Você está na página 1de 12

Environmental Modelling & Software xxx (2012) 1e12

Contents lists available at SciVerse ScienceDirect

Environmental Modelling & Software


journal homepage: www.elsevier.com/locate/envsoft

A dynamic assessment tool for exploring and communicating vulnerability


to oods and climate changeq
Carlo Giupponi a, b, *, Silvio Giove a, b, Valentina Giannini a, b
a
b

Universit Ca Foscari di Venezia e Dipartimento di Economia, S. Giobbe 873, 30121 Venezia VE, Italy
Fondazione Eni Enrico Mattei and Centro Euro-Mediterraneo sui Cambiamenti Climatici e Climate Impacts and Policy Division, Italy

a r t i c l e i n f o

a b s t r a c t

Article history:
Received 30 August 2011
Received in revised form
8 May 2012
Accepted 9 May 2012
Available online xxx

In this article we propose an innovative approach to support a participatory modelling process for the
exploratory assessment of vulnerability within the broad context of climate change adaptation. The
approach provides a simplied dynamic vulnerability model developed within a conceptual model
adopted e but very rarely made operational e by many international organisations such as the
Intergovernmental Panel for Climate Change, the European Union. We propose a procedure in which
disciplinary experts and local actors interact for the identication of the most relevant issues with
reference to a specic vulnerability problem. Local actors (e.g. representatives of public administrations,
business, NGOs) identify the most relevant issues related to the various dimensions of vulnerability, to be
considered as input variables to contextualise the generalised model in the study case. Quantitative
indicators are provided by disciplinary experts to describe past and future trends of variables, and their
trajectories are combined to explore possible future vulnerability trends and scenarios. A non additive
aggregation operator is proposed to allow experts and actors to pro vide their preferences through ad hoc
questionnaires, thus overcoming the oversimplications of most of the current vulnerability indices,
which are usually either additive (fully compensatory) or multiplicative (non compensatory), and
providing transparent and robust management of subjectivity and analysis of the deriving variability and
uncertainty in model outputs. Input data for the demonstration of the model derive from the European
Project Brahmatwinn, with reference to the Assam State in India.
 2012 Elsevier Ltd. All rights reserved.

Keywords:
Vulnerability assessment
Participatory modelling
Multiple criteria
Flood
Climate change

Software availability
The model presented in this paper has been developed with the
Simile software by Simulistics, a system dynamics and object-based
modelling and simulation software for complex dynamic systems
in the earth, environmental and life sciences. A free edition of the
package can be downloaded at: http://www.simulistics.com/
products/simile.php. The visual interface of the model is provided
in Fig. 4, while the corresponding Simile (sml) le can be obtained
from the corresponding author.
1. Introduction
The Himalayan region, characterised by heavy seasonal precipitations (monsoons), is one of the areas in the world most
q Thematic Issue on the Global Change Modelling.
* Corresponding author. Universit Ca Foscari di Venezia e Dipartimento di
Economia, S. Giobbe 873, 30121 Venezia VE, Italy. Tel.: 39 041 234 9126; fax: 39
041 234 9176.
E-mail address: cgiupponi@unive.it (C. Giupponi).

vulnerable to ood hazards. The frequency of extreme events has


been increasing in recent years and is projected to increase even
more in the future due to climate changes (Mirza, 2010). Physical
factors affecting the vulnerability of the area are both natural and
socio-economic (Shrestha and Aryal, 2010), the former include the
shrinking of glaciers and the risk for Glacial Lake Outburst Floods
(GLOF), and river discharges as affected by the seasonality of
precipitations, the latter include the impact on hydroelectric
energy production, irrigation, and a cascade of wider effects on
local societies (e.g. side effects on health).
The management of natural resources (NR) requires capabilities
for integrating different disciplinary expertises including, but not
limited to, specic skills to deal with the complexity of dynamic
social and ecological systems.
The concept of integration of disciplines and tools in the eld of
water management nds its denition in what has been described
by the Global Water Partnership as Integrated Water Resources
Management (IWRM): a process which promotes the coordinated
development and management of water, land and related
resources, in order to maximize the resultant economic and social

1364-8152/$ e see front matter  2012 Elsevier Ltd. All rights reserved.
doi:10.1016/j.envsoft.2012.05.004

Please cite this article in press as: Giupponi, C., et al., A dynamic assessment tool for exploring and communicating vulnerability to oods and
climate change, Environmental Modelling & Software (2012), doi:10.1016/j.envsoft.2012.05.004

