean, wheat, potato, in response to different growth and
environmental conditions including climate change (Williams et al. 1984; Easterling et al.
1993, 2003; Brown and Rosenberg 1999; Mearns 2001).

59

EPIC uses a single model to simulate all crops, but the values of heat unit are
unique for different cultivars (Williams 1989; Easterling et al. 2003). EPIC can simulate
crop growth for both annual and perennial crops. For annual crops, such as rice, the
simulation period runs from planting to harvest or until the accumulated heat units equal
the potential heat units for the given crop. For perennial crops, which generally maintain
their root systems over time, the growth period starts when the average daily air
temperature exceeds the base temperature of the plant (Williams 1989).
The EPIC model is designed to simulate physically based processes that influence
soil erosion and crop growth processes. The major components in EPIC are comprised of
weather simulation, hydrology, erosion-sedimentation, nutrient cycling, plant growth,
tillage, soil temperature, and crop management. The model operates on a daily time basis
at the single farm scale (Williams et al. 1984). EPIC estimates potential yields by taking
into account four key factors: crop characteristics, weather, soil fertility, and soil
properties Williams (1989). In the simulation, EPIC employs radiation-use efficiency
(RUE) concept to convert PAR into the biomass. Crop RUE crop photosynthesis, crop
biomass, and yield increase with increasing atmospheric CO2 concentration (Stockle et al.
1992a). Atmospheric CO2 concentration also influences water use efficiency (WUE) by
altering the stomatal conductance of crops, leading to changes in evapotranspiration; the
changes in WUE are reflected in the EPIC simulation result. The potential biomass is
determined by the stresses from five factors: water, nitrogen, phosphorus, temperature,
and aeration (Easterling et al. 2003). These stresses determine the potential biomass, so

60

modifying the harvest index value in the model to simulate the stresses results in altered
simulated yields. Besides yields, EPIC simulates many other outputs including
evapotranspiration, WUE, crop potential growth, and growth constraints, such as water
stress, nitrogen and phosphorus stress, heat stress, and others (Easterling et al. 1992).
In this dissertation, EPIC has been modified to allow consideration for the effects
of dynamically rising CO2 fertilization on crop photosynthesis and evapotranspiration,
which follows methods employed in previous climate change analyses (Easterling et al.
1992; Stockle et al. 1992b; Brown and Rosenberg 1999). To adjust EPIC for dealing with
CO2 fertilization effects, Easterling et al. (1992) adjusted the LUE (light-use efficiency)
and the LE (latent heat of vaporization and evaporation) parameters to calculate crop
photosynthesis and potential evapotranspiration under rising CO2 concentrations,
respectively. Since LUE is sensitive to the atmospheric CO2 concentration, the changes in
LUE value due to elevated CO2 concentration will alter the rate of photosynthesis and
crop biomass. The equation for finding the appropriate LUE value for a specific CO2
concentration is provided by Stockle et al. (1992a).
EPIC uses the Penman-Monteith formula to estimate the evapotranspiration
resulting from crop sensitivity to changing CO2 and crop environmental conditions
(Williams et al. 1989). According to Stockle et al. (1992a), the rising CO2 concentration
increases crop stomatal resistance, which reduces crop transpiration and moderates the
effect of water stress on crop growth and yield. Several studies have demonstrated the
decreases in evapotranspiration values and the increases in simulated WUE values

61

(defined as the ratio of economic yield in kg/ha and evapotranspiration in mm) after
increasing the CO2 concentration values (Stockle et al. 1992b; Brown and Rosenberg
(1999). Full details of EPIC simulation processes and crop sensitivity analysis are
described in full by Stockle et al. (1992a; 1992b), Williams et al. (1989), and Brown and
Rosenberg (1999).
In addition to crop photosynthesis and evapotranspiration, EPIC is used to
represent the agricultural performance of farms under various farm and climate
conditions (Easterling et al. 1992; 1993). EPIC requires both climate- and farm
management-related data, such as weather, soils, weather, crop variety, management
practice, tillage, fertilization, irrigation, and other farm characteristics. By assigning
different sets of farm characteristics and operation inputs, EPIC can simulate the changes
in yields and crop responses related to farm management decisions under various climate
circumstances (Easterling et al. 1993).
I needed to assemble the input data to simulate rice yields for the study sites
representing Thailands regions. After gathering the necessary data, I converted it into the
format required for model input. The input datasets were comprised of program control
codes, daily climatic data, weather station data, study site data, physical and chemical soil
data, crop data, crop management data, operation schedule data, and operational codes.
To operate the model, I created three main input files (daily weather and soil data, plus
the operations schedule for each study site) to be used in each procedure of the model
simulation as shown in Figure 2.2. Observed weather data and SRES climate projections

62

were used as the baseline and future change scenarios, respectively, to address crop
sensitivity under climate change. In the next section, the preparation of input data in
particular of future climate, soil, crop, and management practices will be described.

2.3.2 EPIC input data

2.3.2.1 Climate change projection data


Climate change datasets used in this research were based on the most extreme
scenarios among all SRES storylines; SRES A1B scenario were developed considering
rapid changes in economic growth, high efficient technologies, and global population
increase by year 2050 before its declination (IPCC 2007). Two global climate models
selected for this study were generated by the CSIRO Atmospheric Research, Melbourne,
Australia (CSIRO) and the Center for Climate System Research, Tokyo, Japan (MIROC).
The datasets of these global climate projections, hosted by the Program for Climate
Model Diagnosis and Intercomparison (PCMDI) at the Lawrence Livermore National
Laboratory (LLNL), and were used in the 2007 IPCC Fourth Assessment Report.

63

Observe
climatic data

Projected climatic data


from climate model

Observed daily statistics

Projected changes in
monthly statistics

Stochastic Weather Generator

Generated daily
weather for present
climate

Generated daily
weather for future
changed climate

Impact Models

Climate & Socioeconomic changes

Exposure units

Impacts

Figure 2.3: Methods applied to this study (adapted from Kates et al. 1985)

64

I made the decision on the selection of climate models based on the availability of
datasets required to be used as climate parameters of the crop model. The minimum
requirement datasets were the maximum temperature, minimum temperature, and total
rainfall at monthly time scale. However, not all global climate models generated
minimum temperature variables to be used; the CSIRO and MIROC climate models were
selected under this reason. The datasets for the period of 2016-2054 and the 20th century
baseline climate simulation for the period of 1970-1999 from each model were derived.
Next, the datasets were downscaled from the global scale data to the regional scale data.
Table 2.2 shows the description of climate models used in this study. Both modelsimulated datasets were generated for the monthly data from January 2000 to December
2099. The CSIRO was developed at the global resolution of 192x96 km consisting of
1200 time slices which was smaller compared to the MIROC datasets at the resolution of
128x64 km consisting of 804 time slices. These datasets were downloaded in a NetCDF
file format and decoded into multidimensional datasets in ArcGIS 9.3. The decoded data
were further downscaled to the regional scale at four selected provinces (selection
process is described in chapter 3). The monthly data would be directly extracted and used
if the grid fell in the location of the study area; otherwise, the data from 2-4 nearest grids
to each study area, which fell inside the country, were averaged before use. The grid
points for extracting climate dataset for four study areas are shown in table 2.3.
In order to run the EPIC model, daily weather data such as maximum and
minimum temperature, total precipitation are minimum requirement variables. However,

65

the projected datasets from CSIRO and MIROC climate models were generated at
monthly time scales and required the conversion from monthly data into daily data. This
research employed weather generator in EPIC to generate sequences of precipitation (wet
or dry state) occurring for a one day period independent of the other climate variables.
Table 2.2: Description of climate models
Climate models
CSIRO MK 3.5
MIROC 3.2

Institution
CSIRO Atmospheric Research,
Melbourne, Australia
Center for Climate System Research,
Tokyo, Japan

Spatial resolution
192x96

Scenario
A1B Run1

128x64

A1B Run1

Table 2.3: Grid points for each study areas


Climate models
CSIRO MK 3.5

Case study province


Northern region: Payao
Upper Central Plain: Pitsanulok
Lower Central Plain: Lop Buri

MIROC 3.2

Northeastern region: Roi Et


Northern region: Payao
Upper Central Plain: Pitsanulok
Lower Central Plain: Lop Buri
Northeastern region: Roi Et

Grid points (Latitude, Longitude)


19.58, 99.37
17.71, 99.37
15.85, 99.37
15.85, 99.37
13.98, 99.37
15.85, 103.12
20.93, 101.25
18.14, 101.25
15.35, 101.25
18.14. 101.25
15.35, 101.25
15.35, 104.06

WXGEN is the EPICs built-in stochastic weather generator developed to


generate a sequence of daily weather variables such as precipitation, maximum and
minimum temperatures, and solar radiation from a monthly data at interest locations from

66

long-term observed climate means (Easterling et al. 1993; Niu et al. 2009). In the absence
of measured weather variables, WXGEN could be used as an alternative means to satisfy
the input requirement of the EPIC model (Wallis and Griffiths 1995). In this research, I
used WXGEN to generate daily weather data simulated from CSIRO and MIROC climate
models for 2016-2054 periods.

2.3.2.2 Soil data


Clay, loamy and sandy soils were major soil characteristics used to simulate rice
yields to compare rice yield grown under different types of soil at each study area. Three
soil series found in rice land on the study areas were selected to represent each group of
soil characteristics; Nan (Ns), and Roi Et (Re), Mae Sai (Ms) soil series represented clay,
loam, and sand, respectively. The biophysical and chemical information of these soil
series were surveyed by Land and Development Department (LDD) in Thailand and the
last updated data were available for 2008 period. In order to simulate yield, EPIC
required at least seven parameters of soil data to be used as soil input parameters, such as
soil texture (percentage of sand, silt, and clay), bulk density, pH, field capacity, wilting
point, water-holding capacity, and organic matters. Other soil parameters can be
estimated by EPIC itself.

67

2.3.2.3 Crop growth and crop management data


Information on farm management practices and crop growth characteristics were
major input parameters in simulating rice yields. This research focused on Khao Dawk
Mali 105 (KDML105) rice cultivar. The information of KDML105 rice and its constraint
factors, such as water stress, potential heat unit, and salinity factors were used to identify
the basis of plant functions. For the farm management data, the input data required to run
EPIC included cropping schedules of KDML105, fertilization, irrigation, tillage, etc.
These farm management data required specific timing of individual operations either by
date or by fraction of the growth period. The data were collected from both literature
reviews and official document of Agricultural department, Thailand. These data were
initially used as input parameters and then were changed to adjust the model performance
for each study sites. Then, the adjusted input parameters were applied to the EPIC model
and simulated rice yields for two time periods 2016-2025 (centered at 2020) and 20452054 (centered at 2050). The simulated yields represented the productivity under the
baseline crop management scenario assuming no adjustment or changes in agricultural
practice and technology from current period to response to the stresses from future
climate change.

68

2.3.2.4 Adaptation options


Besides climate change, the third phase of analysis also takes into account social
factors. Hypothetically, social factors act simultaneously with climate factors that
influence farm adaptation, resulting in the changes in crop production. Chuku and Okoye
(2009) mention several types of social factors that affect agricultural adaptation, which
individual farmers can use to cope with climate change, and that influence production:
income/asset management, government programs and insurance, farm production
practices, and technological development. Among these adaptation options, the last two
options, farm practices and technological development, are generally implemented in the
studies using the process-based simulation. For example, Easterling et al. (1993; 2003)
integrated farm adaptation into the EPIC simulation model to study how changes in farm
practices influence crop yields under climate change. Lists of changes in farm practices
include shifting planting and harvesting dates, altering fertilization, and changing
irrigation amounts. To simulate possible yields resulting from these kinds of
technological improvements, Roudier et al. (2011) altered crop parameters to represent
either changes in plant physiology or improvements in crop cultivars to resist heat.
Implementing adaptations in process-based simulations make it possible both to explore
ranges of adaptations and quantify potential yields responding to climate and
socioeconomic interactions.
Despite an ample range of potential adaptation options, this dissertation limits the
choices to the current production sensitivities and adaptive capacities of farm households

69

generated by the second phase of analysis. Using this information, I constrain future
adaptation options by assuming that the current sensitivities and adaptive capacities likely
determine the abilities of the farm households to produce rice under future climate and
socioeconomic conditions. The study therefore aims to identify those adaptation options
specifically associated with the study areas without ignoring their basic production
characteristics. From this information base, I develop adaptation scenarios for EPIC to
simulate potential yields under climate change. Further details of the selected adaptation
scenarios and model parameterizations will be discussed in Chapters 5 and 6.

2.4 The justification for this research and its focus


Similar to most vulnerability assessments, a vital initial step is to define
vulnerable exposure units of concern. For this dissertation, I focused on the farm
household production unit. There is general agreement that the farm household is one of
the most sensitive groups affected by climate variability and change (Parry 1985;
Adejuwon 2006; Oxfam International 2009). Indeed, the farm household is normally the
first group to experience and be affected by first-order climate impacts through changed
crop yields. This highly sensitive and vulnerable group is defined by households that
often live on the socioeconomic margin, operating its farms with limited profits and small
buffers against hardships and with greater concern for the survival of household members
than household wealth (Parry 1985). These farm households are more likely to be found
in developing countries (Parry 1985), including Thailand (Ishii 1975). Increasing the

70

resilience of vulnerable farm households and their communities can be a stepping stone
to addressing the national and global challenges of climate change adaptation as well as
food security policy (Oxfam International 2009). Thus, the farm household level and the
changes in crop yields that so profoundly affect it is an appropriate and important place to
focus this research.

71

CHAPTER 3
PRINCIPAL COMPONENT ANALYSIS OF VULNERABILITY
This chapter describes the principal component analysis (PCA) used to identify
key indicators associated with vulnerability of Thai rice production. To reduce the 106
rice production proxies in each of Thailands 62 provinces2 into a few understandable
principal components, separate analyses are conducted for each dimension of the
Vulnerability Scoping Diagram (VSD): exposure, sensitivity, and adaptive capacity.
Before presenting those analyses, this chapter first provides an overview that explains the
assumptions of PCA, the typical steps in a PCA, and the way the PCA was employed
here.

3.1 Overview of PCA


As stated in Chapter 2, PCA is a mathematical procedure used for several
purposes including seeking patterns of common variation among observed variables and
reducing a large number of observed variables to a smaller number of components
(Tabachnic and Fidell 2007). PCA is the simplest of a larger family of factor analytical
procedures based on eigenvectors, or characteristic vectors. The eigenvectors are
2

The identification of the 106 rice production proxies was described in Chapter 2.

72

calculated from a matrix of correlation coefficients, so the fundamental assumptions of


that statistic randomly sampled variables, linear relationships among variables, and
normality also apply to PCA (ORourke et al. 2005). Additional assumptions of PCA
are that large variances have important structure and that the relationships among the
components are orthogonal (Shlens 2009).
The steps in a PCA (and any other eigenvector-based analysis) follow the
framework of Yarnal (1993). Before performing the analysis, there are four subprocedures, including the selection of decomposition modes, types of dispersion matrix,
types of analysis, and types of rotation (Figure 3.1). There are several possible modes of
data decomposition from which to choose (Yarnal 1993). Among these, P-mode
decomposition suits my hypothesis that the proxy data are inter-related and there are
groups of variables that could explain or underlie the structure and patterns of the dataset
(Dunteman 1989; Yarnal 1993). Based on this decomposition mode, the correlation
matrix was selected as the more appropriate way to evaluate and measure the correlation
and the dispersion patterns of multiple variables. Note that the covariance dispersion
matrix is actually a better fit than the correlation matrix in depicting spatial deviation of
spatial data, but the component structure may distort if important variables are clustered
and are homogeneous over space and time, as is often the case in climatic variables
(White et al. 1991).

73

Figure 3.1: Eigenvector-based classification framework (adopted from Yarnal 1993)

74
PCA was chosen over factor analysis because factor analysis is both difficult to
use and to justify theoretically (Yarnal 1993); PCA is much easier to use and does not
require complex theoretical justification. Finally, rotation was preferred over non-rotation
because rotation reduces the overlap among the variables and creates components that are
independent of each other; rotation simplifies the structure of the loadings patterns and
makes the PCA results easier to interpret (Dunteman 1989; Tabchnick and Fidell 2001;
Pett et al. 2003; Burstyn 2004). The orthogonal Varimax rotation is commonly used
because the algorithm maximizes the variances of the PCA loadings as well as the
differences between the high and low loadings (see Pett et al. 2003).
After determining the four sub-procedures, I ran several preliminary PCAs to see
if the correlation matrix was factorable and the Varimax rotation provided reasonable and
meaningful solutions. I performed three correlation criteria tests: the correlation matrix
test, Barletts test of sphericity, and the Kaiser-Meyer-Olkin (KMO) test. The criteria for
each test are adjustable and the cutoff points are arbitrary; the results from different
studies may result in unique outcomes (Pett et al. 2003). The criterion used in this
dissertation followed Hair et al. (2009) who suggested that the criterion for the intercorrelation coefficient value (r) for all variables should range from 0.3 to 0.7. A value of r
= 0.30 or above indicates multicolinearity, which the PCA addresses. However, the
multicollinearity begins to obscure underlying structure when r > 0.70. Note that setting
the correlation coefficient cutoff point at 0.7 may discard some important variables and
reduce too much information (White et al. 1991). Thus, the correlations among the

75

climate and socioeconomic variables were examined closely before the elimination.
Bartletts test of sphericity and KMO are used to measure the adequacy of sampling and
determine the strength of the linear association. The criterion for Bartletts test should be
statistically significant (p <= 0.05) and the KMO value for individual variables (referred
to as an anti-image correlation) should be equal or greater than 0.5 (Hair et al. 2009).
The final PCAs were then performed to summarize the relationships among a
large number of proxy variables with a small number of components (Pett et al. 2003;
Burstyn 2004). PCA is recognized as a suitable method for finding indicators with
relatively homogeneous clusters in climatic datasets (White et al. 1991), environmental
and ecological datasets (Wagner et al. 2006), and socioeconomic datasets (Jiang 2010).
PCA assumes that there is significant variability in the dataset, the cause of variability on
individual sources and proxy dataset is large enough to produce stable correlation
coefficients, and the selected eigenvector rotations are meaningful (Shlens 2009).
PCA generates one component for each variable but, because an important
function of the procedure is variable reduction, a crucial aspect of an analysis is
determining an appropriate number of components to retain. I employed two criteria to
determine that number: I retained variables that had eigenvalues greater than 1.0 and
looked for breaks in the slope of the eigenvalue scree plot to decide on the numbers of
extracted components. Pett et al. (2003) provide detail on how to apply these criteria.
The steps outlined above resulted in three components representing exposure, five
components representing sensitivity, and four components representing adaptive capacity.

76

Items with component loadings valued above 0.6 were used to label and describe the
components. The extracted components were then named based on the characteristics of
these items. The remainder of the chapter expands upon the exposure, sensitivity, and
adaptive capacity PCAs.

3.2 Exposure

3.2.1 Preliminary analysis


I performed a preliminary analysis to examine the 59 variables selected as proxies
that account for exposure for each of 62 provinces. The selected variables indicated both
biophysical and social vulnerability associated with seasonal climate variables, such as
maximum and minimum temperature, rainfall, water source availability, crop suitability
zones, farm damages due to extreme climate events, and rice yields.
A major constraint in analyzing the variables was the temporal inconsistency
among the available data sources. The original research design intended to perform PCA
for the years 2000, 2003, and 2008 to represent exposure during an El Nio year, a nonENSO3 year, and a La Nia year, respectively (ESRL 2010). Using three phases of

I interpreted the Multivariate ENSO Index (MEI) to indicate the ENSO phase on a monthly basis. (ENSO
refers to the combined ocean-atmosphere phenomenon, the El Nio-Southern Oscillation.) Negative values
of the MEI during the growing period (June-November) represent the cold ENSO phase or La Nia,
whereas positive MEI values represent the warm ENSO phase or El Nio. Near-zero values indicate neutral

77

ENSO, I planned to compare different growing conditions to which the social and
biophysical subjects were exposed. The impacts of ENSO events (both El Nio and La
Nia) on rice production have been reported to affect the timing of rice planting and
yields (Mitin 2009) in such countries as the Philippines and Indonesia (Naylor et al.
2001; 2006). Rice production during an El Nio or La Nia year requires different farm
adjustments and practices, especially crop water management and crop scheduling
(Naylor et al. 2006; Huang et al. 2010). Using three phases of the ENSO phenomenon
could reveal relationships among rain distribution variables, farm production variables,
and farm households sensitivity or adaptive capacity. Unfortunately, the sets of
socioeconomic variables were not available for El Nio or La Nia years; hence, I
decided to conduct the analysis only for 2003, the non-ENSO year. This dataset provides
a benchmark of farm activities in response to the rainfall distributions of a normal year.
However, it is important to keep in mind that using a non-ENSO year could
obscure relationships that occur with the ENSO-related weather shocks and that induce
changes in rice production, crop production resources, farm decisions, and farm
capacities. In other words, the results will be generalized due to the single non-ENSO
event used.
Working on exposure was complicated because monthly weather data tends to be
serially correlated. Thus, it was essential to conduct a preliminary analysis to screen

conditions, here termed non-ENSO. The time series of MEI are available at
http://www.esrl.noaa.gov/psd/people/klaus.wolter/MEI/table.html

78

correlated variables and determine whether structure among the variables might exist.
The criterion used in this dissertation followed Hair et al. (2009). Only variables that met
this requirement (0.3 r 0.7) were retained for the PCA. Then, the Bartletts test of
sphericity and the Kaiser-Meyer-Olkin (KMO) measure of sampling were performed. In
the preliminary analysis, only 16 acceptable variables remained for exposure.
Agricultural land, crop water, and farm damage variables had weak correlations with the
climate variables and were eliminated from the PCA variables matrix. The SPSS result in
Table 3.1 shows significant values for both Bartletts test of sphericity and KMO. The
former test had an approximate Chi-square value of 725.77 (p= 0.000) indicating that
sufficient correlation existed; the latter measure was quite large (approximately 0.79)
suggesting the adequacy of sample size for the PCA. Thus, these results suggested that I
could be confident that the principle component analysis was appropriate for analyzing
the data and could expect to find structure in the variables. I applied similar preliminary
analyses to the PCA of the other two vulnerability dimensions sensitivity and adaptive
capacity to maintain coherence among the analyses.