C. Giupponi et al. / Environmental Modelling & Software xxx (2012) 1e12

welfare in an equitable manner without compromising the


sustainability of vital ecosystems (GWP, 2000). IWRM now has to
consider also a relatively new dimension: global change.
Within the broadest context of global change studies, an issue of
very high relevance for decision and policy makers in the eld of NR
Management (NRM) is the analysis of adaptation strategies to cope
with climate change phenomena and their current manifestations
as impacts of climate variability and extreme events. New paradigms incorporating the adaptive dimension, which takes into
account future scenarios of change, have been proposed, such as
the concept of Adaptive and Integrated Water Management (AIWM;
see Pahl-Wostl et al., 2010). According to the glossary of the IPCCWG2 (Parry et al., 2007), adaptation should be dened as an
adjustment in natural or human systems in response to actual or
expected climatic stimuli or their effects, which moderates harm or
exploits benecial opportunities. Of great interest here is the fact
that adaptation involves both natural and human systems, which
can be seen as coupled complex adaptive systems (Rammel et al.,
2007) that, according to Berkes and Folke (1998), can be dened
as, or simply socio-ecosystems (SESs). This emphasises the need for
integrated (holistic) and dynamic analysis and assessment tools in
support to climate change adaptation (CCA) efforts, which are
commonly designed through policies and measures aiming in
general at reducing the vulnerability of those systems and at
increasing the sustainability of SESs (Ostrom, 2009).
Vulnerability is dened in many different and contradictory
ways in the literature (see Fssel and Klein, 2006 for an introduction to the main schools of thought, recently updated in Chapter 2
of the IPCC SREX Report; Field et al., 2012). Given the major role
played by the Intergovernmental Panel for Climate change (IPCC) in
the context of CCA we used as a main terminological reference the
most recent IPCC Reports, and in particular the contribution of
Working Group II to the Fourth Assessment Report of the IPPC
(Parry et al., 2007), and the Special Report on CCA, extreme events
and disasters (Field et al., 2012), made whenever possible consistent with the terminology adopted by the Brahmatwinn Consortium (Hutton et al., 2011). Accordingly, vulnerability is here dened
as the propensity or predisposition of a system to be adversely
affected by a specic stressor, and, more specically, the degree to
which the SESs of the case study region may be susceptible to, or
unable to cope with, the adverse effects of climate change on ood
risk. Adaptive capacity is one of the components of vulnerability, to
be increased with CCA, resulting from the combination of the
strengths, attributes, and resources available to society and its
various components, that can be used to prepare for and undertake
actions to reduce adverse impacts, moderate harm, or exploit
benecial opportunities. CCA can thus be intended specically as
the process aimed at moderating harm or exploit benecial
opportunities of the evolving climate and its interactions with the
other main drivers of natural or anthropogenic origin.
In an uncertain future, projections on how vulnerability might
change are needed to understand which adaptation strategies may be
preferable in any given situation. Here science plays a very important
role in supporting policy makers in taking decisions in the elds of
CCA and IWRM within the broader context of sustainability science,
but interactions and mutual learning among the various actors
(disciplinary experts, policy makers, lay people, etc.) is of crucial
importance (Tabara and Pahl-Wostl, 2007). System thinking,
modelling, and visual, besides effective written and verbal, communication means can be considered as fundamental competences to
facilitate improved integration across disciplines (Jeffrey, 2003).
Disciplinary experts have developed different kinds of mathematical models for the simulation of the various components of the
system, typically a river basin: hydrologic models for simulating the
water cycle, land use models for the allocation of land cover patches

and related activities, physiological models for the simulation of


vegetation growth, etc. More recently, attempts were made to
integrate disciplinary models for the development of integrated
modelling environments, and, in particular, Integrated Assessment
Models (IAM; Parker et al., 2002). Moreover, stakeholders participation, at the core of IWRM principles, has gained an ever
increasing role, supported also by a series of environmental regulations and directives in Europe and elsewhere, such as the Water
Framework Directive (EU-WFD; EC, 2002), and have pushed the
research community to develop suitable methods and operational
tools for managers and practitioners. The culture of managing
decision processes through an effective involvement of all the
interested actors is thus becoming increasingly widespread and
nding applications in the real world. Examples are available also in
the Himalayan region (see for example Kelkar et al., 2008), but
usually the integration of scientic components (i.e. hydrologic and
other sorts of modelling) and participatory processes is not fully
developed in functional formalisation.
A new research stream e participatory modelling e has developed in recent times bridging modelling and participation, with the
ambition to provide methods and tools for the development of
IAMs with the active contribution of involved stakeholders
(Siebenhuner and Barth, 2005). Integrated and participative models
are often merged with other tools specically developed for the
management of the process of decision/policy-making and
commonly ascribed to the category of Decision Support Systems or
simply DSS tools (van Delden et al., 2011). These systems provide
comprehensive methodological background and toolboxes targeted
to practitioners to help them plan and implement IWRM principles,
for example within the context of the River Basin Management
Plans (RBMP) required by the EU-WFD (Giupponi, 2007; Giupponi
and Sgobbi, 2007; Rammel et al., 2007).
Many reasons support the efforts for a stronger relationship
between science and policy-making, the main one possibly being
the ambition to enhance the impact and value of the output of
research for our societies. Communication is crucial and it must be
managed and planned carefully through decision-making
processes (Vogel et al., 2007). Prerequisites for improved communication between the scientic and policy communities are the
shared conceptualisation of the problem in question and the use of
common communication tools and languages. Unfortunately, such
requirements have usually been neglected by both communities
and the current situation is that different groups (e.g. politicians,
researchers, but also different disciplinary groups) have different
interpretations, reference models and languages. Shared conceptual models could greatly contribute to more effective collaboration
and participation (Otto-Banaszak et al., 2010) and a signicant
contribution in this direction could come from researchers developing scientic approaches such as mathematical models within
the framework of the existing conceptual models adopted by policy
makers, and making use of effective graphical interfaces (Dennison,
2008). The case of interest for this work is the conceptualisation of
a model for climate change vulnerability assessment. This case can
be considered emblematic of what stated above, since many
different models exist as a consequence of the diverse interpretations of both the problem itself and of the main keywords. There is
one model, however, that is adopted in many policy references in
the eld of CCA, such as the specic European White Paper
(Commission of the European Communities, 2009a), and the Fourth
Assessment Report of the IPCC (Parry et al., 2007). That model also
nds theoretical support in the CCA literature, one of the most cited
papers being the one by Fssel and Klein (2006), in which this
conceptual framework is described in detail. Fig. 1 depicts the main
elements of the conceptual model for vulnerability to climate
change impacts.

Please cite this article in press as: Giupponi, C., et al., A dynamic assessment tool for exploring and communicating vulnerability to oods and
climate change, Environmental Modelling & Software (2012), doi:10.1016/j.envsoft.2012.05.004

C. Giupponi et al. / Environmental Modelling & Software xxx (2012) 1e12

Fig. 1. Conceptual model for vulnerability to climate change impacts, according to the
denitions provided in the AR4 of the IPCC (Parry et al., 2007). Mitigation strategies
focus on limiting emissions, while adaptation is aimed at limiting potential impacts
and vulnerability or increasing adaptive capacity.

Surprisingly, notwithstanding the broadest literature on


vulnerability, not many authors attempted the development of
operational assessment tools and some even declared that
vulnerability simply cannot be quantitatively assessed. We believe
instead that vulnerability assessment tools developed at the
interface between science and policy making could have
a tremendously positive impact towards scientically sound and
socially acceptable adaptation planning and implementation.
The present work is aimed at presenting an operational model
for vulnerability assessment developed upon the conceptual model
reported in Fig. 1, and demonstrating its applicability in a participatory context set up within the activities of the Brahmatwinn
Research Project,1 which was aimed at enhancing capacity to
develop response options for Integrated Water Resources
Management (IWRM) in headwater basins of alpine mountain
massifs (Danube and Brahmaputra). In particular, this paper reports
further developments based upon the nal results of the project,
with specic focus on the assessment of vulnerability to ood risks
under the pressure of expected climatic changes in the Brahmaputra River Basin, and specically in the area of the Indian state of
Assam. It utilises data acquired during the project complemented
with further efforts carried out after its completion, to explore
innovative approaches for operational vulnerability assessment.
In the following section we present the vulnerability conceptual
model (2.1) and its implementation in the case study (2.2), and then
we move to its mathematical formalisation, focusing in particular
on the aggregation algorithm (2.3). Section 3 presents the results of
the explorative application of the operational model, including
sensitivity and uncertainty analyses. Section 4 discusses the
outcomes of the work and provides concise conclusions, identifying
also future research needs.
2. Materials and methods
2.1. The conceptual model for vulnerability assessment
Participatory modelling techniques were adopted in the Brahmatwinn Project to
support the assessment of vulnerability in the context of water resources

1
Twinning European and South Asian River Basins to enhance capacity and
implement adaptive management approaches (BRAHMATWINN; Project no: GOCE
-036952). Research funded by the European Community, SUSTDEV-2005-3.II.3.6:
Twinning European/third countries river basins.

management, facing the projected future impacts of climate change. To


ensure research would address end users e in this case mainly local administrators e
needs, the participation of stakeholders was facilitated through a series of workshops,
which were organized to explore pressing issues and identify possible response
options. The research consortium was made of 17 partners, coming from different
disciplinary elds: climatology, hydrology, geography, social sciences, glaciology,
ecology, law, and informatics. The main disciplines needed to develop response
options for IWRM were thus represented, but so were also the usual problems of
integration of diverse disciplinary elds and communication between scientists and
policy makers. One of the rst issues was then the declaration of the main elements for
communication, i.e. the operational denitions of the main concepts.
According to the denitions provided in the introductory section, in the model
adopted here vulnerability is a function of the character, magnitude, and rate of
climate change and variation to which a system is exposed, its sensitivity, and its
adaptive capacity, resulting from the operational combination of indicators or
indices of potential impacts, representing the consequences of climate change on
natural and human systems and of adaptive capacity, representing the ability to
plan, prepare for and implement adaptation measures. Potential Impacts derive in
turn from the combination between (i) the exposure to ood risk, i.e. in general the
degree, duration, and/or extent to which the system is in contact with, or subject to,
a perturbation related to climatic stimuli, here quantied through the trajectories of
climatic drivers signicant for ood risks, and (ii) the sensitivity of the socioecosystem, i.e. the degree to which a system is affected, either adversely or benecially, by climate-related stimuli. Adaptive capacity here represents the ability of
a system to adjust to climate change (including climate variability and extremes) to
plan, prepare for and implement adaptation measures to moderate potential
damages, to take advantage of opportunities, or to cope with the consequences.
Factors that determine adaptive capacity of human systems include economic
wealth, technology and infrastructure, information, knowledge and skills; it reects
the ability of a system to change in a way that makes it better equipped to deal with
external inuences, by exploiting both its socio-economic coping capacity and
intrinsic resilience. These two indexes are used here in the effort to develop an
operational formalisation of the conceptual model. Resilience has become a very
frequent keyword in the most recent policy documents on CCA (e.g. the EU White
Paper on CCA; Commission of the European Communities, 2009b). Unfortunately,
and similarly to vulnerability, it is also a widely debated concept, whose endless
terminological debate is completely out of the scope of this paper. Resilience is here
dened as the capacity of the socio-ecological system to absorb disturbance and
reorganize while undergoing change so as to still retain essentially the same functions, structure, identity, and feedbacks. Socioeconomic coping capacity is dened as
the ability to plan, prepare for and implement measures, i.e. societys tactical ability
to predict and to act in advance of the hazard, to take steps to counter pressures and
threats, and the capability to recover after being subject to threat or pressure. Factors
that determine the coping capacity of human systems include economic wealth,
knowledge and skills.
In order to make the conceptual model described above and presented in Fig. 1
to develop into an operational tool for dynamic simulations, a series of actions are
needed with the coordinated involvement of both researchers and local actors, as
reported in the ow-chart of Fig. 2. Three main issues emerge on this regard:
i)
ii)
iii)

The identication of data sources for the quantication of input variables;


The mathematical formalisation of causal links, and in particular the aggregation algorithms, including the required normalisation procedures;
The elicitation of experts knowledge for consolidating the implementation
of the generalised model in the application case and for the inclusion of
stakeholders views and preferences, when required.