3.2.2 PCA results


The 16 components were extracted by using PCA with varimax rotation with
Kaiser normalization. Orthogonal rotation was used here to minimize the number of
variables loading highly on a single component. With fewer variables representing a
component, interpretation is simplified (Hair et al. 2010). Figure 3.2 shows the scree plot

79

illustrating that there are 5 components having an eigenvalue greater than 1.0. The slope
of the curve levels after the third component. Therefore, I decided to retain three
components to account for exposure. Results of the three-component analysis are shown
in Table 3.2. Only variables with component loadings greater than 0.60 are shown in the
list.
Table 3.2 summarizes the three components accounting for exposure. The first
component characterizes minimum temperature during the dry season, which accounts
for 28% of total variance (Table 3.3), while the third component characterizes maximum
temperature during the growing period and accounts for 18% of variance. The second
component characterizes agroclimatic conditions during the growing period and also
accounts for 18% of total variance. Together, the three components account for 65% of
the variance.

80

Figure 3.2: Scree plot for the PCA of exposure variables

Table 3.1: The KMO and Barletts test results for exposure variables
Kaiser-Meyer-Olkin Measure of Sampling Adequacy.
Bartlett's Test of Sphericity

Approx. Chi-Square

.793
725.771

Df

120

Sig.

.000

81

Table 3.2: Three-component solution with temperature and moisture items identified as
important on the exposure dimension of vulnerability
Rotated Component Matrix
Component
1

Min Temp 2003_Jan

.912

.076

.259

Min Temp 2003_Nov

.851

.076

.403

Min Temp 2003_Oct

.802

.240

.415

-.711

.191

.141

Max Temp 2003_Dec

.617

.494

.080

Max Temp 2003_Mar

-.191

.829

.121

Max Temp 2003_Sep

.078

.732

.378

-.035

.610

.002

Max Temp 2003_Jun

.082

.316

.829

Max Temp 2003_Aug

.196

.510

.751

SPI 2003_Mar

.146

-.049

.613

SPI 2003_Nov

SPI 2003_Jun

Table 3.3: Percentage of variance explained by the three components retained in the
exposure PCA
Rotation Sums of Squared Loadings
Component
1

% of Variance
28.312

Cumulative %
28.312

18.478

46.789

18.188

64.977

82

3.2.3 Interpretation of PCA results


Component 1: Dry season climate. Component 1 is defined by seasonal climate
variables associated with dry conditions. Most variables loading above 0.60 on this
component are comprised of minimum temperatures during October, November, and
January and maximum temperature in December. This component indicates a general
pattern of positive correlation with temperatures that should occur during the dry season.
In Thailand, the average temperature, especially at night, decreases at the end of the rainy
season in October. The dry season is short and lasts until January. Recall that these
variables represent the climatic conditions of a non-ENSO year.
Besides temperature, the November standardized precipitation index (SPI) has a
high negative loading (0.711) on the first component. The negative loading is caused by
anomalous wet or dry conditions during the November rice harvest. Optimum moisture
content during harvest must be considered to preserve grain quality (Grolleaud 2002).
During the dry season, moisture content normally decreases and 20-25% of moisture
content is suggested as optimum for rice harvesting. Rice quality could be degraded if it
is harvested either at high or low moisture contents: grain that is too dry cracks easily,
especially after rewetting during the milling process; grain that is too wet can be dried,
but the cost of drying may be high (UCDAVIS 2003). Thus, under unfavorable moisture
conditions at harvest, farmers may have to compromise between cost and quality.
In sum, the combination of minimum temperature and an unfavorable SPI in
November can expose crops to unfavorable climatic conditions and reduce yields, thus

83

contributing to the vulnerability of farmers who rely on good yields. For better
understanding, this result may need to combine with the results from other vulnerability
dimensions to explain the above conclusion.
Component 2: Agro-climate for rice. Variables loading above 0.6 on the second
component comprise maximum temperatures in March (0.829) and September (0.732),
and SPI in June (0.610). Considering those variables suggests the component name of
agro-climate for rice. The variables reflect two climatic elements determining
important rice growth stages. Starting from June, Thai farmers generally start sowing or
transplanting rice in this rainy season month. Optimum rainfall at the early stage of plant
growth is important in assuring good rice yields at harvest. A shortage of crop water,
especially during the planting period, poses a critical risk to soil moisture and may
stagnate rice crop growth (Welch et al. 2010).
The maximum temperature in September can potentially affect crop growth and
yield as this month falls between the end of panicle inflorescence and the beginning of
flowering (Wopereis et al. 2009; Manju et al. 2010). KDML 105 rice is sensitive to
photoperiod and the short day length (approx <12 hours) triggers flowering. An increase
in temperature during the panicle initiation of the flowering phase has been associated
with losses in grain yield. Temperature stresses affect crop physiological process by
decreasing dry matter accumulation; influencing productive tillers, reducing grain weight,
and increasing floret sterility (see Manju et al. 2010).

84

For the highest loaded variable, at first glance is seems difficult to explain how
the maximum temperature in March may have an impact on rainfed rice because this
month lies outside the growing period. However, March moisture conditions, which can
be affected by temperature, are partly responsible for successful initiation of the crop.
Above normal temperatures in March can cause excess evapotranspiration and drier than
normal moisture conditions, whereas below normal temperatures lead to much lower
evapotranspiration rates and wetter conditions, thereby affecting the agro-climate for
seedlings.
Component 3: Maximum temperature. Opposite to the first component, most
variables loading highly on the third component represent the maximum temperature
particular to the vegetation phase of rice. The first two variables with high loadings are
maximum temperatures in June (0.829) and August (0.751). The last variable indicates
the March SPI value (0.610), which determines if the condition is wetter or drier than
normal (see the discussion in the previous paragraph). Impacts of temperatures on
different growth phases have been reported (Welch et al. 2010). In June, the rice crop
passes through the critical vegetation phase (from germination to panicle initiation) when
the plant is very fragile. Temperatures that are either too low or too high can interrupt the
growth cycle and reduce yield (see Wopereis et al. 2009). On the one hand, low
temperatures can increase the duration of the vegetation phase required by the plant. On
the other hand, with high temperatures the crop reaches the end of the phase before fully
developing (Manju et al. 2010).

85

3.3 Sensitivity

3.3.1 Preliminary analysis


The correlation matrix for all 59 variables was computed and examined to reduce
as much as possible the redundancy or lack of homogeneity in the component structure.
In this study, the range of inter-variable correlations was arbitrarily accepted from r =
0.30 to 0.83 because agricultural vulnerability data are socially based and fairly high
dependence between variables can be expected. In addition to the remaining 41 variables,
the rainfed household variable had a correlation value over the accepted range (r= 0.932)
but was not eliminated from the analysis because it was deemed an important variable; its
inclusion does not change the structure of PCA results significantly. The Bartletts test of
sphericity was statistically significant and the KMO measure of sampling adequacy for
sensitivity variables was 0.677 (p=0.000) as shown in Table 3.4.
Table 3.4: The KMO and Bartletts test results for sensitivity variables
KMO and Bartlett's Test
Kaiser-Meyer-Olkin Measure of Sampling Adequacy.
Bartlett's Test of Sphericity

Approx. Chi-Square

.677
3608.499

df

861

Sig.

.000

86

3.3.2 PCA results


Based on the 42 variables entered into PCA using varimax rotation, a total of 42
components were extracted for each of 62 provinces. Figure 3.3 shows the scree plot
illustrating that there are 13 components having eigenvalues greater than 1.0. The scree
plot shows a clear drop after the fifth component, so I retained five components to
account for sensitivity. The result of five-component analysis is shown in Table 3.5.
All variables are summarized by five components (Tables 3.5 and 3.6). The first
component characterizes household economy and accounts for 27% of the variance; the
second component characterizes land scale and accounts for 14% of the variance.

87

Figure 3.3: PCA scree plot of the sensitivity components

88

Table 3.4: Five-component solution with items identified as important on the sensitivity
dimension of vulnerability

Rotated Component Matrix

Component
1

P_Rainfed household

.889

-.038

.233

.095

-.149

A_Agri-income for household

.876

.034

.207

-.150

-.113

.840

.333

.101

-.121

.103

P_Income >500 tbt

-.807

-.162

-.183

-.133

.232

P_Income 20-29 tbt

.799

.044

.332

.025

.174

-.793

-.270

-.080

-.295

.168

P_Income 10-19 tbt

.785

.180

.175

.111

-.416

P_Farm with small farm machine

.783

-.214

.157

.397

.039

-.778

-.033

-.212

-.015

-.027

.745

-.223

.277

-.231

.070

P_Irrigated farm household

-.742

-.372

-.187

.051

.057

P_Income 50-99 tbt

-.739

-.054

-.283

.063

.478

P_Income 100-499 tbt

-.728

-.246

-.211

-.036

.258

P_Renting farm land

.692

.073

-.088

.167

-.189

Monocropping

.670

-.119

.665

-.095

.118

P_Diffences in rice area 21-50 rai

-.129

-.895

-.065

.091

.173

P_Rice area 21-50 rai

-.255

-.862

.061

.103

.178

P_Rice area <5 rai

.021

.830

.094

-.172

-.345

P_Differences in rice area 6-10 rai

.130

-.754

-.008

.070

.256

P_ Rice area 6-10 rai

.336

.740

.230

-.315

-.149

P_Farm problem: low crop price

.331

.629

-.122

.078

.144

-.074

.619

-.008

-.089

.210

T_Non-native laborer

.420

-.144

.809

.111

.139

T_Unemployment member

.383

.060

.792

.056

-.008

T_No education member

.187

.318

.783

.094

-.074

consumption
P_Owning farm land

A_Income from non-agri source

P_Good public deep well


P_Farm with agricultural animal

A_Good private shallow well/village

89

Rotated Component Matrix

Component
1

T_Household member working off town

.052

-.076

.774

.110

.116

T_Iliteracy

.173

.069

.769

-.081

-.251

T_Rice area 6-50 rai

.619

-.280

.661

.064

.156

T_Rice area <5 rai

.467

.415

.653

-.110

-.058

Cropping >3 times/year

-.243

-.213

-.015

.792

-.093

Rainfed rice zone Z-1

-.139

-.140

.385

.739

.151

Cropping 2 times/year

-.112

-.156

.419

.730

.035

A_Income from agri commercial

.357

-.112

-.054

.726

.058

P_Owning large farm machine

.190

-.187

-.124

.714

.011

P_Farm problem: expensive fertilizer

.297

.232

-.037

.709

.310

P_Income <10 tbt

.426

.240

.044

-.004

-.783

P_Farm problem: poor transportation

.058

.409

.031

-.085

-.757

P_Income 30-49 tbt

.043

.243

-.086

.395

.713

P_Land under mortgage

.218

.533

.010

.019

-.711

-.045

-.055

.193

-.016

.673

P_Land with full ownership

Note: T = total number; P = percentage; A = average

Table 3.5: Percentage of variance explained by the five components retained in the
sensitivity PCA
Rotation Sums of Squared Loadings
Component
1

% of Variance
27.238

Cumulative %
27.238

14.301

41.539

13.261

54.799

10.229

65.029

9.662

74.690

Extraction Method: Principal Component Analysis.

90

Component 3 reflects human capital and accounts for 13% of the variance. Land
tenure and security of land ownership in the fourth component account for 10% of the
variance. The last component, which accounts for nearly 10% of the variance,
characterizes production capacity. The five components account for nearly 75% of the
variance.

3.3.3 Interpretation of PCA results


Component 1: Household economy. The first component is named the household
economy which measures the level of adaptive capacity of Thai farm household. The
result includes 15 variables characterizing the household economy of Thai farmers and
includes such variables as mode of farm production, household income, and cropping
intensity. These variables portray the sensitivity of farm households to changing climatic
and socioeconomic conditions. Considering all variables together, the component reflects
self-sufficiency of rainfed farm households (0.889) that occupy agricultural land (0.840)
and depend on their yields for household consumption (0.876). Earning income only
from agriculture, the economic capacity of these households is quite low and the annual
income of the households range only from 10-19 thousand baht4/$300-600 US (0.785) or
from 20-29 thousand baht/$600-900 US (0.799). The households practice a single
cropping system (0.670) meaning that the farms rely on natural rainfall without irrigation.

1 US dollar = 33 Thai baht

91

This also implies that these households rely heavily on agricultural production for family
consumption and a small income from excess yields. Crop failure would jeopardize the
food security and economic status of the households.
The scale of farm equipment is another variable that indicates the level of the
households economy. Self-sufficient households tend to possess farm assets that are
suitable for small farmland, such as small machines or equipment (0.783) and agricultural
animals (0.745). Small farm size suggests low economic status of the farm households.
Even though the households may occupy the land, the ability of farmers to
improve farmland may be constrained by the types of land ownership certificate that they
hold. A majority of farmers cultivate crops on their own land thanks to the land
registration and formalization of local property rights issued during the 1980s and 1990s
in Thailand (Phelinas 2001). Nonetheless, only 63% of agricultural households have full
rights of land ownership. The variable that loaded high on this component reflects the
farmers who hold SPK401 document type (0.692), meaning that the farmers are allowed
to cultivate crops or conduct other agricultural activities, but do not have full rights to the
land. SPK 401 only gives the farmers rights to transfer or inherit farmland to family
members, but they cannot use the land for financial creditability or for banking
transactions. As a result, with no other additional financial resource, the economic
flexibility of these households is low.
Five variables negatively load on the first component. These variables indicate
moderate to high income households who earn an income of approximately 50-59

92

thousand baht/$1500-1800 US (-0.739), from 100-499 thousand baht/$3,000-15,000 US


per year (-0.728), or more than 500 thousand baht/$15,000 US per year (-0.807). The
variables indicate the major sources of high income are from non-agricultural activities (0.793). Another highly loading negative variable reflects a rice farm household operating
with an irrigation system (-0.742). Together these negative loadings suggest a portion of
the farm household economy that has moderate to high incomes, additional income
sources, and irrigation.
Component 2: Land scale. The second component comprises 7 variables
representing the size of rice farms, which in turn relates to their sensitivity. Groups of
small landholdings that are smaller than 5 rai5/1 ha (0.830) and 6-10 rai/1-2 ha (0.740)
have positive loadings on this component. With small-scale production, the economic
status of small land holdings tends to be highly sensitive to low crop prices (0.629).
These small farms are associated with good quality, private shallow wells in the village
(0.619), suggesting that households with very small to small farms do not have indoor
plumbing and rely on nearby shallow wells for water.
Two variables have high negative loadings on the changes in rice land scales of 610 rai/1-2 ha (-0.754) and 21-50 rai/3-8 ha (-0.895) between the 2003 and 2008 crop
years. Medium-sized holdings (21-50 rai/3-8 ha) also have a negative loading. This
result is explained by government statistics that recorded an increase in the number of
holdings and the total rice area at the size of 16-50 rais from 2003 to 2008. At the same

1 hectare = 6.25 rai

93

time, there was a considerable decrease in the number of farms of less than 5 rai and 6-10
rai from 2003 to 2008 (OAE 2008). Therefore, although the total area devoted to very
small and small farms is relatively high, the recent trend is towards increasing farm size.
Component 3: Human capital. Component 3 comprises demographic variables
indicating human capital as the major production input and important source of
sensitivity. The variables loaded highly on non-native labor (0.809), which may suggest a
labor shortage in the production area. According to NSO (2008), the labor force engaged
in Thai agriculture has dwindled from 47% in 1998 to 34% in 2008. The increased
demand for non-native labor also may be indicated by the high loading on households
with family members who have moved out of town for employment (0.774).
Some high-loading variables reflect the characteristics of available household
labors and may explain the decrease in farm labor as young people migrate to seek a
better life. Several variables indicate that a proportion of household members aged
between 15-60 years are unemployed (0.792), undereducated (0.783), and illiterate
(0.769). These characteristics of available household members are associated with the
size of farm; i.e., very small (0.653) to medium-sized farms (0.661) have these household
characteristics. The third component shows strong correlation with the fourth component.
Component 4: Production capacity for commercial farm. The seven variables
loading high on this component reflect production capacity, especially of commercial rice
farms, and their sensitivity to changing conditions. The variables involve intensity of
cultivation, mode of production, source of priority income of farmers, and major

94

concerns of the production. The households are associated with more intense farm
cultivation, which can grow rice twice a year (0.730) or three times (0.792) per year. The
rice production areas are located in the nations high suitability rice zone and are
facilitated with irrigation systems and medium to high soil quality (0.739). These
attributes may explain why farm households are able to attain multiple crops per year and
obtain greater yields as a consequence. It is no surprise that most farm households derive
their income from commercial agriculture (0.726).
The capacity for commercial agriculture is also reflected by the size of
agricultural machine and concern with fertilizer cost. These farm households tend to own
large farm machines (0.714), which are necessary and worth the investment for
commercial farms. Ownership of large-scale equipment implies that farmers occupy large
plots of land. The concern over fertilizer prices indicates degree of sensitivity of farm
production (0.709). The cost of fertilizer in Thailand is driven by market fluctuations and
is always considered one of the major concerns of commercial rice farmers. Note that
Thailand lacks the major elements (nitrogen, phosphorous, and potassium) used for
synthesizing chemical fertilizers; the substances need to be imported. More important,
synthesis of inorganic fertilizers is heavily dependent on fossil fuels, which means that
fertilizer prices are subject to the fluctuation of world fuel markets and are beyond the
control of commercial farmers. In many cases, therefore, applying fertilizer is a balance
between the cost of inputs and the economic return, making this aspect of commercial
farming risky.

95

Component 5: Land tenure and security of land ownership. Five high-loading


variables reflect land tenure and security of land ownership of the farm households. The
variables involve household incomes, land tenure and ownership, and farmland
infrastructure. Three of the five variables are negatively loaded and associated with very
poor farm households earning less than 10 thousand baht/$330 US per year (-0.783).
These poor households tend to have poor road conditions and transportation networks to
the farmlands (-0.757) and to have mortgages on their land (-0.711). Considering all
negatively loaded variables implies that these households struggle with financial issues
that limit their capacity to manage farmland. Earning small incomes, these farmers tend
to spend money on their survival, which limits their spending on land modification or
infrastructure improvement. In another words, the ownership of land with secured
financial status would likely lead to infrastructure development and ultimately contribute
to a high positive score on this component.
The positive loadings characterizes relatively higher annual household incomes of
about 30-49 thousand baht/$1,000-1,660 US (0.713) and indicates full ownership with a
land right certificate (0.673). A landholder who holds a land right certificate (Title deed,
NS 3, SK1) has a legal permit for full rights in operating the land or using the land in a
financial transaction. Having land ownership, farmers can secure household incomes as
land can be used to gain financial credit. In contrast, poorer farmers who rent or hold a
mortgage on the land do not meet the basic qualifications to secure credit or take a
second-mortgage on their land (Phelinas 2001). Considering all the variables above,

96

farmers with full land ownership are likely to have high and flexible financial status and
that status facilitates land modification and land or infrastructure improvement, thereby
reducing their sensitivity to environmental and socioeconomic change.