The three following sections will focus on the issues listed above, with specic
reference to the Brahmatwinn project for what concerns the identication of suitable indicators for vulnerability assessment (i), and with the aim of proposing an
operational approach (ii and iii) that could be adopted also elsewhere.

2.2. Case study setting of the vulnerability assessment model


Fssel (2007) developed a conceptual framework of vulnerability with six
fundamental dimensions: temporal reference, sphere (or scale), knowledge domain,
vulnerable system, attribute of concern, and hazard. In this work the knowledge
domain is the socio-ecosystem, in which the human and the natural components are
functionally coupled, and accordingly we include both socio-economic and environmental attributes of concern. We follow that framework to dene the assessment
context of this work. In particular, we focus on the knowledge domain of waterrelated hazards, and specically on oods and related concerns for local livelihood; as a temporal reference we consider a time frame of interest for exploring
long-term dynamics in which climate change may play a relevant role; the sphere of
the application considers both endogenous factors of the vulnerable socioecosystem of the Assam valley and exogenous factors, namely climate change drivers.
The identication of a comprehensive set of indicators for vulnerability
assessment has been dealt with by various authors. Fssel (2010) proposes

Please cite this article in press as: Giupponi, C., et al., A dynamic assessment tool for exploring and communicating vulnerability to oods and
climate change, Environmental Modelling & Software (2012), doi:10.1016/j.envsoft.2012.05.004

C. Giupponi et al. / Environmental Modelling & Software xxx (2012) 1e12

Fig. 2. Flow chart of the activities for the implementation of the vulnerability
assessment model.
a comprehensive system of indicators to be used at the country level, referred to
a conceptual framework close to the one adopted here, based upon the concepts of
sensitivity, exposure, impacts and capacity. In analogy with what proposed by Fssel
we adopt here a hierarchical system of indicators, but we consider also temporal
dynamics and we focus on the issue of aggregating indicator values within
a comprehensive concept of vulnerability and an overall conceptual and visual
framework (see Fig. 1), consistent with relevant policy documents mentioned in the
introductory section. In line with the current literature on vulnerability indices, the
resulting vulnerability index is a non-dimensional measure useful for comparative
analysis, but while the available indices are usually utilised for spatial comparisons
(typically atlases of national or sub-national vulnerabilities), the proposed approach
is aimed at inter-temporal comparisons, thus providing policy makers with a means
to explore past and expected trajectories and thus support the development of CCA
strategies, making use of available quantitative indicators.
In this work the selection of indicators was based upon two main objectives:
providing dynamic and quantitative information for the quantication of the nodes
of the conceptual framework presented in Fig. 1, and balancing the assessment
according to the three main dimensions of sustainable development (environment,
society and economy). Concerning data sources, disciplinary analyses and modelling
carried out by the research consortium produced a wealth of indicators, which
enable quantitative approaches to be applied for the assessment of issues identied
by stakeholders in the participatory process organised during the project implementation through a series of workshops. An innovative synoptic framework was
produced during the project to facilitate the matching between the indicators and
local issues, and thus the communication between researchers and local actors
(LAs), allowing the linking of each indicator proposed by researchers with the
related issues raised by LAs, within a comprehensive Knowledge Integration
Table (KIT; see Giannini and Giupponi, 2011 for details). The KIT is built upon a series
of hierarchical levels: four Themes (Environment, Society, Economy and Governance) are subdivided into a series of Domains (e.g. Forests; Infrastructures;
Livelihood), and further split into Sub-domains (e.g. Forest management; Poverty).
Sub-domains represent the key elements of the KIT, since they are the interface of
a two-way relationship connecting the scientic knowledge provided by the
research consortium in the form of indicators (one or more for each sub-domain) to
the issues and the needs expressed by the local actors involved in the participatory
process of the project. In total the table presents 49 sub-domains to which 190
indicators are allocated, providing possible solutions to the assessment of the state
of local SESs (Giannini and Giupponi, 2011).
One of the projects workshops, held in Kathmandu (November 2008), was
aimed at identifying priority sub-domains in the Upper Brahmaputra River Basin
(see Ceccato et al., 2011 for details). Participants were asked to select the most
relevant sub-domains out of three comprehensive lists e economic (ECON), environmental (ENV) and social (SOC) e extracted from the KIT. Each participant was
asked to select the rst, second and third most relevant sub-domains to be used as
evaluation criteria for the analysis of alternative CCA strategies to cope with ood
risk in the Study area, out of the three lists by assigning decreasing scores (3e1). The
nine most important sub-domains identied by summing up the scores of local

actors are: Poverty; Population dynamics; Infrastructure pressures (social dimension); Employment; Energy production; Agricultural production (economics); Basin
morphology; Vulnerability; Forest management (environment).
Those results were utilised as the basis for the explorative implementation of the
proposed assessment model in the case study, with vulnerability selected as the output
of the model, while the other sub-domains are used as independent variables to
operationalise the conceptual model as reported in Fig. 3 in which the model framework depicted in Fig. 1 is implemented in a dynamic system analysis (Sterman, 2000)
modelling environment2 for scenario simulations with policy makers. In this way
a bidirectional ow of information, i.e. between the preferences and priorities
expressed by local actors and the evidences acquired by the researchers involved in the
project, was implemented and framed within a widely adopted policy framework.3
As depicted in Fig. 3, three of the sub-domains selected by local actors were
connected to the quantication of sensitivity to ood risk: Basin Morphology, here
described as the morphology of the river bank and its maintenance (a bad basin
morphology leads to negative impacts towards sensitivity); Infrastructure pressure,
here dened as the sensitivity induced by alterations of the river ow deriving from
barrages, dams, etc. which may increase the sensitivity, in particular in an area with
high seismicity (a bad infrastructure pressure leads to negative impacts towards
sensitivity); and Population dynamics here associated to the number of people
exposed to ood risk. Two sub-domains out of those selected at the Brahmatwinn
workshops were used to contribute to the concept of socio-ecosystem resilience:
Agricultural production and Forest management, for their signicant contribution
to build resilience of the territory potentially affected by ood events. Socioeconomic coping capacity was in turn considered as being positively affected by good
status of the three remaining sub-domains: Employment rate; Energy production;
and Poverty reduction.
The most suitable indicators in terms of appropriateness with respect to the case
study considered and data availability were selected out of the KIT to describe the
trends of the sub-domains of interest over the past and also in future projections.4 In
general, the trajectories of indicators were based on available information and
projections for the years 1980, 2000, 2020 and 2050. Whenever specic projections
were available, two distinct trajectories were developed for each indicator to
represent the trends according to the A1B and B1 IPCC-SRES scenarios (Nakicenovic
and Swart, 2000). The list of sub-domains and selected indicators is shown in
Table 1. Moreover, in order to provide a quantitative trajectory of climate-related
stimuli to be used for the quantication of the exposure to ood risk, another
indicator was extracted from the modelling outcomes of the project (Dobler et al.,
2011) and taken from the climate domain of the KIT, for its specic relationship
with extreme ooding events: the yearly maximum 5-day precipitation (PX5D).
Worth to remember here that, given the exploratory character of the work, suboptimal proxies were considered acceptable for the sake of demonstrating the
implementation of the tool, and a single indicator per sub-domain was used.
At this stage the problem of indicator aggregation into intermediate and nal
nodes of the conceptual model for vulnerability assessment emerged.
2.3. Aggregation operators
The vulnerability concept, like other terms such as sustainability are
commonly adopted and interpreted in the current language, but they are not easy to
conceptualise and formalise into a unique and commonly accepted framework, as
stated above. Beyond some rationality properties, such as monotonicity, which avoid,
ceteris paribus, the decreasing of the overall satisfaction degree (the aggregated
value) if a (benet) criterion increases, it is indeed impossible to provide a single
mathematical formula that objectively computes its value. Vulnerability reects
neither a physical variable (which could be objectively measured, despite possible
error measures) nor the output of a well-dened static or dynamic model (described,
for instance, by differential equations) that assumes a theoretical and practical sense
and is univocally interpreted. Because of that, no model can measure strictu sensu the
vulnerability, and its assessment requires consideration of both quantitative and
qualitative information and subjective judgements (Field et al., 2012).
The aim of the approach described below is to go beyond the fully compensatory
normalised additive procedure commonly adopted by the current literature for
concise vulnerability indices. In this direction, what can be done is the assessment of
a coherent subjective probability, see (Coletti and Vantaggi, 2006), also known as
a fuzzy possibility distribution. In fact, given the multifaceted nature of vulnerability,

2
The dynamic assessment model has been developed within the modelling tool
Simile by Simulistics (http://www.simulistics.com/).
3
We considered this as being acceptable given the methodological and explorative character of the work, but in case of future application of the proposed
approach, the identication of the descriptive variables for vulnerability assessment
would require a much deeper and interactive collaboration of all the disciplinary
experts together with LAs.
4
The authors gratefully thank in particular Prof. Nayan Sharma (Indian Institute
of Technology Roorkee) for providing most of the estimations for the trajectories of
indicators.

Please cite this article in press as: Giupponi, C., et al., A dynamic assessment tool for exploring and communicating vulnerability to oods and
climate change, Environmental Modelling & Software (2012), doi:10.1016/j.envsoft.2012.05.004

C. Giupponi et al. / Environmental Modelling & Software xxx (2012) 1e12

Fig. 3. Preliminary formalisation of the vulnerability assessment model.