3.4 Adaptive Capacity

3.4.1 Preliminary analysis


The correlation matrix was computed and examined for the redundancy and
homogeneity that existed in the component structure. For adaptive capacity, 27 variables
were remained for the interpretation; all have correlation values over r = 0.30 but almost
half of those variables range between r = 0.70 and 0.9 which indicate high correlated
among each other. The high correlation for this component can be expected due to the use
of a socially based dataset. For the adaptive capacity, proxy data are strongly bounded to
social experience, which may seem redundant in the analysis results. The correlations
may also be highly redundant when describing the properties of infrastructure at different
scales. The example proxies are the scales of market channel types of land problem, types
of agricultural facilities, and types of agricultural production difficulties experienced by
farmers. As these proxies are crucial in describing the potential adaptive capacity of the
farmer and production system in detail, the exclusion of these proxies may obscure an
important feature that enhances the level of agricultural adaptive capacity. Moreover, the
elimination of the variables also decreases the KMO values indicating inadequacy of

97

variables. Thus, in this dissertation, I decided to extend the high bound of r value to 0.9 to
keep as much as possible detail for the interpretation process. The variables with r value
exceed 0.9 were eliminated and remained only the variable with 0.3 r 0.7. After the
preliminary step for this component, the Bartletts test of sphericity is significant and the
KMO measure of sampling adequacy for sensitivity variables is high at 0.735 (p=0.000)
as shown in Table 3.7.

Table 3.6: The KMO and Bartletts test results for adaptive capacity variables
KMO and Bartlett's Test

Kaiser-Meyer-Olkin Measure of Sampling Adequacy.


Bartlett's Test of Sphericity

Approx. Chi-Square

.735
2164.450

df

528

Sig.

.000

3.4.2 PCA results


Matching the 27 variables used in the PCA, the PCA with varimax rotation
extracted 27 components for each of the 62 provinces. Figure 3.4 shows the scree plot
and illustrates that 14 components have eigenvalues greater than 1.0. Based on the slope
of the scree plot, I retained three components to account for the adaptive capacity. The
result of three-component analysis is shown in Tables 3.8 and 3.9.

98

Figure 3.4: PCA scree plot of the adaptive capacity components

99

Table 3.7: Three-component solution with items identified as important on the adaptive
capacity dimension of vulnerability
Rotated Component Matrix
Component
1

T_Community request: loan source

.890

.243

.063

T_Market channel: Co-op

.780

.263

.274

T_Community request: reduce interest rate

.776

.056

-.022

T_Community request: provide farm land

.748

.177

-.054

P_Owing agricultural animal

.739

.259

.264

T_Community request: crop price insurance

.711

.023

.057

T_Community request: building pipeline

.710

.216

.052

T_Market channel: farmers local shop

.705

.245

.223

T_Community request: poverty and debt

.657

.050

-.049

P_Moving reason: seasonal off- farm

-.641

-.560

-.315

P_Agricultural storage

-.600

-.418

-.357

P_Using land-line phone

-.356

-.830

-.265

A_Owing pick-up car

-.365

-.780

-.093

P_Internet accessibility

-.203

-.773

-.152

P_All-season good road condition

.313

.738

.198

A_Owning small agricultural vehicle

.254

.707

-.046

P_Owning large farm equipment

-.059

.683

-.384

P_Owning small farm equipment

.478

.680

.063

P_Low local wage

.284

-.611

.053

P_Rice area 6-10 rai

.030

-.068

.901

P_Rice rarea 21-50 rai

.196

-.053

-.808

P_Rice area >50 rai

-.291

.206

-.784

P_Rice area <5 rai

-.192

-.182

.747

P_Lack of farm management knowledge

.382

.362

.691

P_Farm adopting self-sufficient farm practice

.299

.464

.631

P_Farmer partially own and rent land

.339

.294

.614

P_Using large farm machine >50HP

-.209

-.349

-.601

Note: T = total number; P = percentage; A = average

100

Table 3.8: Percentage of variance explained by the three components retained in the
adaptive capacity PCA

Rotation Sums of Squared Loadings


Component
1

% of Variance
25.099

Cumulative %
25.099

21.207

46.307

17.613

63.919

Tables 3.8 and 3.9 suggest that the three components summarize the variables
adequately. The first component characterizes social capital and social network,
accounting for 25% of the variance. Component 2 characterizes financial capacity, which
accounts for 21% of the variance. Last, the third component reflects infrastructure and
household mobility and accounts for 17% of the variance. Collectively, the components
account for nearly 64% of the variance.

3.4.3 Interpretation of PCA results


Component 1: Social capital and social network. Eleven variables loading above
0.60 represent social capital and social network, all of which suggest the adaptive
capacity of farm households. The majority of the variables reflect farm financial concerns
that require assistance from the government and financial institutions, for example,
providing more sources of financial credit and loans (0.890), decreasing interest rates
(0.776), allocating more farm lands (0.890), providing reasonable crop price guarantees

101

(0.711), developing water canal networks (0.710), and solving poverty and farmers debt
problems (0.657). These associations imply that most farm households are poor and
operate the farm in debt. Another variable suggesting an economic system based on poor
farmers is their use of farm animals (0.739), such as buffalo and cows, which is suitable
for small production units.
The availability of local markets may provide an additional source of finance for
farmers. For example, agricultural co-ops (0.780) and local agricultural markets operated
by groups of farmers (0.705) represent market channels and informal credit markets that
poor farmers can access. The availability and accessibility of local markets also benefit
small farm production because the distance from farm to market is shorter and requires
less time and money for transportation. The short distance between farm location and
market is widely recognized as a benefit, especially for the quality of crops as that can be
damaged while delivering or transferring products (Morgan and Munton 1971).
Two variables with negative loadings are agricultural storage (-0.600) and moving
off-farm during the non-farming season (-0.641). These negative loadings may imply that
establishing or providing agricultural storage within the community and offering more
local jobs during the non-farming season may contribute to a positive score on this
component. Storage provides a suitable place for farmers to keep their produce before
delivery and helps reduce damages that may occur to that produce. Increasing local
employment may encourage householders not to leave their home province, which may
in turn help to alleviate the problem of farm labor shortage. However, both these

102

variables depend on the effectiveness of government policy and the cooperative nature
and structure of the community.
Component 2: Infrastructure and household mobility. Eight variables characterize
farmland infrastructure and household mobility. Again, these variables suggest the
capacity of farm households to adapt to a changing physical and socioeconomic
environment. Variables with high positive loadings involve year-round good road
condition to the farmland (0.738), which is associated with the types of vehicle used by
farm households. The examples of vehicles are agricultural trucks (0.707), small tractors
(0.680), and large tractors (0.683). The scale of vehicle and farm machinery implies the
economic status and the operational scale of the farm household, which is far above the
poverty line.
The high negative loadings on four variables reflect communication technologies
used by the households. The means of communication loaded negatively on the landline
telephone (-0.830) and internet access (-0.773). Note that, in some areas the landline
phone is required for the internet connection. Four-door cars (-0.780) indicate great
personal mobility of family members. This assumption is supported by a negative loading
on low local wages (-0.611), which discourages people from staying and working in their
home province.
Despite the fact that internet accessibility may be considered a push factor for
labor migration, it can benefit farm production within the community if used in a positive
sense. Internet and phone communication can be used for agricultural technology

103

transfer; they are fast, cheap, and relatively easily updated. Local job creation strengthens
the community and lessens farm labor shortage. Farmers who are able to use the internet
indicate a new, well-educated generation, which means that the ability to adopt new farm
knowledge is higher.
Component 3: Financial capacity. The third component is represented by eight
variables that indicate financial capacity of small farm households to adapt to changing
conditions. The variables include the scale of farm operation, scale of farm equipment,
and choice of farm practice. For example, small rice farms scaled from 6-10 rai/1-2 ha
(0.901) or less than 5 rai/1 ha (0.747) load highly on this component. Small farm
householders are characterized by a lack of farm knowledge (0.691) and a lack of farm
labor (0.614). The choices of farm operation also imply the financial capacity of farmers:
the adoption of self-sufficient farm techniques (0.631) indicates small-farm operation and
likely associated financial constraints.
The negative values loaded highly on large-scale farms with large machines. The
medium-sized farms of 21-50 rai/3 to 8 ha (-0.808) or larger than 50 rai/8 ha (-0.784)
reflect relative wealth of farmers and high adaptive capacity of farmers financial status.
If these variables were positively loaded, it would indicate improved financial capacity of
small farm holders should they become more prosperous.

104

3.5 Conclusions
The results from using VSD and PCA (Figure 3.5) show that it is possible to
identify and interpret key indicators that influence the Thai rice production system. Three
climate components were indicated as key sets of stresses for exposure to climatic
variation and change. Five components regarding labor, land, capital, and production
system were suggested to be key indicators that affect farm household and farm
production sensitivity. Last, three components indicating the capacity of the system in
handling vulnerabilities corresponded to social capital, financial capacity, and
communication and infrastructure. However, the effects of each component on individual
farm households may be unequally distributed over space. Therefore, in the next chapter
mapping will be used to explore the patterns of vulnerability across Thailand and
highlight those areas that are most and least vulnerable to climatic and socioeconomic
change.

105

Figure 3.5: Final VSD with key vulnerability indicators

106

CHAPTER 4
VULNERABILITY MAPPING
This chapter aims to explore the vulnerability patterns of individual provinces and
the overall vulnerability patterns of the country. The objective is to select case study
provinces that have the lowest vulnerability component scores based on the analysis
results from the previous chapter. In that chapter, 11 vulnerability indicators (i.e.,
principal components) were extracted from the PCA. Three components reflect the
physical vulnerability (exposure) and eight components reflect the social vulnerability
(five for sensitivity and three for adaptive capacity) of the rice production system.
Here, the component scores of the vulnerability dimensions for each of the
provinces were computed using the component score coefficient matrix generated by the
PCA. For each component, I calculated the standardized component score (1) by
accumulating the multiple values of the variables coefficient (C) with a variables
standardized score on original observed data (Z).
Component score 1= {(C1* Z1)+(C2*Z2)+..+ (Cn*Zn}

(4.1)

Where n = number of variables that contribute to the component. Then, I mapped


the standardized component score in each of 62 provinces, classified into five equalinterval classes to represent the degrees of vulnerability from highest to lowest

107

vulnerability. The maps for exposure, sensitivity, and adaptive capacity, respectively, are
shown and discussed in the following three sections.

4.1 Physical vulnerability


Two exposure components extracted from the PCA establish the bounds for
climate conditions minimum and maximum temperatures which are critical for the
growing period of current rice production. The first component, minimum temperature
during the dry period in the non-ENSO year, is important to the rice harvest and is shown
in Figure 4.1A. The map distinguishes areas of lowest minimum temperature in dark
grey to highest minimum temperature in red6 and shows that several provinces in the
Northeast region have component scores in the lowest class where the areas of low
minimum temperature during the dry period are located. Many provinces in the West and
in the Central Plain have component scores in the highest class indicating that these areas
have high minimum temperature compared to the Northeast region.
Based on the agricultural literature mentioned in the previous chapter, both higher
and lower minimum temperatures adversely affect moisture content of the rice grain,
which in turn decreases the quality of grain and influences harvest management of the
farm. This fact implies that the provinces in the Northeast, West, and Central Plain

Throughout this chapter, I use the convention that provinces in red represent very high
component scores, in pink represent high scores, in white represent moderate scores, in light grey
represent low scores, and in dark grey represent very low scores.

108

(assuming they harvest rice during November) require enhanced farm management
practices in maintain grain quality and good crop price. Minimum temperatures appear to
be more favorable for harvest in the north Central Plain and North regions. However, it
should not be assumed that production in the central and northern regions will be less
vulnerable to changes in minimum temperature only that harvest climatic conditions
are typically more favorable in these regions.
This finding demonstrates that with todays climate some regions are more
vulnerable than others are and climatic variation over space requires variable
management strategies to maintain rice production. Nevertheless, it is difficult to know
which production areas would be more vulnerable in El Nio or La Nia years and if
climate change were to cause harvest climatic conditions to shift. For example, the winter
floods in 2009-10 damaged vast areas of mature, ready-to-harvest rainfed rice in
Thailand. Despite the vulnerability denoted in Figure 4.1A, rice production in the upland
Northeast region only suffered minor damage to rice fields. In contrast, the lowland
central farm areas, where 1 million to 23 million tons of paddy rice experienced
significant damage, felt the greatest impacts (Reuters 2010).
The component that indicates optimum agro-climate conditions for rice is shown
in Figure 4.1B. The areas suggested by the PCA for having an optimum maximum
temperature in September and an optimum rainfall during the growing period in June
(i.e., high to very high component scores) are located in provinces in the central to
western regions extending into the North region of Thailand. The examples of current

109

major rice production areas are Suphan Buri and Nakon Sawan provinces. The areas with
moderate high maximum temperature that are currently produce rice crops are Uthai
Thani, Kampeang Phet, Pitsanulok, Sukhothai, and others. Many provinces in the
Northeast region where glutinous rice for household consumption is cultivated have low
to very low scores on the agro-climatic component.
The component scores for the third component, high to low maximum
temperatures, is shown in Figure 4.1C. The areas with highest component scores
indicating high maximum temperature during the growth period are located in the lower
Central Plain. The majority of the nations rice production is currently cultivated in this
area. In contrast, the extreme North and Northeast areas of the country, as well as three
coastal provinces in the extreme Southeast, have the lowest component scores. On the
one hand, as maximum temperature in June is critical for the vegetation development
phase of rice, the distribution of component scores implies that provinces in the lower
Central Plain gain production benefits from higher levels of maximum temperature
compared to other regions. On the other hand, it is doubtful that this region will be the
first to experience increased benefits from increased temperature at this stage of the
growing cycle in the future (See section 3.2.3). Once the maximum temperature threshold
for the rice crop has been reached in the lower Central Plain, the production belt may be
expected to move to those areas currently possessing moderate to low component scores
in maximum temperature level such as the upper Central Plain or upper Northeast.

110

Figure 4.1A: Exposure component 1: Minimum temperature

Figure 4.1B: Exposure component 2: Agro-climate

111

Figure 4.1C: Exposure component 3 Maximum temperature

112

4.2 Social vulnerability

4.2.1 Sensitivity
Five sensitivity components contributing to the vulnerability of farm households
and rice production systems are household economy, land scale, human capital, land
tenure and security, and production capacity. The most vulnerable farm households and
production areas are located in the Northeast region where small-scale production
dominates. The least vulnerable groups are located in the lower and upper Central Plain
regions where large commercial farms prevail.
Figure 4.2A shows that the farm household economy in each region is not equally
distributed. The areas of very low agricultural economy with an average household
income per year less than 30 thousand baht/$1,000 US are located along the border areas
in the Northeast of Thailand. Low income households are located in the heart of the
Northeast where most households cultivate rainfed rice (especially glutinous rice)
primarily for household consumption. The major source of household incomes in this
region is from selling excess agricultural yields. In contrast, provinces with the highest
farm household incomes have average annual household incomes that range from 50-500
thousand baht/$1,600-16,000 US and are located in the Central Plain region where most
commercial farms operate. As industrialization expands into the central region, the
diversity of income sources becomes more diverse. As demonstrated by the proportion of

113

household income variable, high-income households are not dependent on agriculture as


the only income source.
Changes from one farm size to another farm size (e.g., the combinding of small
farms into the larger farms, or the fracturing of larger farms into smaller farms) could
indirectly reflect the consequences of agricultural economic transactions and land use
(Phelinas 2001). Figure 4.2B shows that some provinces in the Central Plain region have
high component scores for the changes in small (6-10 rai/1-2 ha) and medium (21-50
rai/3-8 ha) farmland scales, suggesting increasing consolidation of smaller farms into
larger holdings. The lowest component scores, which indicate that a change from one
farm size to another size is much less likely, are found in small holdings of 6-10 rai/1-2
ha or less in the North region where small farms continue to dominate.
In any case, transactions are likely to involve landowners (not mortgagees or
renters) with relatively high economic status. Small-farm holders tend to have low
household economic status and to be more sensitive to production stress, but changes in
land occupied by this group are limited (Phelinas 2001). Land represents a valuable asset,
family wealth, and high social value; land can be bequeathed to descendants. Thus, the
decision to sell land and leave the current family or subsequent generations landless
would likely be distressing to poor farm households. Selling the land would be the last
resort and would not expect to be predominant in those regions where small-farm holders
are the majority.

114

The transition from the agricultural sector to industrial and urban sectors as a
result of economic development leads to high mobility in younger, more-educated farm
household members. As the income return from non-farm employment is generally
higher than the return from farm work, people are encouraged to leave their rural home
towns for the employment opportunities of larger urban centers. The empirical evidence
also supports the idea that higher income households result when a large proportion of
income comes from non-agricultural activities (Isvilanonda 2008). This trend is similar in
those households that invest both capital and time in high education levels; the choice of
work is made on whether or not the reward is worth the investment (Huffman 2001).
These assumptions are reflected in the distribution of the human capital component
(Figure 4.2C). The component scores indicate the mobility of household members in
combination with educational levels as reflected in labor productivity and employment
opportunity for the agricultural economy across Thailand. In the Northeast region, the
component scores for several provinces are moderately low, indicating low educational
attainment and a shortage of farm labor. Relatively high scores (i.e., higher educational
attainment and little farm labor shortfall) are found in two eastern provinces but most
importantly throughout the lower Central Plain region where manufacturing, business,
and financial institutions are centralized and basic infrastructure (roads, irrigation
networks, etc.) are more developed. The mobility of farm labor, especially among better
educated household members, from the Northeast to the Central Plain has long been
acknowledged as a national problem.

115

Except for the most rugged landscapes, the possession of land tenure and land
ownership security seems to be equally distributed among the farm households across
most of the country and does not seem to be a major source of vulnerability (Figure
4.2E). Other socioeconomic factors such as human capital and production capacity
warrant more attention as sources of sensitivity in the farm economy and farm production
systems.
The Central Plain has higher farm household economic and production capacity
than other regions. As indicated by the high component scores on human capital (Figure
4.2C) and production capacity (Figure 4.2D), components 3 and 4 denote farm
households possessing higher educational and economic levels, occupying larger
holdings on fertile and irrigated farmland, and engaging in commercial farm production.
Therefore, farm households in the Central Plain currently have higher economic status, a
more prosperous production environment, and more flexible production inputs compared
to other regions. Moreover, business and industry are centered in this region. As the
consequence, high mobility and flexibility in agricultural economic transactions are
expected. These transactions are reflected in the observed changes in farmland use and
farm size.

116

Figure 4.2A: Sensitivity component 1: Household economy

Figure 4.2B: Sensitivity component 2: Land scale

117

Figure 4.2C: Sensitivity component 3: Human capital

Figure 4.2D: Sensitivity component 4: Production capacity

118

Figure 4.2E: Sensitivity component 5: Land tenure and security of land ownership

119

4.2.2 Adaptive capacity


Three principal components that contribute to the degree of adaptive capacity are
social capital/network, financial capacity and infrastructure, and household mobility.
These components could either mitigate or exacerbate the vulnerability that currently
exists due to physical exposure and socioeconomic sensitivity. As shown in Figure 4.3A,
B, and C, the influence of each component is not equally dispersed across Thailand and
not all provinces have equal access to these adaptive mechanisms due to differences in
the geographical suitability, economic development, and agricultural production
environment in each area. Overall, high component scores show greater relative
socioeconomic and production capacity in the lower and upper Central Plain in contrast
to the low component scores in the Northeast.
Figure 4.3A indicates very low social capital and very weak social networks in the
southern areas of the Northeast. Low social capacity characterizes the north-central part
of this region, as well as the nations North. Relatively high social capacity is dispersed
across the lower Central Plain and other isolated southern provinces.
In Thailand, farm size determines agricultural household income and farm
management type (Isvilanonda 2001). Very low and low component scores on
agricultural financial capacity are mainly located in the North where small upland rice
farms are dominant (Figure 4.3B). The ability to expand farmland and even to
consolidate farms is limited due to the mountainous terrain, which in turn limits the
development of basic infrastructure such as irrigation, transportation, and communication

120

networks. This condition reduces accessibility of farm households to new agricultural


technology compared to other regions. Combining this geographical constraint with low
financial status means that most farm households have less opportunity to improve farm
output and household finances. To offset these constraints, northern farmers adopt
affordable self-sufficient farm production practices that are dependent on family farm
labor and are more suitable to the capabilities of small farms. In addition, the low
financial standing of the small farm household economy is partly offset by small and
informal credit and financial sources and local-oriented market channels (Barry and
Robison 2001).
The lower and upper Central Plain has high component scores on the financial
capacity component indicating high flexibility and high farm opportunity (Figure 4.3B).
The production systems in the central region are relatively larger and inputs are more
accessible and abundant due to the favorable production environment: flat, fertile land
with well-developed irrigation and transportation networks (Figure 4.3C). With large
farms and good transportation, the use of large farm machinery to reduce labor demand
and increase farm output characterizes this component. Moreover, these capabilities
allow producers in the Central Plain to have better access to new agricultural innovations
and better opportunity in improving farm financial status compared to farmers in most
other regions.