fuzzy logic seems to be particularly appropriate for the management of this type of
problem, even if not the only possible approach. In this context, all the (normalised)
factors that are used to compute the vulnerability index are transformed into real
numbers between zero and one (normalisation phase), and can be subsequently
treated as fuzzy variables. To aggregate them into a single value (usually moving
bottom-up in the decision tree), suitable aggregation algorithms need to be selected,
in accordance with the logic of the conceptual model, but also according to the
elicited preference of the decision makers (DMs). The elicitation of preferences can
be a complex and time-consuming task. We do not address this aspect in depth (see
Meyer and Pointhire, 2011 for details), but we simply remark a general rule: as soon
as the complexity increases (trying to include a wide range of possible preference
structures), the number of parameters increases (in some cases exponentially), and
the identication procedure becomes challenging, and the questionnaire used to
elicit information from DM would require too many questions, thus making it
difcult if not impossible to deal with all of them at once. This is one reason why
simplistic approaches are applied in many cases, for example simple additive or
multiplicative combinations of e in some case weighted e normalised values of
indicators.
Multi Criteria Theory and in particular Multi Attribute Value Theory (MAVT;
Belton and Stewart, 2002), suggest many different methods for the elicitation of the
preference structure (and of the relative parameters). But again, for what stated
above, no a priori reason exists to afrm that one method is better than another; the
choice depends on the characterisation of the mathematical properties required,
which also reect the attitude of the DMs, such as their tendency towards pessimism
or optimism. In order to move to our methodological proposal, let us recall rst that
an Aggregation Operator is a non-decreasing multi-values function with some
properties which can be used to obtain an aggregate score; among them, the most
commonly used are the weighted averages (WA) with the EWA (Equally Weighted
Average) as a particular case of WA with equal weights, that are specic cases of the
quasi-arithmetic means operators (Grabish et al., 2009).
A preliminary step for the aggregation of indicators is normalisation. Even if
(possibly) sensitive to outliers, for the sake of simplicity in this study the Min-Max
normalisation algorithm was applied to the values of indicators collected with
different units (e.g. US$, km2, etc.), in which the extreme values were those observed
in the time series considered, and including in the calculation both the scenarios
envisaged in order to maintain the comparability of trajectories; given the wide
spread observed for the variables inside their domain, we feel that for this study the
Min-Max normalisation can be accepted, see (Ebert and Welsch, 2004).
Having calculated the normalised values of the indicators, the aggregation
problem can be formalised according to MAVT as the identication of the algorithm
and parameters which, at each node, aggregate the values of the sub-nodes into
a single value. For what above said, in the present work, we did not select a priori
a single method to aggregate the basic fuzzy variables, and we explored three
different approaches5:

5
We recall that the main scope of the aggregation is to identify an efcient proxy
of the subjective probability expressing the reasoning of DM(s). The vulnerability
index is nothing but the estimation of such probability, that is the (subjective)
probability that the considered scenario is vulnerable, given the values of the
factors (the variables) associated to it at a certain time.
6
The allocation of weights can be done with xed budget allocation (see Belton
and Stewart, 2002. Multiple criteria decision analysis: an integrated approach.
Kluwer Academic Publishers, Boston.) in which participants are asked to allocate
a xed number (budget) of points (100, in this case) amongst the attributes to be
weighted (in this case the selected sub domains) at the participatory workshops.

(i) The simple WA procedure, which requires only the denition of one weight
for each variable,6 and is partially compensative;
(ii) The GA (Geometric Average) approach which, again, requires only the denition of weights, but, in contrast with (i), is totally non compensative;
(iii) A more complex parametric procedure, the Choquet integral with Non
Additive Measures (NAM), able to represent a wide possibility of preference
structures, both compensative or not.
In the case (i), an additive way of reasoning needs to be assumed (or checked),
while (ii) supposes a complete not compensative way of reasoning. Which of the two
is the best proxy of scientic reasoning and DM(s) preference structure for the
vulnerability estimation? The rst one is a linear method, the second one is not,
since it avoids compensation. Anyway, both of them are characterised by a xed
structure (only the weights can vary), while the Choquet integral (iii) is a more
general approach, assigning a suitable weight not only to a single attribute, but also
to any combination of them, making it possible to implement synergic or redundant
interactions among them. By denition the Choquet integral of the vector x1 ; .; xn
with respect to m (a fuzzy measure on A) is dened as:

Cm x1 ; .; xn

n 
 

X
xi  xi1 $m Ai

(1)

i1

Where i indicates that the indices have been permuted so that xi1  xi , and
being Ai fi; i 1; i 2; .; ng, An1 B; x0 0.
The Choquet integral simply adds the marginal gains of each added attribute
considering the NAM associated to the coalition, and, in this sense, it is an extension
of the WA, but a powerful extension, since a suitable tuning of the measures permits
to represent different ways of reasoning. In fact, despite the WA, the Choquet
integral is not necessarily compensatory and can include interactions between
aggregation factors. In particular, if the measure of a subset is greater than the sum
of the measures of one of its partitions, we represent a synergic effect, if equal, no
interaction (i.e. the WA), if lower, a redundancy effect is represented. A limit case of
redundancy is the MAX operator; in this case, it sufces that a single variable
assumes a satisfactory value to render the aggregated values satised (disjoint
effect, or Orness). In the opposite case, we have the MIN operator, where if only one
variable is null, the aggregated value is null (conjoint effect, or Andness). In the
intermediate cases, many other possibilities can be dened, like the OWA (Ordered
Weighted Average) operators (Yager, 1993).
WA appeared to be suitable in particular for expressing the possible compensatory effects of the various components of adaptive capacity (Orness), while the GA
appeared suitable for implementing the Andness concept describing the combination between exposure and sensitivity to produce potential impacts (e.g. with very
high exposure, but no sensitivity, the potential impact should be considered negligible). Both WA and GA require few parameters (the weights assigned to the
intermediate and to the root nodes), but are characterised by a rigid mathematical
structure. Conversely, the NAM with the Choquet integral represents a more
generalised approach and was thus selected for the vulnerability assessment model,
but the price to pay is the assignment of a series of parameters to be elicited through
questionnaires.
2.4. Assignment of aggregation parameters
A suitable questionnaire was designed for the elicitation of aggregation weights,
with one question per aggregation node, asking for the measures (i.e. the aggregation weights) to be assigned to the converging variables in any possible combination
of worst and best levels. Following Despic and Simonovic (2000), where all the [0,1]
combinations of criteria x1 ; .; xn are considered for each node of the aggregation