121

Figure 4.3A: Adaptive capacity component 1: Social capital and social network

Figure 4.3B: Adaptive capacity component 2: Financial capacity

122

Figure 4.3C: Adaptive capacity component 3: Infrastructure and household mobility

123

4.3 Calculating overall vulnerability


With the complexity of the agricultural production system, it is difficult to judge
whether any single production area is currently more or less vulnerable than others.
Based on the vulnerability maps presented here, a province may show great exposure to
climate stresses and sensitivity to socioeconomic factors, but diminished impact due to its
high adaptive capacity. Thus, assessment of individual provincial must take all
components into consideration.
Consequently, I assess the overall vulnerabilities of four regions farm households
and rice production systems in this section. The four regions are the main rice producing
regions and include the North, Northeast, upper Central Plain, and lower Central Plain. I
combine all component scores from each component and for each province into a single
value to compare and rank the provinces by region. I arbitrarily standardize the scores for
each component from -2, -1, 0, 1, 2, corresponding respectively to the dark grey through
red categories mapped in this chapter and with -2 meaning highest vulnerability and +2
meaning least vulnerability. To simplify the calculation and without theory to suggest the
importance of one component over another, I assume all components contribute equally
to vulnerability. See Rygel et al. (2006) for a detailed explanation of the pros and cons
associated with various weighting schemes used to determine overall vulnerability.
The results of the calculation (not shown) provide a ranking of overall
vulnerability by province, by region. I will only focus on the top-ranked, least vulnerable

124

provinces. The numbers of least vulnerable provinces located in the North are Payao,
Chieng Rai and Chien Rai. For the Northeast provinces, Kalasin, Chaiyapum, and Roi Et
are in the top rank. The top candidates for the upper Central Plain are Pitsanulok, Pichit,
and Kamphang Phet provinces. For the lower Central Plain, the list includes Lop Buri,
Uthai Thani, and Chinart provinces. Among these provinces, I select only four provinces
one from each region as case study sites to represent the regional production under
future climate change. The case study provinces are Payao, Roi Et, Pitsanulok, and Lop
Buri, representing the production areas of the North, Northeast, Upper Central Plain, and
Lower Central Plain, respectively (Figure 4.4).
The selection of case study sites from the list of least vulnerable provinces is
based on the datasets required for running the EPIC crop simulation model. The most
important dataset is the observed climate data, which forms the baseline climate data in
EPICs weather generator. For this research, the selected case study must have
continuous data available from 1970 to the present. Moreover, the last phase of the
analysis uses projected data from global climate models generated at gridded location
points; therefore, the location of each province must be an acceptable distance from each
other so that the grid points do not overlap and create homogenous climate projections for
the four regions. The selected least-vulnerable provinces fulfill these criteria and form the
backdrop for the simulation of Thai rice production under climate change presented in the
following chapter.

125

Figure 4.4: Location of the four case study provinces

126

CHAPTER 5
CLIMATE AND CROP YIELD SCENARIOS
The third phase of analysis uses EPIC crop simulation to evaluate crop sensitivity
to the impacts of future climate change. I develop two types of scenarios to explore
possible impacts through simulation: climate change scenarios and adaptation scenarios.
The climate change scenarios emphasize regional changes experienced with two global
climate models: CSIRO Mk3 and MIROC. The adaptation scenarios concern changes in
agricultural technology and practice that are likely to occur in response to climate change.
I include one scenario that adopts these farm adaptation strategies and another that does
not adopt them. The combinations of climate model scenarios and adaptation scenarios
lead to four scenarios that cover some of the possible impacts of climate change on rice
production (Table 5.1).
Table 5.1: Scenarios established for the four case studies
Scenario
1
2
3
4

Description
CSIRO Mk3 climate change with CO2 rising by 1% per year; no adaptation
MIROC3.2 climate change with rising CO2 by 1% per year; no adaptation
CSIRO Mk3 climate change with CO2 rising by 1% per year; with adaptation
MIROC3.2 climate change with CO2 rising by 1% per year; with adaptation

127

5.1 Climate projection scenarios


The analysis compares projections of the CSIRO Mk3.5 (hereafter CSIRO) and
the MIROC3.2 (hereafter MIROC) models for two time slices (STF and LTF) under an
A1B emission scenario. The seasonal changes in Tmin and Tmax are the focus as these
variables are the determinants of impacts on plant functions (i.e. evapotranspiration and
photosynthetic rate) and on production yields. The increasing trends of climate change
are consistent with some climate impact studies (Wassmann et al. 2009) in that the
overall mean temperature for some regions of Southeast Asia is likely to increase.
Despite that similarity, the distribution patterns and magnitude of the two model
projections are different in detail.
Figure 5.1 shows that the changes in Tmax over time at all study areas are similar
for the two models with a small increase from STF to LTF. The magnitude of Tmax under
CSIRO during the growing season (June-November) is relatively higher than that of
MIROC. Figure 5.2 shows that magnitudes of Tmin during the growing season are also
similar between the two models. However, the average modeled changes in Tmin between
two models are different in that CSIRO projects Tmin to decrease slightly during the dry
period (November to December) of LTF at Payao and Roi Et provinces. Trends of Tmin
are projected to increase for all study areas under the MIROC at both time intervals.
The diurnal temperature range (DTR = Tmax-Tmin) can be used as an indicator of
optimal agro-climatic conditions for crops; the wider the DTR is, the less optimal agroclimatic conditions. From Figure 5.3, the range of DTR under CSIRO is narrowest during

128

June-July and becomes wider during September-November, especially when compared to


MIROC during the latter period. This finding suggests that modeled agro-climatic
conditions are excellent in June and July especially under CSIRO and that conditions are
very good under MIROC and only fair under CSIRO during September-November.
As part of the exploration of future climate, this research additionally compared
simulations of 20th century climate (20c2m in the IPCC AR4, available from the Program
in Climate Model Diagnosis and Intercomparison, PCMDI; http://www.pcmdi.llnl.gov)
generated by both CSIRO and MIROC with observed climate (not shown). I found that
both models underestimate 20th century temperatures for all four study areas. The
inability of the models to simulate observed temperatures suggests that projections of
future climates should never be used as predictions but instead should be treated as
plausible scenarios of potential future conditions. Nonetheless, CSIRO and MIROC 20th
century baselines are presented in Figures 5.1, 5.2, and 5.3 to suggest the direction of
possible future climate changes.
For precipitation, both climate models reflect variability of seasonal precipitation
in all study areas (Figure 5.4). Rainfall is projected to increase in May-August the wet
season in Thailand under CSIRO. For MIROC, rainfall is projected to increase during
the dry season and decrease during the wet season. However, MIROC projections of LTF
rainfall are excessively low (including many months with no measurable rainfall)
compared to the simulated 20th century baseline, casting serious doubts about this
models future rainfall projections for this region.

129

Two major uncertainties arise from the use of global-scale climate projections for
regional-scale climate assessment (Lobel et al. 2007; Sun et al. 2007). The spatial scale of
GCMs is relatively course, often as much as a 2-degree grid box; the distance between
cell centroids is much larger than 100 km, which is not only wider than the provinces
used in this research, but also greater than the distance between the centroids of the study
areas. Downscaling is therefore the preferred treatment for regional climate data. The
availability of downscaled data and the techniques used to derive those data when they
are available, however, are poor for developing countries (Sun et al. 2007).
Consequently, this research used direct output of GCMs. Because it is difficult to find
grid points that fall within the boundaries of the four study areas, I averaged climate

130

Figure 5.1: Maximum temperature compared between CSIRO (left) and MIROC (right).
In this and subsequent figures, LOP = Lop Buri, PIT = Pitsanulok, PYO = Payao, and
RET = Roi Et.

131

Figure 5.2: As in Figure 5.1, but for minimum temperature

132

Figure 5.3: As in Figure 5.1, but for diurnal temperature range

133

Figure 5.4: As in Figure 5.1, but for average rainfall

134

values from the two closest grid points to address this issue. Thus, the first uncertainty
results from the mismatch between the position of the simulated grid point and the
centroid of the four study provinces.
The second uncertainty derives from the mismatch between the temporal scale of
the available modeled data and the time steps required by the EPIC crop model. In this
study, both GCMs provide monthly scale data that need to be converted into daily data.
However, the underestimation of seasonal precipitation, in particularly by MIROC,
results in low quantities of daily rainfall after the conversion and consequently affect the
simulated yields.
These spatial and temporal uncertainties highlight the importance of using
multiple climate projections when assessing climate change impacts and require the use
and interpretation of simulated climate data with caution. For this dissertation, I tried not
to alter the projected climate data as much as possible. Thus, despite the anticipated low
temperature and precipitation projections, these monthly scale data were not edited
further before data conversion for the EPIC crop model. Instead, I used the data as is.
In summary, the CSIRO scenario suggests elevated temperatures and normal moisture
conditions with rising CO2 concentrations. MIROC also suggests higher temperatures but
projects significant constraints on crop water availability in the long-term future.

135

5.2 Crop and crop management scenarios

5.2.1 Crop and crop management baseline parameterization


Crop and farm management parameter sets were developed as Scenario 1 and 2
baselines to represent the current production systems in the four case study provinces.
The scenario datasets depict general crop management and practices suggested by the
Bureau of Rice Research and Development, Ministry of Agriculture and Cooperatives,
Thailand (BRRD 2010). This dissertation focuses on the production of Khaw Dok Mali
105 (hereafter KDML 105), better known as Thai jasmine rice. For the baseline crop and
crop management parameter sets, the data were obtained from the Land Development
Department (LDD). These data have been used to simulate rice yields for a projection of
Southeast Asia START Regional Center under the Global Change System for Analysis,
Research and Training (START) and for an article by the Bureau of Rice Research and
Development (BRRD 2011).
These baseline crop parameter sets were then slightly modified in EPIC to
represent rice production in each of the study areas. The final parameter sets for Scenario
1 and 2 are not a detailed description of an actual farm as some parameters were adjusted
in order to run the simulation model to simulate rice yields close to the actual yields for
the period 1999-2009 recorded by the Office of Agricultural Economic (OAE 2010). As a
result, the values of parameter sets for each study area may differ slightly.

136

5.2.1.1 Crop parameter


The rice growing season is determined by the major planting and harvesting days
reported by BRRD (2011). Planting and harvesting dates may vary from place to place
and time to time, but the crop schedule used in this study represents a generalized crop
schedule for Thailand. Therefore, rainfed rice production runs from early June to midNovember.
The phenology of KDML105 depends on air and water temperature and on
photoperiod (day length) differently in each development phase. KDML 105 has three
major developing phases: the vegetative phase (from germination to panicle initiation),
the reproductive phase (from panicle initiation to flowering), and the maturation phase
(from flowering to maturity). Most parts of the developing phases are influenced by
temperature and timing; only the flowering phase is dominated by photoperiod. KDML
105 is a photoperiod-sensitive variety, so the flower induction is triggered in a short-day
period (less than 11 hr/day) that falls in October-November in Thailand. In other words,
no matter when the rice is grown, KDML 105 will blossom only in these months
(Craufurd and Wheeler 2009). Figure 5.5 shows details of rice development stages used
to develop crop and crop practice parameters.

137

03
S

10

14-15

18-19

ST

PI

Weeks
Stages

Figure 5.5: Phenology of KDML 105 (expressed in number of days); sowing starts in
May and transplanting in June (adopted from Wopereis et al. 2009; BRRD 2011).
Note: S= sowing; T= transplanting; ST= start of tilling; PI= panicle initiation; F=
flowering; M= maturity.

Rice management practice schedule needs to match the cropping environment and
timing of rice development. In Thailand, the planting season generally begins in May to
June and coincides with the rainy season so that rice will gain enough water. Sowing
starts in May (or June if using transplants). The life cycle takes about 18-19 weeks, with
one month required for rice to transition from panicle initiation to flowering, and another
month for the flowering stage to reach maturity. The rice harvest occurs in November to
December, which is approximately 90 days after the initiation of flowering. Note that the
flowering stage of KDML105 will start only when day length is shorter than 11 hours.
Thus, it takes 125-143 days (i.e. 18-19 weeks) for KDML105 to reach maturity
(Wopereis et al. 2009; BRRD 2010). Table 5.2 shows planting schedule input data for
crop management module in EPIC for the Scenario 1 and 2.

138

5.2.1.2 Soil data


Although clay soils are generally more favorable for current rice production,
under changes in climate, in growing environment or in other factors, it is possible that
new growing areas on different types of soils could emerge. To explore the potential
changes in rice yields under various soil types, I select three soil series that are
predominant in the four study areas (LDD 2011). Nan (Na; a silty clay), Roi Et (Re; a
sandy loam), and Mae Sai (Ms; a silty clay loam) are used to represent the characteristics
of clay, loam, and sand soils, respectively. The dominant soil characteristics followed the
soil classification of LDD (2005). The examples of soil properties used as input data
(Table 5.3) were collected from the Land and Development Department, Thailand.

Table 5.2: Agronomic and management parameter input data of KDML105 for Scenarios
1 and 2
Parameter description
Starting date
Planting date
Harvesting date
Plants/m2
Row spacing
Irrigation
Potential heat unit
Total water requiremnt
Base temp.to estimate phonological phases
Floodwater depth
Planting method
Fertilizer (N) application
- 16-16-8 (20 kg/rai)
- Urea (7kg/rai)

Source: BRRD (2010)

Input data
May 25
June 1
November 25
280
10 cm
Automatic
1900
1563.0
10 Deg C
80 cm
Transplanted
1st time-June 25; 2nd time Oct 15
Oct 25

139

Table 5.3: Characteristics of selected soils used for Scenarios 1 and 2


Soil Texture / Soil series
Soil order
Taxonomic name

No. to horizontal layers


Sand (%)
Silt (%)
Clay (%)
pH water
Carbon (%)
Ca (meg/100 g)
Mg (meg/100 g)
Na (meg/100 g)
CEC
FC
PWC
AWC
Hydrologic group

Silty Clay
Nan (Na)
Alfisols
Fine, Mixed,
Isohyperthermic
4
5.50
49.00
45.50
4.80
0.64
4.30
3.30
0.60
15.10
0.37
0.22
0.15
D

Sandy Loam
Roi et (Re)
Ultisols
Fine-loamy, mixed,
isohyperthemic
Typic Rhodustalfs
5
67.90
11.60
16.20
5.20
0.33
1.60
0.30
0.20
2.50
0.14
0.09
0.04
A

Silty Clay Loam


Mae sai (Ms)
Alfisols
Fine-silty, Mixed,
Isohyperthermic
5
1.00
63.30
35.70
6.50
1.04
15.30
4.90
0.30
22.00
0.38
0.22
0.16
D

Source: LDD (2005)

Note: soil data is shown only the first horizontal layer (Source: BRRD 2010)
Source: BRRD (2010)

5.3 Simulation results for scenarios 1 and 2


This section reports the results of simulated rice yields in response to the impacts
of climate variability and change at the four case study sites. The simulation results are
derived from the CSIRO and MIROC climate scenarios, with current agricultural
practices assuming no adjustments at two time scales (STF and LTF). Using the EPIC
crop model with the interaction effects from multiple production factors, the

140

simultaneous effects of CO2 and climate change under the current crop management were
taken into account. The results generated from EPIC are shown in Figure 5.6 including
yield production, evapotranspiration, and irrigation requirements as a result of the
influence of climate and of management factors on crop photosynthesis and
evopotranspiration processes.

5.3.1 Crop yields


As expected, the simulated yields project a decrease due to the impacts of rising
temperature and CO2 concentration (Figure 5.6). Under current farm practice, the most
negative yields for both STF and LTF result under the CSIRO Scenario 1. The decrease
in yields for all study areas with all soil characteristics are projected to range from 2-54%
in STF and 15-58% in LTF. Production with sandy soil is most detrimental.

141

Payao

Pitsanulok

Lop buri

Roi et
Figure 5.6: Simulated yields from EPIC under Scenario 1 and 2

142

5.3.2 Evapotranspiration and water use efficiency


Besides yields, the EPIC simulation provides water use efficiency (WUEF) and
evapotranspiration (ET) values. These two variables are correlated with crop yields and
plant functions responding to climate changes in Scenario 1 and 2.
Figures 5.7 and 5.8, respectively, show the effects of climate changes under
CSIRO and MIROC on crop WUEF and ET. Focusing on CSIROs projected 0.8-1 C
increase with great amounts of rainfall during the growth period, WUEF values for all
study areas and at both time intervals are minimal. Despite the rising CO2 levels (at 380
ppm), ET rates are therefore high for all study areas for both STF and LTF. Thus,
simulated yields decrease under Scenario 1.
For Scenario 2, MIROC projects a 1.00-1.30 C increase with a significant
decrease in rainfall. WUEF rates are high for MIROC compared to the values for
CSIRO,with noticeable increases in WUEF values from STF to LTF. The sharp increases
in WUEF values from STF to LTF for crops with loam soil are especially promising. In
contrast, ET values are low compared to those of CSIRO, decreasing dramatically from
218-248 mm/day in STF to 22-31 mm/day in LTF. These trends in WUEF and ET are
consistent at all study areas. For MIROC, simulated yields in STF generally decrease
except Payao and Lop Buri on clay soil. As reported earlier, changes in yield vary in
direction for LTF.

143

Payao

Pitsanulok

Lop buri

Roi et
Figure 5.7: Crop water use efficiency (WUEF) simulated by EPIC

144

Payao

Pitsanulok

Lop buri

Roi et
Figure 5.8: Evapotranspiration (ET) simulated by EPIC

145

5.3.3 Discussion: impacts of combined crop-climate relationships on yields

5.3.3.1 Scenario 1
Yields might be expected to increase because CSIRO projects more rain water
available for crop growth. Indeed, the simulation results show that the increment of
rainfall during the growth period increases ET values. WUEF values indicating dry
matter per unit of water are also minimal. In this case, however, the changes in rainfall
may have little impact on rice yields because the irrigation parameter is set to Auto in
EPIC to ensure that the rice crop will have enough water throughout the growing period,
which is realistic for the four study areas. Therefore, in areas under irrigation or with the
likelihood of adequate available water, changes in precipitation, stomatal resistance, and
humidity have nominal effects on rice yields and ET values. Conversely, the crop
becomes more sensitive to changes in other climatic factors such as temperature and solar
radiation that affect ET and reduces simulated yields (Brown and Rosenberg 1999).
Among these other climatic factors, rising temperature can explain the decreased
yields in the first scenario. Generally, rising temperature with elevated CO2 induces heat
stress in crops. In order to deal with this stress, plants reduce the apertures of leaf stomata
to decrease evapotranspiration and subsequently increase water use efficiency. However,
under the irrigation Auto setting in EPIC, stomatal resistance and water use efficiency
have small effects on yields.

146

Here, the increase in temperature might drive the decreased rice yields in Scenario
1. However, the investigation of single-variable effects on a crop may require different
analysis techniques, which is beyond the scope of this study. It is worth noting that
Brown and Rosenberg (1999) developed a sensitivity analysis to explore single and
multiple crop-climatic factor effects on yields. In their research, decreased yields were
projected for an irrigated crop (corn) under the combination of increasing rainfall and
rising temperature with rising solar radiation, similar to the crop-climatic conditions in
this dissertation. Their research showed that no combination of crop-climatic factors was
sufficient to benefit production under this circumstance; the corn yields were projected to
decrease because of the rising temperature. Applying and testing the assumptions and
findings of Brown and Rosenberg to rice yields must wait for future research.
In addition to the simple temperature factors discussed above, DTR values may be
relevant to the simulated changes in crop yields (Lobell 2007; Le 2010). Limited research
has shown that temperature fluctuations between Tmax and Tmin can reduce the optimum
length of the growing season and can be harmful to crop yields. In the present research,
CSIRO demonstrates wide ranges of DTR during August-November, which is a critical
period for the flowering and grain-filling phases of rice. However, because studies of the
relationship between DTR and crop yield are limited, the true nature of the impacts of
DTR are still not well understood (Le 2010).

147

5.3.3.2 Scenario 2
The MIROC results do not demonstrate the huge crop yield losses seen in the
CSIRO simulations (Figure 5.6). Under Scenario 2, the productions in almost all case
studies during STF gain relative yield benefits from the combination of crop and climate
factors. These relative yield benefits are even more noticeable during LTF. To be
consistent with Scenario 1, I use the findings of Brown and Rosenburg (1997) as
reference. Their findings demonstrated that under a water-stressed condition, ET and
crops are highly sensitive to rainfall rather than other climatic factors; ET and crop yields
are reduced in this circumstance. Again, with the auto-irrigation setting in EPIC, rainfed
crop yield will not be affected by either increased or decreased rainfall. Instead, the
interacting effects of rising temperature, increasing CO2 levels, and physiological factors
perhaps cause the changes in crop yield simulated in this study.
According to Kimball et al. (2002), rising temperature associated with elevated
CO2 may induce heat and water stresses in crops. As noted earlier, plants reduce the
apertures of stomata to decrease ET and increase WUEF in order to deal with these
stresses. As demonstrated in the results of the present study, calculated WUEF values are
high at all study sites and relatively high in LTF (Figure 5.7). As a result, photosynthesis
may increase as well as crop yield. The results of Scenario 2 show that some provinces
gain benefits in the form of small increases or decreases in yields (under clay in both
periods and under loam in LTF) compared to the devastating losses of CSIRO. The

148

benefits, nonetheless, do not fully offset the lost yields under all soils and in all study
sites. Payao (North) and Roi Et (Northeast) gain higher benefits (or alternatively have
lesser losses), especially in LTF, compared to other provinces.

149

CHAPTER 6
ADAPTIVE STRATEGY SCENARIOS
This chapter focuses on crop sensitivity to future climate change, which also
accounts for the interactions of socioeconomic drivers in the form of farm adaptation
strategies. The PCA analysis the second phase analysis of this dissertation
established that strong adaptative capacities, including social network, financial capacity,
technological transfer, and transportation, exist at the farm household level in the four
case study areas. These findings suggest wide ranges of possible adaptation strategies for
constructing scenarios 3 and 4. The adaptation strategies can range from a simple change
in crop schedule to the implementation of high-innovation crop techniques that are
affordable and accessible to farm households in both the short term and long term.
Here, I design the adaptation scenarios by taking into account the types of
adaptation approaches, the capacity of farm households to adapt, and the time scale
needed to achieve adaptation. Two adaptation approaches, the escape and the tolerate
approaches, are suggested for the agricultural sector to overcome stresses from projected
climates (Jagadish et al. 2010). The first approach involves shifting farm schedules to
avoid a crops exposure to heat stress during its critical growth period. The second
approach includes using improved crop varieties that are tolerant of stressful agronomic
or climatic conditions. These two scenario sets are anticipated to be viable and accessible
to Thai farmers in the future. However, each approach requires different levels of

150

agricultural research and development (R&D) and different time scales for the
technology to be accessible to farm households.
Therefore, I establish two parameter sets to represent various levels of
technological readiness and time scale. Min represents a minimum level of technological
advancement and farm practice, whereas Max represents a maximum level of
technological advancement and its full accessibility to farm households. Combining these
two parameter sets with adjustments in planting schedule generates four sub-scenarios for
the adaptation options (Table 6.1). This research uses two time intervals, the short-term
future (STF, 2019-2025) and the long-term future (LTF, 2045-2054), to represent the
time span that farm households take before implementing adaptation options into their
production system. See section 6.1, below, for more detail.
The EPIC crop model then integrates the adaptation scenarios and simulates
potential yields under different climate projections: scenario 3 uses CSIRO Mk3.5 and
scenario 4 uses MIROC3.2 climate model scenarios (hereafter, CSIRO and MIROC). For
scenarios 3 and 4, only production with clay soil types are simulated owing to the limited
soil management data available for future conditions. The results of scenarios 3 and 4
provide the general trend of potential rice production with various adaptation strategies
under the climate projections of CSIRO Mk3 and MIROC3.2, respectively.
In this chapter, section 6.1 describes the details of adaptive strategies in response
to climate variation and changes projected by CSIRO and MIROC and developed for
EPIC simulation. Section 6.2 addresses the simulated yields from Scenarios 3 and 4.

151

Section 6.3 discusses the impacts of combined crop-climate and elevated CO2
relationships on yields.

Table 6.1: Four options in adaptive strategies designed for Scenarios 3 and 4
Option

Description

Option 1 No Sc + Min

No adjustment in planting schedule with minimum crop


improvement level

Option 2 No Sc + Max

No adjustment in planting schedule with maximum crop


improvement level

Option 3 Sc + Min

Adjusted planting schedule with minimum crop


improvement level

Option 4 Sc + Max

Adjusted planting schedule with maximum crop


improvement level

6.1 Crop technology/ practice scenarios


Based on the two adaptation approaches I used here, the first escape approach
refers to changes in crop planting or harvesting dates that could be managed at the farm
level to overcome crop yield losses. This adjustment is considered to be relatively lowcost and easy to implement (Easterling et al. 2003). Farmers can operate the farm and
make their own adjustment to suit agro-climatic conditions in the short run.
Implementing the escape approach into the EPIC model is challenging. Because
of the uncertainty of how projected increases in maximum temperature and minimum

152

temperature, high variations of DTR, and indefinite rainfall distributions may affect crop
yield and crop function, the combinations of multiple climatic stresses make it difficult to
address specific response strategies for agriculture. I used a simplified crop schedule
adjustment option (hereafter, Sc) by having farmers respond to projected climate
conditions by harvesting one month earlier in the calendar year. Thus, the rice flowering
time is assumed to initiate in September instead of October and the maturity phase is
assumed not to be exposed too long to unfavorable climatic conditions such as heat stress
and wider DTR. The default option for this approach is No Sc, indicating no change in
crop scheduling (see Table 6.1).
The second approach is the tolerate approach, in which agricultural research
develops in ways that enable crops to tolerate the climatic conditions projected to exist in
the future. Given the projections of heat stress, water scarcity, flooding, and salinization,
improvements on current KDML105 varieties to tolerate these stresses and help them
survive and function can be expected. Agronomic research is already attempting to alter
crop flowering genes to stimulate early crop maturation and harvest so that the
biotechnology will be in place should climate change projections prove to be accurate
(Wassmann et al. 2010).
These research directions require different durations and capital investments to
become ready for adoption by farmers. Therefore, to address possible trends in the
tolerate approach, I developed two options representing scales of research advancement
and time. Option 1 (Min) represents a lower bound of agricultural research progress in

153

improving and adopting intensive farm-practice strategies to increase crop production.


This option refers to farm technologies that could develop and distribute to farm
households relatively quickly, such as a new fertilization application. Option 2 (Max)
represents a higher bound of agricultural research progress that would require more time
to develop and disseminate, such as crop engineering. Therefore, to affect yields, the
EPIC model was parameterized to represent crop technology/practice for Min and Max
differently. Crop module parameters, such as those for crop phenological attributes (e.g.
leaf area, panicle size, etc.), are set higher for Max compared to Min.
The design of adaptation scenarios takes into account the time interval that
farmers are expected to take before changing their farm practices or implementing new
farm technologies with regard to climate change. According to Easterling et al. (2003),
the time needed for full adoption of agricultural innovation ranges from 10 to 30 years.
Approximately 30 years are required for new cultivar development, but it might take only
10 years to improve fertilization application or crop water management. However, the
length of time for these innovations to be diffused and to be adopted at the farm scale
may vary for several reasons. For example, not all farmers become aware of changes in
agro-climatic conditions and respond to it at the same time, and not all farmers have the
same capacity in managing their farms. Nonetheless, the PCA results indicate high
adaptive capacity exists in the four case study provinces and that farm households are
financially able to invest in and gain access to new farm innovations with less difficulty
than other provinces in Thailand. It is uncertain, however, that the capacity of households

154

in the case study areas will remain competitive in the long-term future. With this concern,
I set two adaptation time intervals in an attempt to capture possible consequences from
climate and farm adaptations over time. The two time intervals are STF (2019-2025) and
LTF (2045-2054).

6.2 Simulation results for Scenarios 3 and 4


This section reports the results of simulated rice yields in response to the impacts
of climate change after adjustments in crop technology/practice. The simulation used
projected climate scenarios from CSIRO (Figure 6.1) and MIROC (Figure 6.2) climate
models at two time scales, STF and LTF. The level of CO2 fertilization for STF is 480
ppm and for LTF is 740 ppm (with a baseline CO2 concentration level of 380 ppm).

6.2.1 Crop yields

6.2.1.1 Option 1: No Sc + Min


Despite no shifting in farm schedule, Option 1 provides promising simulated
yields under climate change with CO2 levels at 480 ppm. The simulations show that this
combination of crop technology/practice could help compensate the negative impacts on
yields of projected climate change during STF. For this time interval, Roi Et and Payao

155

provinces have especially significant yield increases of at least 20% from the baseline
scenario for both CSIRO and MIROC.
However, simulated yields are projected to decline in three of the four study areas
under CSIRO and two of the four under MIROC, with a CO2 level of 720 ppm together a
low crop technology/practice level (Min). This result may imply that crop
technology/practice needs to improve to capture the benefits of rising CO2 fertilization
effects.

6.2.1.2 Option 2: No Sc + Max


With the higher-level improvements in crop technology/practice, potential yields
are highest among the four options in all study areas, under both climate scenarios, and
during both time intervals. Simulated yields in all study areas are projected to increase
more than 40% from the baseline during both STF and LTF, although increases are
slightly less during LTF. This result suggests that advancement in technology together
with CO2 fertilization effects could greatly strengthen the ability of rice production to
overcome the negative impacts of climate change.

156

Figure 6.1: Simulated yields from EPIC under Scenario 3 during STF (upper) and LTF
(upper)

157

Figure 6.2: Simulated yields from EPIC under Scenario 4 during STF (upper) and LTF
(lower)

158

6.2.1.3 Option 3: Sc + Min


The changes in farm practice (i.e., adjustment in farm schedule) combined with
low crop technology are inadequate to compensate the negative impacts on yields of
climate change. Simulated yields at all locations during STF are lower than other options;
the greatest decreases in yields from the baseline are in Pitsanulok and Lop Buri
provinces in the upper and lower Central Plains. During LTF for CSIRO, significant
negative results are simulated for almost all locations; except for Roi Et, yields are
estimated to be more than 20% below the baseline. For MIROC, Roi Et again fares very
well, but the other three study areas suffer lesser production losses for this option.

6.2.1.4 Option 4: Sc + Max


EPIC projected strongly positive results for all case studies under the combination
of adjusted farm schedule and higher improvement in crop technology. The positive
yields are most promising during the earlier time interval for both climate scenarios.
During LTF, however, projected yields in some provinces are much less impressive,
especially under CSIRO, which may indicate that this option might not be as well
matched to the climate despite the crop schedule change.

159

6.2.2 Water use efficiency and evapotranspiration


This section demonstrates the effects on water use efficiency (WUEF) and
evapotranspiration (ET) caused by elevated CO2 associated with rising temperature,
which may help explain interactions between crop functions and climate change (Figures
6.3 for Scenario 3 and 6.4 for Scenario 4). Theoretically, the reduction in ET with
elevated CO2 can be expected due to the partial stomatal closures that reduce plant
transpiration demand and indirectly alleviate crop water stress. Water use efficiency
varies as it can either range from high to low or experience no change depending on
growth conditions. Under an adequate water supply, stomatal conductance has nominal
influence on the high WUEF. The benefit of elevated CO2 on stomatal conductance
becomes more obvious under water shortage conditions, enabling plants to require less
water and therefore survive longer (see Kimball and Bernacchi 2006; De Van Geijn and
Goudriaan 1996).
EPIC simulated great increases in water use efficiency (WUEF) from the baseline
value for almost all study areas during STF under both climate change scenarios and for
all options. The increased values from the baseline demonstrate that crops generally gain
benefit from elevated CO2, which helps improve WUEF. The positive impacts of elevated
CO2 on WUEF are assumed to increase during LTF; however, this assumption proves to
be false for some study areas and some options. For example, significant decreases in
WUEF are projected for Payao in most adaptive options under CSIRO (Figure 6.3).

160

Figure 6.3: Water use efficiency simulated by EPIC under Scenario 3 for STF (upper)
and LTF (lower).

161

Figure 6.4: Water use efficiency simulated by EPIC under Scenario 4 for STF (upper)
and LTF (lower)

162

Despite the projected low precipitation under MIROC, great benefits in projected WUEF
result from the elevated CO2.with the relatively high values of WUEF estimated for STF
becoming even greater during LTF, with two exceptions. Although the best cases are
projected for Option 4 (Sc + Max) during STF, two provinces (Payao and Lop Buri) have
strong decreases in WUEF during LTF under this option (Figure 6.4).
Figures 6.5 and 6.6 show that, for all options and all study areas, Scenarios 3 and
4 experience a reduction in simulated ET values from the baseline ET during STF. These
reductions in ET are even greater during LTF in most cases. Across both models, the
greatest reductions in ET values are associated with Option 3 (Sc + Min) and Option 4
(Sc + Max). For CSIRO, only Payao province, where the projected seasonal rainfall
during LTF is much lower than that during STF, has a significant increase in ET with
three of the four options. Slightly greater reductions are estimated for the ET values
under MIROC during LTF.

163

Figure 6.5: Evapotranspiration simulated by EPIC under Scenario 3 for STF (upper) and
LTF (lower)

164

Figure 6.6: Evapotranspiration simulated by EPIC under Scenario 4 for STF (upper) and
LTF (lower)

165

6.2.3 Discussion: impacts of combined crop-climate and elevated CO2 relationships


on yield

The simulation results for Scenarios 3 and 4 suggest that Thai rice production
might benefit from a combination of elevated CO2, climate change, and crop
technology/practice aimed at addressing the challenges of climate change. The elevated
CO2 influences crop physiological processes and consequently helps improve crop water
use efficiency and increases yields. However, the positive results require some changes or
adjustments in current crop technology/practice. Based on four adjustment strategies used
here, Option 2 (No Sc + Max) provides the best results among all options for both climate
scenarios, while Option 3 (Sc + Min) has the worst results for both scenarios.
On the one hand, these results may imply that there is no desperate need to change
the farm schedule in the short-term future without great improvement in crop technology
to compensate the negative impacts from high temperature alone. Even for the most
minimal option (No Sc + Min), the EPIC results still indicate good yield potential for all
locations without rescheduling the crop timetable and merely adopting good farm
practice to intensify yields during STF.
On the other hand, to maintain positive yields in the long-term future, the
simulated results show an urgent need for major advances in crop technology/practice. As
seen in Figures 6.1 and 6.2, the simulated results for both climate scenarios demonstrate
that the lower bound on crop technology/practice is incapable of overcoming the negative

166

impacts from climate change despite the elevated CO2. The combination of high
improvement in crop technology/practice either with crop rescheduling (Option 4) or
without crop rescheduling (Option 2) demonstrates high yield potential for all study sites,
especially under MIROC. Moreover, the relatively low yields simulated in Option 3
imply that adjusting farm schedules without significant advances in crop
technology/practice will not be sufficient for dealing with a changing climate. However,
achieving the right combination of crop scheduling and technology/practice for specific
locations at any one time requires resources that probably exceed those currently
available to individual farmers. Thus, to manage climate change at the farm level calls for
a strong collaboration between farmers, government, and private agencies agencies (e.g.,
the distribution of crop-climate information and seasonal weather forecasts, the building
of communication and basic infrastructure, the establishment of local training workshops
, and the provision of farm financial support programs) to maintain or boost agricultural
potential either in the near or distant future (Mizina et al. 1998; Chuku and Okoye 2009;
Mitin et al. 2009; Wreford et al. 2010).

6.2.4 Integrating the results from all phases

This section integrates all of the results from the three phases of this study. The
previous analysis has shown that Thai rice production is already exposed to high
temperature and currently receives adequate rainfall in most years, but will potentially

167

deteriorate with any increases in temperature and changes in rainfall distribution. Even
though the benefit derived from elevated CO2 could alleviate the negative impacts of
climate stresses on the rice crop, the threshold at which CO2 loses its effectiveness is not
well defined. Anticipating the future stresses from climate change, I explored ranges of
adaptive strategies and estimated their outcomes. I found that outcomes may vary
depending on the time frames of the adaptive strategies that are planned and implemented
in response to climate change, and, more important, on the capability of farmers and
related institutions in the production system. Table 6.2 summarizes the impact results
from the three phases of this dissertation. The structure distinguishes exposure,
sensitivity, and adaptive capacity the vulnerability dimensions as well as overall
vulnerability, separated by the time intervals, STF and LTF. In the short-term future, the
modulations of the impacts on rice production range from negative yields, to no change,
to very positive yields depending on the interaction between climate and human
responses and on the timeline.
Based on the EPIC simulations, rice yields in Thailand are expected to decrease if
the CO2 concentration level were to stabilize at 380 ppm under the CSIRO and MIROC
climate model scenarios. Yields are expected to decrease during the short-term future by
2-54% without adaptation and to become worse in the long term. However, the above
impacts change if socioeconomic factors that influence sensitivity and adaptive capacity
of farm households are considered. Under current farm household capacities, production
levels could be stabilized if crop climatic conditions were to change gradually and

168

farmers were to take action to prevent losses in yields. That is, with current levels of farm
skill and knowledge, Thai farmers are potentially able to adjust to climate changes as
well as adopt essential farm techniques and practices to maintain production to meet farm
household consumption. Examples of technology/practice that are affordable at the farm
level include adopting crop types and varieties suited to the new agro-climatic conditions,
implementing farm irrigation, and applying improved fertilizer, field rotation, and tilling
techniques (Mitin 2009). Thus, if the appropriate actions were to take place in the shortterm future, the projected outcomes of the rice yields could be positive to very positive.

Table 6.2: Summary of results, with (-) negative outcome, (0) moderate outcome, (+) positive outcome, and (++) very positive
outcome

Scenario / Time

Exposure

Sensitivity

Adaptive capacity

Vulnerability

Short term future

(-)
Rice production system in
4 provinces

(-/0 )
Degree of impact of climate
change on exposure unit at the
current socioeconomic level
-rice yield changes

(-/0 )
Current farm production capacity
and supporting accessibility level

(-/0 )
Yield losses approx. 10%

( + with Min )
Degrees of impact of climate
change on exposure unit at
relatively higher from the current
socioeconomic level
-rice yield changes

Current crop technology/practice,


adopting current crop variety

( + , ++ with Max )
Investment in agri. R&D
Adopting new crop varieties and
good farm practice, reducing
production constraint factors
Improvement in production input
capacity and readiness of farm
supporting system
Improvement in farm households
and production socioeconomic
status

(+)
Yield increase approx. 4251% (not vulnerable)

170

Table 6.2 (cont.)


Scenario / Time

Exposure

Sensitivity

Adaptive capacity

(-)
Rice production system in
4 provinces

( - with Min )
Degree of impact of climate change
on exposure unit at the current
socioeconomic level
-rice yield changes

( - with Min)
No significant improve in farm
production capacity and supporting
accessibility from the current

Long term future

Vulnerability
(-)
Yield losses 2-30%

No significant change in farm


households and production
socioeconomic status from the past
(+)
Rice production system in
4 provinces

( + with Max )
Degrees of impact of climate
change on exposure unit at
relatively higher from the current
socioeconomic level
-rice yield changes

( + with Max )
Heavy investment in agri. R&D
Adopt new crop varieties with
stress tolerant ability, reduce
production constraint factors
Improve farm household and
production socioeconomic status
Improve the availability and
accessibility of infrastructure,
market and financial channels,
precise agricultural weather
forecasts, government supporting
programs and policies, social
networking

(+)
Yield increases approx. 4351% (not vulnerable)

Considering the constraints discussed above, it is possible to ask what the


consequences could be if farmers were to implement farm adaptation strategies in the
next 20-30 years. The answers would vary but we should anticipate that agricultural R&D
would play an even bigger role in maintaining and perhaps improving yields for Thai rice
agriculture in the long-term future. The low-cost farming techniques (that is, the No
adaptation scenario) used at the present would no longer be adequate to maintain or raise
yields under projected CO2 levels of 740 ppm and higher temperatures. Technological
engineering appears to be the most desirable adaptation strategy to use farm resources
(i.e., new crop variety, land, water, fertilizers, chemicals, etc.) efficiently and induce
productivity as shown in the results of scenarios 3 and 4.
The best scenario for the long-term future (the Max scenario) anticipates the
readiness of farm households and full development of farm technologies to deal with
climate change. Hypothetically, the farm households that survive and even thrive in this
scenario are those that are able to take advantage of advanced technologies because they
have a strong and flexible financial status, land ownership security, good education,
access to large farm machinery, and other components identified in the sensitivity
analysis. Importantly, this scenario assumes that agricultural technologies are reachable
and affordable for farmers. As shown in the adaptive capacity analysis, the farm
households in the four case study provinces currently have these capacities, including
sound market and financial institutions, agricultural support and crop insurance programs,
basic infrastructure (road, electricity, water, communication line, etc.), and welldeveloped social networks. With these ideal production conditions, the potential yields
are expected to increase 43-51% from the baseline production.