Please cite this article in press as: Giupponi, C., et al., A dynamic assessment tool for exploring and communicating vulnerability to oods and
climate change, Environmental Modelling & Software (2012), doi:10.1016/j.envsoft.2012.05.004

C. Giupponi et al. / Environmental Modelling & Software xxx (2012) 1e12

Table 1
Sub-domains and indicators.
Sub-domain

Acronym

Denition and selected indicator

Basin morphology

BM

Infrastructure pressures

IP

Population dynamics

PD

Agricultural production

AP

Forest management

FM

Employment

Poverty

Energy production

EP

The morphology of the river bank and its maintenance quantied through river bank dynamics. Indicator:
Potential erosion-prone stream bank line (km2 of eroded areas); source: elaborations by Indian Institute
of Technology Roorkee (IIT-R) for the Brahmatwinn Project.
The sensitivity induced by alterations of the river ow deriving from barrages, dams, etc. which may
increase the sensitivity, in particular in an area with high seismicity. Indicator: Hydroelectrical installed
capacity (MW); source: IIT-R and own elaborations on data from the development plans of the Central
Electrical Authority of India (http://www.cea.nic.in/)
Here associated to the number of people exposed to ood risk, through the proxy indicator Population
projections for the State of Assam (units); source: IIT-R and own elaborations on data from the
Socio-economic scenarios for climate change impacts in India (Teri, MoEF, Defra; downloadable
at http://www.decc.gov.uk/)
An activity that contributes to the maintenance of land with positive potential for limiting the impacts
of oods and facilitating post-disaster recovery. Indicator: Gross irrigated area (km2), a proxy of investments
on agricultural production; source: IIT-R.
One of the most important strategies for controlling runoff and erosion risks, thus limiting the probability
of ood events downstream. Indicator: Per capita availability of managed forest land (km2); source: IIT-R.
One of the dimensions of socio-economic wealth of the community, here measured through the proxy
indicators Number of registered factories; source: IIT-R.
A second index of the economic wealth of the population, here derived from the projections of the indicator
People living in extreme poverty for India; source: A. Calzadilla, 2010. Global Income Distribution and
Poverty: Implications from the IPCC SRES Scenarios. Kiel Working Paper No. 1664.
Another index of the wealth of the community, measured with the indicator Total unit sold (M kWh);
source: IIT-R.

tree; 0 corresponds to the worst case, while 1 for the best one.7 If n is the number
of the indicators converging into a node, the respondents are asked to provide
a score (measure), in this case in the scale [0,1], for 2n  2 questions, which are all the
possible combinations among the criteria converging in the considered node (the
number of question is 2n  2 since the border conditions are already xed) (i.e. the
rst and the last cases, where criteria are respectively all worst e 0,0,0 e and all
best e 1,1,1 e are already set at 0 and 1, respectively). The respondents are thus
asked to assign a numerical score to each row of the matrix containing all the
possible combinations of the criteria chosen at each level.8

1
2
3
4
5
6
7
8

(C1,C2,C3)

Weights

Values

(0,0,0)
(1,0,0)
(0,1,0)
(0,0,1)
(1,1,0)
(1,0,1)
(0,1,1)
(1,1,1)

m(0)
m(1)
m(2)
m(3)
m(1,2)
m(1,3)
m(2,3)
m(1,2,3)

0
?
?
?
?
?
?
1

x(1) < x(2) < . < x(n). Now let us consider 3 criteria (sub-domains) such that n 3; let
(x1, x2, x3) (0.3, 0.8, 0.1); rst of all, we have to order these criteria:
(x(1), x(2), x(3)) (0.1, 0.3, 0.8) since x3 < x1 < x2
The Choquet integral Cm(x1, x2, x3) is thus calculated as follows:
1. (1, 1, 1): x(1) $ m(1, 2, 3) 0.1 $ m(1, 2, 3) 0.1 (which corresponds to x(3))
2. (1, 1, 0): (x(2)  x(1)) $ m(1, 2) (0.3  0.1) $ m(1, 2) 0.2 $ m(1, 2)
3. (0, 1, 0): (x(3)  x(2)) $ m(2) (0.8  0.3) $ m(2) 0.5 $ m(2)
To obtain Cm(x1, x2, x3), it is sufcient to sum up all 3 terms. For our specic case,
as illustrated in Fig. 3, only the case of 2 or 3 variables converging into a node had to
be assessed.
In order to simplify the implementation of the procedure in the simulation
software Simile, the parameters of the Mbius transform are later calculated. There
is a twoeway relation between the non additive measures (m), i.e. the elicited
measures, and the Mbius coefcients (m):
mm S

1st mT

(2)

T4S

In the example with three sub-domains the Mbius coefcients are:


m(1) m(1)