172
Despite having strong adaptive capacities, the choices of farmers to adopt the
above-mentioned strategies determine the potential yields. This analysis shows that crop
yields could range from positive to surprisingly negative, which raises concern about
which adaptation strategies are appropriate to use. This research selected 1) employing
new crop cultivars (Max option) to represent the advancement in crop research and
technology, and 2) altering farm schedule (Sc option) to adjust to the changes in climate
conditions. These strategies are expected to increase crop yields. The use of new cultivars
with changes in crop schedule (Sc + Max) simulates high productivity in the short term
rather than in the long term. However, the positive outcomes from using new crop
varieties obscure the effects of farm scheduling practices in the No Sc + Max option. The
results suggest that it is unnecessary to reschedule the harvest time for new rice varieties
that can tolerate heat stress and mature early. Nonetheless, adjustments in crop practices
needed to adapt to future climate change should still be supported with appropriate crop
and climate information.
Variable yields can be expected due to the differentiation of farm decisions and
practices. The sets of low-level adaptation options (Min) simulate decreases in yields for
three of the four study areas in the long-term future. The lower-bound scenarios could
indicate an unpredictable social problem that could cause farmers to fail to employ
appropriate actions to deal with climate change. Alternatively, this lower bound could
represent impediments to the distribution of technology and farm knowledge to farm
households. With these lower-bound options, EPIC estimates decreases of average yields
ranging from -23% to -10% from the baseline when altering the crop harvest time (Sc

173
option). In contrast, the simulation results demonstrate almost no change in the rice yields
when farmers remain on the same crop schedule (No Sc option). The two lower-bound
scenarios provide ranges of possible outcomes stemming from the use of different
adaptive strategies that unequally affect the production areas. The results point out the
complexity of the production system, compounded by uncertainties of climate change and
human choice.
The lesson here is that Thai rice farmers of the future not only call for suitable
biotechnology, but also need to have good farm management practices in place to
maintain and improve yields with significant climate change. Realizing this argument
requires study to focus on the connection between individual producers and the larger
production system. Maintaining or improving yields can only be achieved under balanced
give and take circumstances. On the one hand, responsible institutions for example,
government agencies must give by investing in agricultural R&D and developing
systems that promote the transfer of technological innovations to the farm-level
producers. On the other hand, in order to take farmers must be able to access and adopt
those technologies. To utilize advanced technology, current constraints on farmer
capability and production capacity must be reduced. Ideally, to maximize the likelihood
that rice producers will be able to take advantage of advanced technologies (and as noted
earlier), they should have strong and flexible financial status, land ownership security,
good education, access to large farm machinery, and other components identified in the
sensitivity analysis. Moreover, for technologies to be attainable and affordable for
farmers, the adaptive capacity analysis showed that many other factors need to be present

174
and accessible, including sound market and financial institutions, agricultural support and
crop insurance programs, basic infrastructure (road, electricity, water, communication
line, etc.), and social networks.
To create production potential in the long-term future, it is essential to build
collaborative systems across scales and between private and public sectors to assure that
all of the above production system characteristics exist. Building this foundation cannot
occur overnight, so it would be wise to take action soon by putting policies in place and
developing contingency plans in parallel with R&D. Unless farm production sensitivity
is reduced and its adaptive capacity is increased, negative outcomes from climate change
should be expected.

175

CHAPTER 7
DISCUSSION AND CONCLUSIONS
Because of the complexity of the agricultural system itself and the uncertainty of
climate change, it is difficult to predict with any certainty the best strategies for helping
Thai rice production deal with the climatic future. This difficulty emphasizes the need for
research to explore various scenarios that encompass multiple assumptions about the
future climate, the future socioeconomic system of Thailand, and future agricultural
potential. This dissertations use of the Vulnerability Scoping Diagram, principal
components analysis, and the EPIC crop model has allowed us to understand some
aspects of the sensitivity of the affected systems in qualitative and quantitative terms.
This final chapter provides brief answers to the dissertations three research questions in
section 7.1 and conclusions about the potential resilience of future Thai rice production in
section 7.2.

7.1 Thai rice production and double exposure


This dissertation asks three questions about the vulnerability of Thailands rice
production to the double exposure of climate and socioeconomic change: (1) Who will be
vulnerable to double exposure and what are key indicators of vulnerability of Thai rice
production? (2) What are the key characteristics of places and agricultural practices that
might reduce the vulnerability of rice production to double exposure? (3) What are the

176
consequences on rice production resulting from double exposure to climate and
socioeconomic change?
Results of the VSD analysis and PCA reveal unequally distributed patterns of
vulnerability stemming from the double exposure to climate and socioeconomic change
across the rice production areas of Thailand. My findings here are consistent with
Leichenko and OBrien (2008) that climate and socioeconomic change could not be
considered separately in determining vulnerability because the simultaneously interacting
processes of double exposure could either amplify or dampen the impacts from each
other. Here, the four least-vulnerable provinces Payao, Pitsanulok, Lop Buri and Roi
Et are characterized by the same combination of factors: relatively low exposure of
rice production to climatic stress; low degree of social sensitivity to climatic and
socioeconomic change; and high level of adaptive capacity of the production system.
Hence, the answer to the first research question is that the greatest vulnerability occurs in
those people or places with high exposure to climate change, high sensitivity to climate
and socioeconomic stress, and low adaptive capacity. For this research, household
production is highly exposed to climate stresses when it confronts unfavorable agroclimate conditions during the growing period, experiences heat stress during the
vegetation phase, and suffers low minimum temperature and high moisture during
harvest. The most highly sensitive farm households primarily depend on rice production
for subsistence, have no alternative sources of income, possess low economic status, and
own a small farm. Households with the lowest adaptive capacities have low financial
capacity, low educational attainment, weak social networking, inadequate basic

177
infrastructure, and limited access to transportation and communication networks, and
only own or have access to small-scale farm machinery.
In answering the second research question, the PCA results extracted three key
characteristics social capital and social network, financial capacity, and infrastructure
and household mobility that determine the adaptive capacity of Thai rice production to
overcome the current and projected vulnerabilities to climate and socioeconomic change.
Results from the EPIC model suggest possible adaptive strategies for handling the
impacts on rice production of the projected change in climate. The results highlight the
urgency of reducing current socioeconomic vulnerability so that the overall production
capacity and yield can increase and prepare them for the impacts from climate change.
These findings also emphasize the critical role of agricultural research and technology
development in increasing Thailands rice production capacity in the long-term future.
The consequences on rice production resulting from the double exposure the
third research question can be anticipated with some uncertainty modulated by
interactions between scales. Responding to climate factors alone, rice yields are expected
to decrease with rising temperature; elevated CO2 concentration might not be enough to
compensate for the losses caused by temperature. However, the magnitude of these
losses can be adjusted by interactions between local, national, and international scales.
The role of these scale interactions will be discussed later in this chapter.
The mixed-method approach implemented here provides satisfying outcomes and
helps achieve the objectives of this study. Using VSD and PCA, key climatic and
socioeconomic indicators reflect the current production conditions of Thai rice

178
households that contribute to vulnerability. Although the PCA identifies well-recognized
components of agricultural theory, the findings restate several important production
factors to which the Thai agricultural sector needs to pay particular attention. Moreover,
by adopting multiple climate and production scenarios and applying them to the four
study sites, the methodology facilitates comparison of the potential of rice production and
ranges of impacts across the country. The overall results provide useful information for
decision makers to develop vulnerability reduction plans and strategies aimed at
addressing climate change in the face of concurrent socioeconomic change.

7.2 Is Thai rice production moving towards resilience?


In the past, Thai agricultural production was prosperous with abundant inputs.
Industrialization and the green revolution transformed this picture, exploiting available
resources to scarcity. More recently, the priority has been to overcome the scarcities of
land, water, labor, and capital to maintain productivity as well as to ensure resilience of
agriculture, from local to national scales.

However, with the complexity of the agricultural system, the linkages between
Thai rice production and other factors at different scales or places cannot be ignored. This
section briefly focuses on the possible impacts on Thai rice production resulting from the
actions and policies under the international market regime and interacting regimes on
climate change and environment GHGs emission reduction and global virtual water in
food crop production, respectively. Based on my review of the literature, it is unclear to

179
me whether international actions and policies in response to either issue will bring about
favorable or unfavorable transformations to Thai rice production. Figure 7.1 shows how
Thai rice production at the local/national scale can be influenced by actions at the global
scales.

The linkage between international and national scales as denoted by the regulation
of global crop demand and supply by international market price controls on production
inputs is an important concern. The consequences of technological improvements and
projected increases in crop demand lead future production to become dependent on the
intensive use of fuel-based inputs (e.g., fertilizers and chemicals). The costs of these
inputs at local/national scales fluctuate with swings in global markets and prices.
Therefore, when fuel prices increase, they spin up the costs of production inputs and
subsequently crop prices.

Because of these macro-scale relationships, Thai farmers make adjustments,


either in farm management or profits. When crop prices in the world market are attractive
enough to compensate for high input prices, farmers may grasp this opportunity and take
actions to increase yields, for example, borrow money to increase fertilizer and chemical
inputs, invest more in labor and farm equipment, rent more farmland, or adopt improved
management techniques to increase farm efficiency. Such opportunities may facilitate
increased production for medium-sized to large farms with considerable financial
reserves, but small farm households with less flexible financial status may find these

180
opportunities unaffordable. In contrast, when market prices are low, farmers may turn
from rice to alternative crops that are more profitable.

Figure 7.1: Interactions of global and local/national scales in determining the resilience
of Thai rice production
Government and public agencies may help facilitate rice production whether the
prices of energy or grain go up or down. Some strategies that have been suggested for
encouraging production at all capacity levels include agricultural subsidy programs, crop
insurance, and other supporting market and financial services.

Despite the lack of progress in the ongoing international climate change


negotiations, it is clear that part of the shared responsibilities from developing countries

181
will be requests to reduce greenhouse gases (GHGs) from agricultural activities and that
those reductions will influence the Thai agricultural system. Rice production alone
releases a great amount of methane and nitrous oxide,7 so Thailand and other riceproducing nations will be considering ways to reduce these agricultural emissions as an
important component of any climate change mitigation plan. Even without knowing what
the final share of responsibility will be, future technology and practice standards will be
moving toward managing on- and off-farm GHG emissions in a measurable, reportable,
and verifiable manner to meet the international standard (see Nelson 2009; FAO 2009).
This shift will influence decision making from local to national levels and perhaps the
transform agricultural processes in the country, including, for example, specifically
requiring the measurement of carbon fluxes from cultivation or more generally changing
the management of land and water (Bockel et al. 2008; FAO 2009). From a positive
perspective, such a transformation could bring about collaborations between riceconsuming developed countries and rice-producing developing nations focusing on green
technology investment if partnering countries recognize the mutual benefits of producing
food for the increased population of the future.

The selection of adaptive measures in response to either external or internal


production factors depends on the direction in which the Thai agricultural sector wants to

Methane (CH4) is the 2nd most important anthropogenic GHG being responsible for nearly one quarter of
the anthropogenic radiative forcing from the naturally occurring GHGs (CO2, CH4, and N2O)..Nitrous
Oxide (N2O) represents roughly 7% of the anthropogenic radiative forcing. Because CH4 and N2O are
much more efficient than carbon dioxide in radiative forcing, the reduction of CH4 or N2O emissions is
suggested to be an effective way to reduce climate change (Wassmann et al. 2009).

182
take future production and on the readiness of production capacity at that time (Nelson et
al. 2007). Failure to improve marginal elements of present production conditions will put
a double burden on the Thai rice production system when facing future economic and
environmental regimes. In addition to the local impacts of climate change, any of the
international factors mentioned above as well as an infinite variety of other
socioeconomic influences, could act as the trigger that could shift the rice production
system to exceed a critical threshold, transforming it or, in the most undesirable scenario,
collapsing the system. For Thailand to avoid losing its comparative advantage in
international trade depends on strong and effective collaborative policies and plans from
public and private institutions to maintain productivity level and socioeconomic wellbeing from national to local scales.

Another possible direction would be for Thailand to accept the potential loss of its
export market and voluntarily transform the rice production system into one aimed only
at self-sufficient production. Rice is fundamental and meaningful in Thai culture, so it is
probably safe to assume that rice production will grow sufficiently to meet national
demand. However, in the unlikely event that culture were to change and demand to
decrease due to a shift in peoples dietary preference, say, to meat and dairy products like
the West, then there would be excess rice to export. In any case, the hope is that suitable
technology would become ready in time for the Thai rice production system to recover or
to transform successfully into a new resilient state.

183
Consequently, it is wise and essential for Thai rice agriculture to establish
adaptive measures to reduce the vulnerability to present climate and socioeconomic
constraints, as well as likely short- and long-term conditions. Acting now gains
immediate benefit for society and the production system at a relatively lower cost than
the price of delaying action to the future (Burton 1996).

However, dealing with complex and dynamic interactions among the climate
system, the socioeconomic system, and rice agriculture means that research on impacts
and vulnerabilities presents considerable uncertainties. Despite suggestions from the
literature that decision-making should be based on reliable climate change predictions
(Smith et al. 1996), it is neither possible to predict the nature and timing of impacts, nor
to know when to adapt and how the Thai government and people will perceive and
respond to impacts when they do occur. Thus, I agree with Pelling (2011) that there will
be no precise technological guidance or best adaptation solution for dealing with double
exposure. Instead, future work must aim at filling the research gaps in developing
suitable methods for impact and vulnerability assessment, assuring the accuracy of
downscaled global data for local use, and facilitating an understanding among decisionmakers of the risks involved with double exposure prepared for both short term and long
term periods.

184

Appendix A
Socioeconomic variables and proxies for sensitivity and adaptive capacity

185
Table A.1: Socioeconomic variables and proxies for sensitivity and adaptive capacity
Variables
Proxies
Social and Demographic
Demographic
Households: farm household
Households: member by age group
Households: member by gender
Households: operating rainfed rice farm
Households: operating irrigate rice farm
Households: member with disabilities
Education
Population 15-60 years: household member, no
education
Population 15-60 years: household members by
education level
Employment
Population 15-60 years: household members with
employment
Population 15-60 years: household member,
unemployment
Villages: household member working off town
Villages: household member work out, permanently
Villages: household member work out, off-farm
season
Population: farm labor, non-native
Annual household
Households: average income less than 10 thousand
income
baht
Households: average income 10-19 tbt
Households :average income 20-29 tbt
Households: average income 30-49 tbt
Households: average income 50-99 tbt
Households: average income 100-499 tbt
Households: average income more than 500 tbt
Source of income
Households: income from non-agriculture source
Households: income from agricultural production
for commerce
Households: income from household production of
service for own consumption
Economic
Land characteristic
Size of land holding
Rice land ownership
Type of land ownership
Rights in occupied land

Dependency on rice
production
Frequency of
cultivation

Source

Period

CDD
CDD
CDD
NSO
NSO
CDD
CDD

2009
2009
2009
2008
2008
2009
2009

CDD

2009

CDD

2009

CDD

2009

CDD
CDD
CDD

2009
2009
2009

CDD
NSO

2009
2004

NSO
NSO
NSO
NSO
NSO
NSO
NSO
NSO

2004
2004
2004
2004
2004
2004
2008
2008

NSO

2008

Households: rice household


Households: rice landholding by land size
Households: rice area changes by land size
Households: owning farm land
Households: renting farm land
Households: land holding with rights document
Households: land holder with full ownership
Households: land holder with partial ownership
Households: land holder, under mortgage
Land units: households growing jasmine rice
Land units: households growing sticky rice
Households: doing 1 crop/year

CDD
CDD
CDD
CDD
CDD
CDD
CDD
CDD
CDD
NSO
NSO
CDD

2009
2009
2009
2009
2009
2009
2009
2009
2009
2008
2008
2009

Households: doing 2 crop/year


Households: doing >3 crop/year

CDD
CDD

2009
2009

186
Variables
Farm machine

Proxies
Households: using agricultural animal
Households: owning, small farm machine
Households: owning, large farm machine
Households: renting, small farm machine
Households: renting, large farm machine
Rainfall dependency
Land units: rice land with irrigation system
Land units: rice land, no irrigation system
Water source
Well units: village, individual-owned shallow well
Well units: village, public shallow well
Well units: village, individual-owned deep well
Well units: village, public deep well
Farm financial problem
Villages: no agricultural land
Villages: no crop seed
Villages: low crop price
Villages: expensive fertilizer
Villages: poor transportation
Villages: lack of farm capital
Villages: lack of farm labor
Villages: lack of farm production knowledge
Infrastructure & Communication
Communication means
Households: using land-line telephone
Households: using mobile
Households: using internet
Electricity
Villages: no electricity
Road
Villages: road by conditions, good in all seasons
Villages: road by conditions, poor in all seasons
Farm facility/service
Villages: rice bank
Villages: agricultural animal (cow, buffalo) bank
Villages: agricultural storage
Villages: rice mill
Villages: farm machine by power, less than 20 HP
Villages: farm machine by power, 20-50 HP
Villages: farm machine by power, more than 50 HP
Market channel
Villages: scale, local agricultural market
Villages: scale, co-op store
Villages: scale, farmer-operated group market
Villages: scale, individual agricultural store
Natural source
Soil surface water
Villages: quality of water, very good
Villages: quality of soil surface water, medium
Villages: quality of soil surface water, poor
Water supply &
Villages: quality of waste water treatment, no treatment
sanitation
Villages: quality of waste water treatment, good
Social capital /Network
Knowledge transfer
Households: using organic fertilizer
Households: adopting self-sufficient farm practice
Households: adopting organic farm

Source
CDD
CDD
CDD
CDD
CDD
LDD
LDD
CDD
CDD
CDD
CDD
CDD
CDD
CDD
CDD
CDD
CDD
CDD
CDD

Period
2009
2009
2009
2009
2009
2009
2009
2009
2009
2009
2009
2009
2009
2009
2009
2009
2009
2009
2009

CDD
CDD
CDD
CDD
CDD
CDD
CDD
CDD
CDD
CDD
CDD
CDD
CDD
CDD
CDD
CDD
CDD

2009
2009
2009
2009
2009
2009
2009
2009
2009
2009
2009
2009
2009
2009
2009
2009
2009

CDD
CDD
CDD
CDD

2009
2009
2009
2009

CDD

2009

CDD
CDD
CDD

2009
2009
2009

187
Variables
Health service
Governmental support

Proxies
Households: adopting chemical-safe farm practice
Villages: accessible clinic and hospital
Villages: community request, provide farmland
Villages: community request, improve land quality
Villages: community request, loan source
Villages: community request, financial fund source
Villages: community request, poverty and debt
Villages: community request, reducing interest rate
Villages: community request, building canal
Villages: community request, building pipe line
Villages: community request, crop price insurance
Villages: community request, crop sell guarantee

Source
CDD
CDD
NSO
NSO
NSO
NSO
NSO
NSO
NSO
NSO
NSO
NSO

Period
2009
2009
2008
2008
2008
2008
2008
2008
2008
2008
2008
2008

188

Bibliography
Allan, J. 1998, Virtual Water: A Strategic Resource Global Solutions to Regional
Deficits. Ground Water, 36: 545546.
Acosta-Michlik, L., U. Kelkar, and U. Sharma. 2008. A critical overview: Local evidence
on vulnerabilities and adaptations to global environmental change in developing
countries. Global Environmental Change 18 (4):539-542.
Adejuwon, J. O. 2006. Food crop production in Nigeria. II. Potential effects of climate
change. Climate Research 32 (3):229-245.
Adejuwon, J.O. 2008. Vulnerability in Nigeria: A national-level assessment. In Climate
Change and Vulnerability. eds. Leary, N., C. Conde, J. Kulkarni, A. Nyong, J.
Adejuwon, V. Barros, R. Lasco, and J. Pulhin. 2009.. Earthscan Publications Ltd.
Adger, W., and P. M. Kelly. 1999. Social Vulnerability to Climate Change and the
Architecture of Entitlements. Mitigation and Adaptation Strategies for Global
Change 4 (3):253-266.
AIACC. 2006. Vulnerability to climate change related water resource changes and
extreme hydrological events in Southeast Asia for the final report of the Assessments
of impacts and Adaptation to Climate Change (AIACC), AS07. Washington, D.C.:
The International START Secretariat.
Ayanlade, A. and O.T. Odekunle. 2009. GIS approach in assessing seasonal rainfall
variability in Guinea Savanna part of Nigeria. In 7th FIG Regional conference:
Spatial data serving people: Land governance and the environment- Building the
capacity. Hanoi, Vietnam 19-22 October 2009.
Attanandana, T., and S. Kunaporn. 2005. Rice soils and rice cultivation system of
Thailand.
Ayanlade, A., and O.T. Odekunle. 2009. GIS Approach in Assessing Seasonal Rainfall
Variability in Guinea Savanna Part of Nigeria GIS Approach in Assessing Seasonal
Rainfall Variability in Guinea Savanna Part of Nigeria. In Seventh FIG Regional
Conference, Spatial Data Serving People: Land Governance and the EnvironmentBuilding the Capacity. Hanoi, Vietnam.
Bachelet, D., D. Brown, M. Bhm, and P. Russell. 1992. Climate change in Thailand and
its potential impact on rice yield. Climatic Change 21 (4):347-366.

189
Barrett, C.B. and E. Mutambatsere. 2005. Agricultural markets in developing countries.
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1142518 (Last accessed: 22
September 2011)
Barry, P. J., and L. J. Robison. 2001. Chapter 10: Agricultural finance: Credit, credit
constraints, and consequences. In Handbook of Agricultural Economics Volume 1A.
eds. Gardner, B.L. and G.C. Rausser. 513-571.
Belliveau, S., B. Smit, and B. Bradshaw. 2006. Multiple exposures and dynamic
vulnerability: Evidence from the grape industry in the Okanagan Valley, Canada.
Global Environmental Change 16 (4):364-378.
Berry, P. M., M. D. A. Rounsevell, P. A. Harrison, and E. Audsley. 2006. Assessing the
vulnerability of agricultural land use and species to climate change and the role of
policy in facilitating adaptation. Environmental Science & Policy 9 (2):189-204.
Bockel, L., A. Gentien, M. Tinlot, and M. Bromhead. 2008. Module 103: From nationally
appropriate mitigation actions (NAMAs) to low-carbon development in agriculture:
NAMAs as a pathway at country level. In FAO Policy Learning Program: Overview
of program modules and sessions. FAO.
Bouman, B., E. Humphreys, T. Tuong, and R. Barker. 2007. Rice and Water. 187-237.
Academic Press
Bouman, B. A. 2007. A conceptual framework for the improvement of crop water
productivity at different spatial scales. Agricultural Systems 93 (1-3):43-60.
Boer, R. and A. Faqih. 2004. Current and future rainfall variability in Indonesia. In: An
Integrated Assessment of Climate Change Impacts, Adaptation and Vulnerability in
Watershed Areas and Communities in Southeast Asia for the report of the
Assessments of Impacts and Adaptation to Climate Change (AIACC) (Annex C),
AS21. Washington, D.C.: The International START Secretariat. P. 95-126.
Braun, J.V. 2009. Food-security risks must be comprehensively addressed.
http://www.cabdirect.org/abstracts/20103045613.html;jsessionid=B9A08DC08843E
798B4568D10EC5DD073 (last accessed 6 June 2011).
BRRD. Khao Dawk Mali 105. http://www.ricethailand.go.th/brrd/rice_tech.htm (last
accessed 8 May 2011).
Brown, R. A., and N. J. Rosenberg. 1999. Climate change impacts on the potential
productivity of corn and winter wheat in their primary United States growing
regions. Climatic Change 41 (1):73.

190
Bureau of Rice Research and Development (BRRD). Rice Knowledge Bank.
http://www.brrd.in.th/rkb2/index.php.htm (last accessed 7 March 2011).
Buddhaboon, C., K. Kunket, K. Pannangpetch, A. Jintrawet, S. Kongton, and S.
Chinvanno. 2008. Effect of Climate Change on Rice Production in Southeast Asia:
A Case Study in Thailand. In Better Air Quality Workshop, 1-4. Bangkok, Thailand.
Burstyn, I. 2004. Principal component analysis is a powerful instrument in occupational
hygiene inquires. Ann. Occup. Hyg. 48(8):655-661.
Carlo, C., and R. Tol. 2002. Sustainability and Economics: A Matter of Scale? Integrated
Assessment 3 (2-3):151-159.
Challinor, A.J., T.R. Wheeler, P.Q. Craufurd, and J.M. Slingo. 2005. Simulation of the
impact of high temperature stress on annual crop yields. Agricultural and Forest
Meteorology 135: 180-189.
Challinor, A., T. Wheeler, Ch. Garforth, P. Craufurd, A. Kassam. 2007. Assessing the
vulnerability of food crop systems in Africa to climate change. Climatic change. 83:
381-399.
Challinor, A.J, F. Ewert, S. Arnold, E. Simelton, and E. Fraser. 2009. Crops and climate
change: progress, trends, and challenges in simulating impacts and informing
adaptation. Journal of Experimental Botany 60(10): 2775-2789.
Chapagain, A.K., A.Y. Hoekstra, A.Y., and H.H.G. Savenije. 2006. Water saving through
international trade of agricultural products. Hydrology and Earth System Sciences
10(3): 455-468.
Chapagain, A.K. 2009. Water footprint of rice: Quantifying the rainbow of virtual water
fluxes related to rice. Re-Thinking paradigms: Water and food security. 4th Marcelio
Botin Foundation water workshop, Santander, Spain. 22-28 September 2009.
Chavas, J.P. 2001. Chapter 5: Structural change in agricultural production: Economics,
technology and policy. In Handbook of Agricultural Economics Volume 1A. eds.
Gardner, B.L. and G.C. Rausser. 263-261.
Chen, R. S., and R. W. Kates. 1994. Climate change and world food security: Editorial.
Global Environmental Change 4 (1):3-6.
Chhetri, N.B. 2007. Understanding the Process of Agricultural Adaptation to Climate
Change: Analysis of Climate-Induced Innovation in Rice Based Cropping System of
Nepal. PhD Diss, The Pennsylvania State University.

191
Childs, B. C. Interpolating Surfaces in ArcGIS Spatial Analyst. Education 2004 (JulySeptember):4.
Chinvanno, S., S. Souvannalath, B. Lersupavithnapa, V. Kerdsuk, and N. Thuan. 2008.
Climate risks and rice farming in the Lower Mekong River Basin: a place-based
approach. In Climate change and Adaptation, eds. N. Leary, J. Adejuwon, V.
Barros, I. Burton, and R. Lasco. London, U.K.: Earthscan.
Christensen, J.H., B. Hewitson, A. Busuioc, A. Chen, X. Gao, I. Held, R. Jones, R.K.
Kolli, W.-T. Kwon, R. Laprise, V. Magana Rueda, L. Mearns, C.G. Menendez, J.
Raisanen, A. Rinke, A. Sarr and P. Whetton. 2007. Regional Climate Projections. In
Climate Change 2007: The Physical Science Basis. Contribution of Working Group
I to the Fourth Assessment Report of the Intergovernmental Panel on Climate
Change., eds. Solomon, S., D. Win, M. Manning, Z. Chen, M. Marquis, K.B.
Averyt, M. Tignor and H.L. Miller, 94. Cambridge, United Kingdom and New York,
USA.; Cambridge University Press.
Chuku, C.A. and C. Okoye. 2009. Increasing resilience and reducing vulnerability in subSaharan African agriculture: Strategies for risk coping and management. African
Journal of Agricultural Research 4(13): 1524-1535.
CIA. 2010. The World Factbook: Thailand. Central Intelligence Agency. Available at
https://www.cia.gov/library/publications/the-world-factbook/geos/th.html (last
access 23 August 2010).
Creswell, J. W. 2008. Research Design: Qualitative, Quantitative, and Mixed Methods
Approaches 3rd ed. Sage Publications, Inc.
Craufurd, P. Q., and T. R. Wheeler. 2009. Climate change and the flowering time of
annual crops. Journal of Experimental Botany 60 (9):2529 -2539.
Cruz, R.V., H. Harasawa, M. Lal, S. Wu, Y. Anokhin, B. Punsalmaa, Y. Honda, M.
Jafari, C. Li and N. Huu Ninh. 2007: Asia. In Climate Change 2007: Impacts,
Adaptation and Vulnerability. Contribution of Working Group II to the Fourth
Assessment Report of the Intergovernmental Panel on Climate Change, eds. Parry,
M.L., O.F. Canziani, J.P. Palutikof, P.J. van der Linden and C.E. Hanson.
Cambridge University Press, Cambridge, UK, 469-506.
Cutter, S. L. 1996. Vulnerability to environmental hazards. Progress in Human
Geography 20 (4):529-539.
Cutter, S. L. 2003. The Vulnerability of Science and the Science of Vulnerability. Annals
of the Association of American Geographers 93 (1):1.

192
Dabi, D.D., A.O. Nyong, A. A. Adpetu, and V. I. Ihemegbulem. 2008. Past, present and
future adaptation by rural households of Northern Nigeria. In Climate Change
Vulnerability. eds. Leary, N., C. Conde, J. Kulkarni, A. Nyong, J. Adejuwon, V.
Barros, R. Lasco, and J. Pulhin. 2009.. Earthscan Publications Ltd.
Deininger, K. and G. Feder. 2001. Chapter 6: Land institutions and land markets. In
Handbook of Agricultural Economics Volume 1, Part A. eds. B.L. Gardner and
G.C.Rausser. London, U.K.: Earthscan. p. 288-331.
Dockerty, T., A. Lovett, K. Appleton, A. Bone, and G. Snnenberg. 2006. Developing
scenarios and visualisations to illustrate potential policy and climatic influences on
future agricultural landscapes. Agriculture, Ecosystems & Environment 114 (1):103120.
DPI. 2011. Irrigation water use efficiency benchmarking.
http://www.dpi.vic.gov.au/dpi/vro/vrosite.nsf/pages/lwm_farmwater_water_use (last
accessed 16 September 2011).
Dunteman, G.H.1989. Principal component analysis. Sage Publications, Newbury Park,
Calif.
Easterling, W. E., N. J. Rosenberg, M. S. McKenney, C. Allan Jones, P. T. Dyke, and J.
R. Williams. 1992. Preparing the erosion productivity impact calculator (EPIC)
model to simulate crop response to climate change and the direct effects of CO2.
Agricultural and Forest Meteorology 59 (1-2):17-34.
Easterling, W. E., R. C. Piere, N. J. Rosenberg, M. S. McKenney, L. A. Katz, and K. M.
Lemon. 1993. Agricultural Impacts of and Responses to Climate Change in the
Missouri-Iowa-Nebraska-Kansas (MINK) Region. In Towards an integrated impact
assessment of climate change: MINK study., ed. N. J. Rosenberg, 184. Washington
D.C.: Kluwer Academic Publishers.
Easterling, W. E. 1997. Why regional studies are needed in the development of full-scale
integrated assessment modelling of global change processes. Global Environmental
Change Part A 7 (4):337-356.
Easterling, W.E., N. Chhetri, and X. Niu. 2003. Improving the realism of modeling
agronomic adaptation to climate change: simulating technological substitution.
Climatic Change 60: 149-173.
Eriksen, S., and P. Kelly. 2007. Developing Credible Vulnerability Indicators for Climate
Adaptation Policy Assessment. Mitigation and Adaptation Strategies for Global
Change 12 (4):495-524.

193
ESRL. 2010. Multivariate ENSO Index (MEI). Earth System Research Laboratory.
Availabel at http://www.esrl.noaa.gov/psd/people/klaus.wolter/MEI/table.html.
(Last access: 1 July 2010).
FAO. Country Profiles - Thailand.
http://www.fao.org/countryprofiles/index.asp?lang=en&iso3=THA&subj=4 (last
accessed 25 March 2010).
FAO. 1997. Small-scale irrigation for arid zones: Principles and options.
http://www.fao.org/docrep/W3094E/w3094e04.htm#TopOfPage (last accessed 8
June 2011).
FAO. 2009a. How to feed the world in 2050. High-level expert forum in 12-13 October
2009, Rome, Italy. http://www.fao.org/wsfs/forum2050/wsfs-forum/en/ (last
accessed 6 May 2010).
FAO. 2009b. Anchoring agriculture within a Copenhagen agreement: A policy brief for
UNFCCC parties. http://foris.fao.org/static/data/nrc/policy_brief_sbstabonn.pdf (last
accessed 17 August 2010).
Felkner, J., K. Tazhibayeva, and R. Townsend. 2009. Impact of Climate Change on Rice
Production in Thailand. American Economic Review 99 (2):205-210.
Fssel, H., and R. Klein. 2006. Climate Change Vulnerability Assessments: An Evolution
of Conceptual Thinking. Climatic Change 75 (3):301-329.
Fssel, H. 2007. Vulnerability: A generally applicable conceptual framework for climate
change research. Global Environmental Change 17 (2):155-167.
GAIN. FAS Overseas reporting and the Global Agriculture Information Network.
Available at: http://gain.fas.usda.gov/Lists/Advanced%20Search/AllItems.aspx (last
accessed 23 February 2010).
Grolleaud, M. 2011. Overview of the phenomenon of losses during the Post-harvest
system. http://www.fao.org/docrep/004/ac301e/AC301e00.htm (last accessed 7 May
2011).
Hair, J. F., W. C. Black, B. J. Babin, and R. E. Anderson. 2009. Multivariate Data
Analysis 7th ed. Prentice Hall.
Hallegatte, S. 2009. Strategies to adapt to an uncertain climate change. Global
Environmental Change 19 (2):240-247.

194
Heslop-Thomas, C. W. Bailey, D. Amarakoon, A. Chen, S. Rawlins, D.D. Chadee, R.
Crosbourne, A. Owino, K. Polson, C. Rhoden, R. Stennett, and M. Taylor. 2008.
Vulnerability to climate-induced highland malaria in East Africa. In Climate Change
and Vulnerability. eds. Leary, N., C. Conde, J. Kulkarni, A. Nyong, J. Adejuwon, V.
Barros, R. Lasco, and J. Pulhin. 2009.. Earthscan Publications Ltd.
Hoekstra, A.Y. and A.K. Chapagain. 2008. Globalization of water: Sharing the planet's
freshwater resources. Blackwell Publishing, Oxford, UK.
Huffman, W. E. 2001. Chapter 7 Human capital: Education and agriculture. In
Agricultural Production, 333-381.
IFPRI. 2010. Food security and climate change: Challenges to 2050 and beyond.
Available at http://www.ifpri.org/publication/food-security-and-climate-change
(aspx (last accessed 23 August 2010).
Iglesias, A., C. Rosenzweig, and D. Pereira. 2000. Agricultural impacts of climate change
in Spain: developing tools for a spatial analysis. Global Environmental Change 10:
69-80.
IPCC. 2007. Climate Change 2007 - The Physical Science Basis: Working Group I
Contribution to the Fourth Assessment Report of the IPCC eds. S. Solomon, D. Qin,
M. Manning, Z. Chen, M. Marquis, K. Averyt, M. Tignor, and H. Miller. Cambridge
University Press.
Ishii, Y., 1975. Thailand: a rice growing society. Hawaii: The University Press of
Hawaii, Honolulu.
Isvilanonda. S, N. Poapongsakor, and M. Hossain.1991. Rice Supply and Demand in
Thailand: Recent Trends and Future Outlook, IRRISocial Sciences Division Paper
Series No 99-01. International Rice Research Institute, Los Bafos, Philippines.
Isvilanonda, S. and M. Hossain. 2000. Dynamics of rice farming in Chao Phraya Delta: a
case study of three villages in Suphan Buri province. In Proceedings of the
international conference : the Chao Phraya Delta : historical development,
dynamics and challenges of Thailand's rice bowl, 12-15 December 2000, Kasetsart
University, Bangkok.
Isvilanonda, S. 2001. Income disparity and income inequality of agricultural households.
Applied Economics Journal 8 (1):29-42 (In Thai).

195
Isvilanonda, S., and S. Fukui. 2002. Global Competitiveness of Thai Rice: Before and
after the Currency Crisis. http://www.lib.kobe-u.ac.jp/handle_kernel/00317864 (last
accessed 8 June 2011).
Isvilanonda, S., and W. Kongrith. 2009. Thai Households Rice Consumption and Its
Demand Elasticity. http://muse.jhu.edu/content/oai/journals/
asean_economic_bulletin /v025/25.3.isvilanonda.html (last accessed 8 June 2011).
Jagadish, S.V.K., K. Sumfleth, G. Howell, E. Redona, R. Wassmann, and S. Heuer. 2010.
Temperature effects on rice: significance and possible adapatation. In ed. Wassmann
R. 2010. Advanced technologies of rice production for coping with climate change:
no regret options for adaptation and mitigation and their potential uptake.
Proceedings of the Workshop on Advanced Technologies of Rice Production for
Coping with Climate Change: No Regret Options for Adaptation and Mitigation
and their Potential Uptake held on 23-25 June 2010 in Los Baos, Philippines. IRRI
Limited Proceedings No. 16. Los Baos (Philippines): International.
Jiang, C. 2010. Research on Logistics Network Infrastructure Based on HCA and DEAPCA Approach. Journal of Computers, 5(4):533-540.
Johnston, D. B. 1981. Rice Cultivation in Thailand: The Development of an Export
Economy by Indigenous Capital and Labor. Modern Asian Studies 15 (1):107-126.
Kasperson, J. X., R. E. Kasperson, and B. Turner II eds. 1995. Regions at risk:
comparisons of threatened environments. Tokyo-New York-Paris: United Nations
University Press.
Kates, R. W., J. H. Ausubel, and M. Berberian (eds.), 1985. Climate Impact Assessment:
Studies of the Interaction of Climate and Society, ICSU/SCOPE Report No. 27, John
Wiley.
Kelly, P., and W. Adger. 2000. Theory and Practice in Assessing Vulnerability to
Climate Change and Facilitating Adaptation. Climatic Change 47 (4):325-352.
Kimball, B. A., J. Zhu, L. Cheng, K. Kobayashi, and M. Bindi. 2002. Responses of
agricultural crops of free-air CO2 enrichment. In Advances in Agronomy. 77: 293362.
Kripalani, R. H., S. V. Singh, N. Panchawagh, and M. Brikshavana. 1995. Variability of
the summer monsoon rainfall over Thailandcomparison with features over India.
International Journal of Climatology 15 (6):657-672.

196
Kumar, M.N., C. S. Murthy, M. V. R. Sesha Sai, and P. S. Roy. 2009. On the use of
Standardized Precipitation Index (SPI) for drought intensity assessment.
Meteorological Applications 16 (3):381-389.
Kupkanchanakul, T. 2000. Bridging the rice yield gap in Thailand. In Bridging the rice
yield gap in the Asia-Pacific region, RAP Publication., eds. M. K. Papademetriou, F.
J. Dent, and E. M. Herath, 146-156. Bangkok, Thailand: FAO.
Lambin, E. F., B. L. Turner, H. J. Geist, S. B. Agbola, A. Angelsen, J. W. Bruce, O. T.
Coomes, R. Dirzo, G. Fischer, C. Folke, P. S. George, K. Homewood, J. Imbernon,
R. Leemans, X. Li, E. F. Moran, M. Mortimore, P. S. Ramakrishnan, J. F. Richards,
H. Sknes, W. Steffen, G. D. Stone, U. Svedin, T. A. Veldkamp, C. Vogel, and J.
Xu. 2001. The causes of land-use and land-cover change: moving beyond the myths.
Global Environmental Change 11 (4):261-269.
Lanza, L. G., J. A. Ramrez, and E. Todini. Stochastic rainfall interpolation and
downscaling. Hydrology and Earth System Sciences 5 (2):139-143.
Le, P. V. 2010. Climate change impacts on crop yields: Reexamine evidence from
decreasing diurnal temperature range. PhD diss, University of California at
Berkeley.
Leary, N., C. Conde, J. Kulkarni, A. Nyong, J. Adejuwon, V. Barros, R. Lasco, and J.
Pulhin. 2008. Climate Change and vulnerability. Earthscan Publications Ltd.
Leary, N., C. Conde, J. Kulkarni, A. Nyong, J. Adejuwon, V. Barros, R. Lasco, and J.
Pulhin. 2008. Climate Change and adaptation. Earthscan Publications Ltd.
LDD. 2005. The miracle of soil: Soil series for economic crops in Thailand. Land and
Development Department: Thailand.
LDD. 2010. Soil characteristics in Thailand. Soil in Thailand. http://www.ldd.go.th/
thaisoils_museum/INDEX.HTM (last accessed 25 March 2010).
Liverman, D. 2008. Assessing impacts, adaptation and vulnerability: Reflections on the
Working Group II Report of the Intergovernmental Panel on Climate Change.
Global Environmental Change 18 (1):4-7.
Lobell, D. B. 2007. Changes in diurnal temperature range and national cereal yields.
Agricultural and Forest Meteorology 145 (3-4):229-238.

197
Luers, A. L., D. B. Lobell, L. S. Sklar, C. L. Addams, and P. A. Matson. 2003. A method
for quantifying vulnerability, applied to the agricultural system of the Yaqui Valley,
Mexico. Global Environmental Change 13 (4):255-267.
Luers, A. L. 2005. The surface of vulnerability: An analytical framework for examining
environmental change. Global Environmental Change Part A 15 (3):214-223.
Manju, Z., S. D. Singh, S. N. Kumar, P. K. Aggarwal, and R. C. Harit. 2010. Impact of
elevated temperature at different phenological stages on the growth and yield of
wheat and rice. Indian Journal of Plant Physiology 15 (4).
Matthews, R. B., M. J. Kropff, T. Horie, and D. Bachelet. 1997. Simulating the impact of
climate change on rice production in Asia and evaluating options for adaptation.
Agricultural Systems 54 (3):399-425.
McKee, T.B., N.J. Doesken, and J. Kleist. 1993. The relationship of drought frequency
and duration to time scales. In Proceedings of the 8th Conference of Applied
Climatology, 17-22 January, Anaheim, CA. American Meterological Society,
Boston, MA. 179-184.
Mearns, L.O., T. Mavromatis, E. Tsvetsinkay, C. Hays, W. Easterling. 1999.
Comparative responses of EPIC and CERES crop models to high and low spatial
resolution climate change scenarios. Journal of Geophysical Research 104(D6):
6623-6646.
Mearns. L.O., F. Giorgi, P. Whetton, D. Pabon, M. Hulme, and Lal, M. 2003. Guidelines
for use of climate scenarios developed from regional climate model experiments.
Data Distribution Centre of the Intergovernmental Panel on Climate Change.
Mitchell, T. D., and M. Hulme. 1999. Predicting regional climate change: living with
uncertainty. Progress in Physical Geography 23 (1):57-78.
Mitin 2009. Documentation of Selected Adaptation Strategies to Climate Change in Rice
Cultivation. East Asia Rice Working Group (EARWG) http://www.eastasiarice.org/
Meehl, G.A., T.F. Stocker, W.D. Collins, P. Friedlingstein, A.T. Gaye, J.M. Gregory, A.
Kitoh, R. Knutti, J.M. Murphy, A. Noda, S.C.B. Raper, I.G. Watterson, A.J. Weaver
and Z.-C. Zhao, 2007: Global Climate Projections. In: Climate Change 2007: The
Physical Science Basis. Contribution of Working Group I to the Fourth Assessment
Report of the Intergovernmental Panel on Climate Change, eds. Solomon, S., D.
Qin, M. Manning, Z. Chen, M. Marquis, K.B. Averyt, M. Tignor and H.L. Miller.
Cambridge University Press, Cambridge, United Kingdom and New York, NY,
USA.

198
Matthews, R. B., M. J. Kropff, T. Horie, and D. Bachelet. 1997. Simulating the impact of
climate.
Mitchell, T. D., and M. Hulme. 1999. Predicting regional climate change: living with
uncertainty. Progress in Physical Geography 23 (1):57-78.
Mitin, A. 2009. Doumentation of selected adaptation strategies to climate change in rice
cultivation. East Asia Rice Working Group: Philippines.
Moser, S. C. 2010. Now more than ever: The need for more societally relevant research
on vulnerability and adaptation to climate change. Applied Geography 30 (4):464474.
Morgan, W.B. and S.C. Monton. 1971. Agricultural Geography. Methuen: London.
Naylor, R.L, W.P Falcon, D. Rochberg, and N. Wada. 2001. Using El Nino/Southern
Oscillation climate data to predict production in Indonesia. Climate Change 50: 255265.
Naylor, R.L., D.S. Battisti, D.J. Vimont, W.P. Falcon, and M.B. Burke. 2007. Assessing
risks of climate variability and climate change for Indonesia rice agriculture. PNAS
104(19): 7752 .
Niu, X., W. Easterling, C. J. Hays, A. Jacobs, and L. Mearns. 2009. Reliability and inputdata induced uncertainty of the EPIC model to estimate climate change impact on
sorghum yields in the U.S. Great Plains. Agriculture, Ecosystems & Environment
129 (1-3):268-276.
Nielsen, T.K. 2004. Lecture note: Paddy cultivation. http://www.kellnielsen.dk/
download/Paddy.pdf
Nelson, D.R., W.N. Adger, and K. Brown. 2007. Adaptation to Environmental Change:
Contributions of a Resilience Framework. In Annual Review of Environment and
Resources. 32: 395-419.
Nelson G.C., M. Rosegrant W., J. Koo, R. Robertson, T. Sulser, T. Zhu, C. Ringler, S.
Msangi, A. Palazzo, M. Batka, V. Valmonte-Santos, E. Ewing, and D. Lee. 2009.
Climate change: Impact on agriculture and costs of adaptation. Washington, D.C.:
International Food Policy Research Institute (IFPRI).
Nelson G.C., M. Rosegrant W., A. Palazzo, I. Gray, C. Ingersoll, R. Robertson, S.
Tokgoz, T. Zhu, T. Sulser B., C. Ringler, S. Msangi, and L. You. 2010. Food

199
security and climate change: Challenges to 2050 and Beyond. IFPRI Issue Brief 66
pp. 8.
Nelson G.C., M. Rosegrant W., A. Palazzo, I. Gray, C. Ingersoll, R. Robertson, S.
Tokgoz, T. Zhu, T. Sulser B., C. Ringler, S. Msangi, and L. You. 2010. Food
Security, Farming, and Climate Change to 2050: Scenarios, Results, Policy Options.
Washington, D.C.: International Food Policy Research Institute.

Nelson, R., P. Kokic, S. Crimp, H. Meinke, and S. Howden. 2010a. The vulnerability of
Australian rural communities to climate variability and change: Part I-Conceptualising and measuring vulnerability. Environmental Science & Policy 13
(1):8-17.
Nelson, R., P. Kokic, S. Crimp, P. Martin, H. Meinke, S. Howden, P. de Voil, and U.
Nidumolu. 2010b. The vulnerability of Australian rural communities to climate
variability and change: Part II--Integrating impacts with adaptive capacity.
Environmental Science & Policy 13 (1):18-27.
NESDA, Office of the National Economic and Social Development Board.
http://www.nesdb.go.th/ (last accessed 27 February 2010).
Nguyen, N.V. 2002. Overview: Global climate changes and rice food security. Rome,
Italy: FAO
Niu, X., W. Easterling, C.J. Hays, A. Jacobs, L. Mearns. 2009. Reliability and input-data
induced uncertainty of the EPIC model to estimate climate change impact on
sorghum yields in the U.S. Great Plains. Agriculture, Ecosystems and Environment.
129: 268-276.
OAE. 2010. Office of Agricultural Economics, Thailand. Available at:
http://www.oae.go.th/ewt_news.php?nid=4542 [Accessed January 24, 2010].
O'Brien, K. L., and R. M. Leichenko. 2000. Double exposure: assessing the impacts of
climate change within the context of economic globalization. Global Environmental
Change 10 (3):221-232.
O'Brien, K., R. Leichenko, U. Kelkar, H. Venema, G. Aandahl, H. Tompkins, A. Javed,
S. Bhadwal, S. Barg, L. Nygaard, and J. West. 2004. Mapping vulnerability to
multiple stressors: climate change and globalization in India. Global Environmental
Change Part A 14 (4):303-313.

200
O'Brien, K. L., S. Eriksen, L. P. Nygaard, and A. Schjolden. 2007. Why different
interpretations of vulnerability matter in climate change discourses. Climate Policy
7:73-88.
ORourke, N. L. Hatcher, and E.J. Stepanski. 2005. A step-by-step approach to using SAS
for univariate & multivariate statistics. 2nd ed. Wiley.
Osman-Elasha, B. and E.A. Sanjak. 2008. Livelihoods and drought in Sudan. In Climate
Change and Vulnerability. eds. Leary, N., C. Conde, J. Kulkarni, A. Nyong, J.
Adejuwon, V. Barros, R. Lasco, and J. Pulhin. 2009. Earthscan Publications Ltd.
OSullivan, D., and D. Unwin. 2003. Geographic Information Analysis 1st ed. Wiley.
Oxfam International. 2009. People-Centered resilience: Working with vulnerable farmers
towards climate change adaptation. Oxfam International Briefing Paper.
http://www.oxfam.org/policy/people-centered-resilience (Last accessed 22
September 2011).
Parry, M. L., C. Rosenzweig, A. Iglesias, M. Livermore, and G. Fischer. 2004. Effects of
climate change on global food production under SRES emissions and socioeconomic scenarios. Global Environmental Change 14 (1):53-67.
Patt, A., R. J. T. Klein, and A. de la Vega-Leinert. 2005. Taking the uncertainty in
climate-change vulnerability assessment seriously. Comptes Rendus Geosciences
337 (4):411-424.
Pearsall, H. 2009. Linking the stressors and stressing the linkages: Humanenvironment
vulnerability and brownfield redevelopment in New York City. Environmental
Hazards 8 (2):117-132.
Pelling, M. 2010. Adaptation to Climate Change: From Resilience to Transformation 1st
ed. Routledge.
Pett, M.A. Lackey, N.R., and Sulivan, J.J. 2003. Making sense of factor analysis.
Thousand Oaks, CA: Sage.
Phlinas, P. M. 2001. Sustainability of rice production in Thailand. Nova Publishers.
Philippe Roudier, Benjamin Sultan, Philippe Ouirion, and Alexis Berg. 2011. The impact
of future climate change on West African crop yields: What does the recent
literature say?. Global Environmental Change. 21(3) 1073-1083

201
Polsky, C., J. Allard, N. Currit, R. Crane, and B. Yarnal. 2000. The Mid-Atlantic Region
and its climate: past, present, and future. Climate Research 14 (3):161-173.
Polsky, C., R. Neff, and B. Yarnal. 2007. Building comparable global change
vulnerability assessments: The vulnerability scoping diagram. Global Environmental
Change 17 (3-4):472-485.
Reilly, J., N. Hohmann, and S. Kane. 1994. Climate change and agricultural trade : Who
benefits, who loses? Global Environmental Change 4 (1):24-36.
Reilly, J. M. ed. 2002. Agriculture 1st ed. Cambridge University Press.
RDIC. 2011. Rural Development Information Center.
http://61.19.244.12/rdic/?mod=report_jbt (last accessed 7 June 2011).
Rosegrant, M., M. Paisner, S. Meijer, J. Witcover. 2001. Global Food Projections to
2020: Emerging Trends and Alternative Futures. International Food Policy Research
Institute, Washington, D.C.
Rosenberg, N. J. 1992. Adaptation of agriculture to climate change. Climatic Change 21
(4):385-405.
Rosenzweig, C., and M. L. Parry. 1994. Potential impact of climate change on world food
supply. Nature 367 (6459):133-138.
Rtter, R.P., T.R. Carter, J.E. Olesen, and J.R. Porter. 2011. Crop-Climate models need
an overhaul. Nature Climate change 1: 175-177.
Roudier, P., B. Sultan, P. Ouirion, and A. Berg. 2011. The impact of future climate
change on West African crop yields: What does the recent literature say? Global
Environmental Change. 21(3): 1073-1083
Ruiz-Ramos, M. and M.I. Minguez. 2010. Evaluating uncertainty in climate change
impact on crop productivity in the Iberian Peninsula. Climate Research 44: 69-82.
Rygel, L., D. Osullivan, and B. Yarnal. 2006. A Method for Constructing a Social
Vulnerability Index: An Application to Hurricane Storm Surges in a Developed
Country. Mitigation and Adaptation Strategies for Global Change 11 (3):741-764.
Saseendran. S.A., K.K Singh, L.S. Rathore, G.S.L.H.V.P. Rao, N. Mendiratta,
K.L.Narayan, and S.V. Singh. 1998. Evaluation of the CERES-Rice version 3.0
model for the climate conditions of the state of Kerala, India. Meteorol. Appl. 5:
385-392.

202
Shlens, J. 2009. A tutorial on principal component analysis.
http://www.snl.salk.edu/~shlens/ (last accessed 15 August 2011).
Shivakoti, G. 2005. Asian irrigation in transition. SAGE.
Sivakumar, M. V. K. 2006. Climate prediction and agriculture: current status and future
challenges. Climate Research 33 (1):3-17.
Sivakumar, M. V. K., and J. Hansen. 2010. Climate Prediction and Agriculture:
Advances and Challenges Softcover reprint of hardcover 1st ed. 2007 ed. Springer.
Smit, B., and J. Wandel. 2006. Adaptation, adaptive capacity and vulnerability. Global
Environmental Change 16 (3):282-292.
Smith, J. B., N. Bhatti, G. V. Menzhulin, R. Benioff, M. Campos, B. Jallow, F.
Rijsberman, M. I. Budyko, and R. K. Dixon. 1996. Adapting to Climate Change: An
International Perspective 1st ed. Springer.
Stockle, C.O., P.T. Dyke, J.R. Williams, N.J. Rosenberg, and C.A. Jones. 1992. A
method for estimating the direct and climatic effects of rising atmospheric carbon
dioxide on growth and yield of crops: Part I- Modification of the EPIC model for
climate change analysis. Agricultural Systems 38: 225-238.
Stockle, C.O., P.T. Dyke, J.R. Williams, C.A. Jones, and N.J. Rosenberg. 1992. A
method for estimating the direct and climatic effects of rising atmospheric carbon
dioxide on growth and yield of crops: Part II- Sensitivity analysis at three sites in the
Midwestern USA. Agricultural Systems 38: 239-256.
Sun, L., H. Li, M. N. Ward, and D. F. Moncunill. 2007. Climate Variability and Corn
Yields in Semiarid Cear, Brazil. Journal of Applied Meteorology and Climatology
46 (2):226-240.
Tabachnick, B. G., and L. S. Fidell. 2006. Using Multivariate Statistics 5th ed. Allyn &
Bacon.
TACFS. 2003. Good agricultural practices for Thai Hom Mali rice, Thai agricultural
commodity and food standard TACFS 4400-2003.
Timsina, J. and E. Humphreys. 2006. Performance of CERES-Rice and CERES-Wheat
models in rice-wheat systems: A review. Agricultural System 90: 5-31.
Tilman, D., K. G. Cassman, P. A. Matson, R. Naylor, and S. Polasky. 2002. Agricultural
sustainability and intensive production practices. Nature 418 (6898):671-677.

203
Tobey, J. A., and G. V. Chomo. 1994. Resource supplies and changing world agricultural
comparative advantage. Agricultural Economics 10 (3):207-217.
Tol, R. S. J., and G. W. Yohe. 2007. The weakest link hypothesis for adaptive capacity:
An empirical test. Global Environmental Change 17 (2):218-227.
Turner, B. L., R. E. Kasperson, P. A. Matson, J. J. McCarthy, R. W. Corell, L.
Christensen, N. Eckley, J. X. Kasperson, A. Luers, M. L. Martello, C. Polsky, A.
Pulsipher, and A. Schiller. 2003. A framework for vulnerability analysis in
sustainability science. Proceedings of the National Academy of Sciences of the
United States of America 100 (14):8074-8079.
USDA. 2010. Briefing Rooms: Rice. USDA Economic Research Service. Available at
http://www.ers.usda.gov/Briefing/Rice/ (Last access 23 August 2010).
Wagner, P. F., R. ONeill, L. T. Tran, M. Mehaffey, T. Wade, and E. R. Smith. 2006.
Regional Vulnerability Assessment for the Mid-Atlantic Region, ReVA, US EPA.
Wallis, T. W. R., and J. F. Griffiths. 1995. An assessment of the weather generator
(WXGEN) used in the erosion/productivity impact calculator (EPIC). Agricultural
and Forest Meteorology 73 (1-2):115-133.
Wandiga, S.O., M. Opondo, D. Olago, A. Githeko, F. Githui, M. Marshall, T. Downs, A.
Opere, P.Z. Yanda, R. Kangalwe, R. Kabumbuli, E. Kirumira, J. Kathuri, E. Apindi,
L. Olaka, L. Ogallo, P. Mugambi, R. Sigalla, R. Nanyunja, T. Baguma, and P.
Achola. 2008. In Climate Change and Vulnerability. eds. Leary, N., C. Conde, J.
Kulkarni, A. Nyong, J. Adejuwon, V. Barros, R. Lasco, and J. Pulhin. 2009..
Earthscan Publications Ltd.
Wassmann, R., S. V. K. Jagadish, S. Heuer, A. Ismail, E. Redona, R. Serraj, R. K. Singh,
G. Howell, H. Pathak, and K. Sumfleth. 2009. Chapter 2 Climate Change Affecting
Rice Production: The Physiological and Agronomic Basis for Possible Adaptation
Strategies. Advances in Agronomy. 101. 59-122. Academic Press.
Wassmann, R., Y. Hosen, K. Sumfleth. 2009. Reducing methane emissions from irrigated
rice. In Agriculture and climate change: An agenda for negotiation in Copenhagen.
Wassmann, R., S. V. K. Jagadish, K. Sumfleth, H. Pathak, G. Howell, A. Ismail, R.
Serraj, E. Redona, R. K. Singh, and S. Heuer. 2009. Chapter 3 Regional
Vulnerability of Climate Change Impacts on Asian Rice Production and Scope for
Adaptation. In Advances in Agronomy, 102: 91-133. Academic Press.

204
Wassmann R. 2010. Implementing the Clean Development Mechanism in the land use
sector: status and prospects. In No regret options for adaptation and mitigation
and their potential uptake. ed. R. Wassman. IRRI Limited Proceedings No. 16. Los
Baos (Philippines): International Rice Research Institute. p 63-66.
Wassmann R, S.V.K. Jagadish, S.B. Peng, K. Sumfleth, Y. Hosen, and B.O. Sander.
2010. Rice production and global climate change: scope for adaptation and
mitigation activities. . In No regret options for adaptation and mitigation and their
potential uptake. ed. R. Wassman. IRRI Limited Proceedings No. 16. Los Baos
(Philippines): International Rice Research Institute. p 67-77.
Watson, J.G. 2011. Principles, assumptions, and data requirements for principal
component analysis and chemical mass balance receptor models.
http://gate1.baaqmd.gov/pdf/1328_Principles_Assumptions_Data_Requirements_Pri
ncipal_Component_Analysis_Chemical_Mass_Balance_Receptor_Models.pdf (Last
access 15 August 2011).
Wattanakij, N. W. Thavorntam, and C. Mongkolsawat. 2006. Analyzing spatial pattern of
drought in the Northeast of Thailand using Multi-Temporal Standardized
Precipitation Index (SPI). In Proceeding of the Asian Association on Remote
Sensing, 9-13 October 2006, Ulaanbataar, Mongolia.
Welch, J. R., J. R. Vincent, M. Auffhammer, P. F. Moya, A. Dobermann, and D. Dawe.
2010. Rice yields in tropical/subtropical Asia exhibit large but opposing sensitivities
to minimum and maximum temperatures. Proceedings of the National Academy of
Sciences 107 (33):14562 -14567.
White, D. M. Richmand, and B. Yarnal. 1991. Climate regionalization and rotation of
principal components. International journal of climatology 11: 1-25.
Wilbanks, T. J., and R. W. Kates. 1999. Global Change in Local Places: How Scale
Matters. Climatic Change 43 (3):601-628.
Wilbanks, T. J. 2006. Chapter 2: How scale matters: Some concepts and findings. In
Bridging Scales and Knowledge Systems: Concepts and Applications in Ecosystem
Assessment, eds. F. Berkes, D. Capistrano, W. Reid, and T. Wilbanks, 21-36.
Washington, D.C.: Island Press.
Williams, J.R., C.A. Jones, and P.T. Dyke. 1984. A modeling approach to determining
the relationship between erosion and soil productivity. 1984. Trans. Am. Soc. Agric.
Eng. 27(1): 129-144.

205
Williams, J.R., C.A. Jones, J.R. Kiniry, and D.A. Spanel. 1989. The EPIC crop growth
model. Trans. Am. Soc. Agric. Eng. 32: 497-511.
Williams, J. R., P. T. Dyke, W.W. Fuchs, V. W. Benson, O. W. Rice and E. D. Taylor.
1990. EPIC--Erosion/Productivity Impact Calculator: 2 User Manual. A. N.
Sharpley and J. R. Williams, ed's. U.S. Department of Agriculture Technical
Bulletin No. 1768. 127 pp.
Wolfe, D.W. 1994. Physiological and growth responses to atmospheric carbon dioxide
concentration. In Handbook of Plant and Crop Physiology. ed. Pessarakli M. Marcel
Dekker, New York. pp. 223-242.
Wopereis, M. C. S., T. Defoer, P. Idinoba, S. Diack, and M.-J. Dugue. 2009.
Participatory learning and Action Research (PLAR) for Integrated Rice
Management (IRM) in inland valleys of Sub-Saharan Africa: Technical Mannual.
Cotonou, Bennin: Africa Rice Center.
Wreford, A., D. Moran, and N. Adger. 2010. Climate change and agriculture: Impacts,
adaptation and mitigation. OECD Publishing.139 p.
Van De Geijn, S.C. and Goudriaan, J. 1996. Chapter 5: The effects of elevated CO2 and
temperature change on transpiration and crop water use. In Global climate change
and agricultural production: direct and indirect effects of changing hydrological
soil and plant physiological processes. eds. F. Bazzaz and Sombroek, W. FAO:
Italy.
Vsquez-Len, M., C. T. West, and T. J. Finan. 2003. A comparative assessment of
climate vulnerability: agriculture and ranching on both sides of the US-Mexico
border. Global Environmental Change 13 (3):159-173.
Yarnal, B. 1993. Synoptic climatology in environmental analysis: A primer. John Wiley
& Sons: London. 256 p.

VITA
Ratchanok Sangpenchan
rus174@psu.edu
Education

Penn State University, Geography, M.Sc., 2009


Sripatum University, Information Technology, M.S.IT. 1998
University of Kasetsart, Integrated Pest Management, B.Sc., 1996

Professional Experience
1996 2006 Researcher, Information and Technology Transfer Department at
Kasetsart Agricultural and Agro-industrial Product Improvement
Institute (KAPI), Kasetsart University, Thailand.

1997-2000

Deputy Secretary of KU-JICA Research Project,


Kasetsart University, Thailand

Você também pode gostar