Below an example with three sub-domains (C1; C2; C3) is expanded, in which
the expert would be asked to provide the 23  2 6 values of weights m(1); m(2);
m(3); m(1,2); m(1,3); m(2,3):
Very importantly, the monotonicity principle should be respected, meaning that
the if a combination where only one criterion is best is given a certain measure m,
all combinations including that criterion in the best case should be given
a measure at least equal to m. In practice, for the monotonicity principle line 5 cant
have a measure lower than measures in lines 2 and 3, line 6 lower than 2 and 4, and
line 7 lower than lines 3 and 4.
In case of a greater relevance of one of the sub-domain in determining the state of
the aggregated index (for example lets say that Basin morphology is considered more
important than the others, for determining the Sensitivity), (measures of rows 2; 3;
and 4 could become 0.20; 0.15; 0.15); and in case we thought that good status of basin
morphology combined with a good status of infrastructures could provide synergic
effect the weight in row could be 0.45 (greater than the sum of 0.20 and 0.15).
Let (x1,., xn) be the values of the normalized criteria; rst of all, we need to
order this vector which will become (x(1),., x(n)), in such a way that

Note that the measures should be assigned only for the extreme cases of
combinations of very bad performance (0) and optimal performance (1) of the subdomains considered, while in the calculus the conditions are greater than.
8
Given the impossibility of testing the model with local decision makers because
of the completion of the project, Masters and PhD students were involved in testing
the questionnaire in its subsequent versions and renements.

m(2) m(2)
m(3) m(3)
m(1,2) m(1,2)  [m(1) m(2)]
m(1,3) m(1,3)  [m(1) m(3)]
m(2,3) m(2,3)  [m(2) m(3)]
m(1,2,3) m(1,2,3)  [m(1,2) m(1,3) m(2,3)] [m(1) m(2) m(3)]
where the coalition coefcient m(T) can be both positive, negative or null; if
positive, it means that there is synergic interaction between the criteria (indicators)
belonging to the coalition T while if negative, there is redundancy interaction (or
conicting). If null, no interaction exists.
Using the Mbius coefcients, given that the Choquet integral is computable as:
Cm x1 ; x2 ; .; xn

X
T4N

mm T min fxi g
iT

(3)

with three sub-domains (x1, x2, x3), the Choquet integral is calculated as follows:
Cm(x1, x2, x3) m(1) $ x1 m(2) $ x2 m(3) $ x3 m(1,2) $ min(x1, x2) m(1,3) $ min(x1,
x3) m(2,3) $ min(x2, x3) . m(1,2,3) $ min(x1, x2, x3)

Please cite this article in press as: Giupponi, C., et al., A dynamic assessment tool for exploring and communicating vulnerability to oods and
climate change, Environmental Modelling & Software (2012), doi:10.1016/j.envsoft.2012.05.004

C. Giupponi et al. / Environmental Modelling & Software xxx (2012) 1e12

Fig. 4. System dynamic model adopted for vulnerability assessment. As compared to Fig. 3, uncertainty coefcients are added to each sub-domain and a Choquet weighting
procedure is introduced at each aggregation node.

As mentioned above, to further characterise the respondents attitudes, it is


possible to compute two indexes, depending solely on the measure values: Orness
and Andness (the second being the complement to 1 of the rst). Using the Mbius
values of the measure, the Orness is computed as follows:
Orness

1 X nt
mm T
n  1 T4N t 1

(4)

Independently from the meaning of the aggregated variable, in general, a non


compensative combination is usually associated to a pessimistic attitude, given that
the respondents evaluation is guided only by the worst case, i.e. the logical and
operator. In this case the Orness index shows values close to zero. Vice versa, the
aggregation becomes totally compensative, corresponding to an optimistic attitude, i.e.
the logical or operator, given that only the best case is considered. With Orness 0.5
we denote addictiveness, and thus no interaction among the indicators values.
The Choquet integral applied to the vulnerability model thus provides the
opportunity to introduce sets of aggregation measures allowing the management of
compensation and addictiveness. Moreover, the implementation of the model in
a system dynamic modelling environment allows the selection of the correct
weights at every time step, depending on the specic combination of values of the
indicators to be aggregated.
The results of the implementation of the aggregation procedure in the system
dynamic model are presented in Fig. 4, which, in comparison to Fig. 3, shows the
introduction of the variable for the calculation of the Choquet integral at every
convergence node. Fig. 4 also shows the introduction of uncertainty coefcients,
allowing the assessment and communication of the effects of uncertain values of
future projections of the indicators, a crucial issue when dealing with hydrology and
climate change (Teegavarapu, 2010). In this prototype a random variable is introduced which makes the trajectories of the indicators oscillate around the values
acquired from 0 (past and present) up to 20% in 2050, and simulations are
repeated 100 times. Different degrees of uncertainty could be introduced in case of
the availability of specic information.

9
A1B: A future world of very rapid economic growth, low population growth
and rapid introduction of new and more efcient technology. Major underlying
themes are economic and cultural convergence and capacity building, with
a substantial reduction in regional differences in per capita income. In this world,
people pursue personal wealth rather than environmental quality. B1: A
convergent world with the same global population as in the A1 storyline but with
rapid changes in economic structures toward a service and information economy,
with reductions in materials intensity, and the introduction of clean and resourceefcient technologies. (http://www.ipcc-data.org/sres/ddc_sres_emissions.html).

3. Results
3.1. Preliminary model runs and model sensitivity
The assessment model presented in the previous section has
been implemented with the indicator values reported in Fig. 4 and
Table 1. As previously stated, two implementations were performed
to represent the behaviour of the socio-ecosystem of the Assam
Valley according to the A1B and B1 IPCC-SRES scenarios.9
The trajectories of the normalised values of the selected indicators are reported in Fig. 5. In brief, the information collected for
the selected indicators showed increasing trends for the three
indicators related to sensitivity: expected limited resources for the
maintenance of investments on the basin morphology, increase
with stabilization of infrastructure pressures on the river, and
increase with future stabilization of the population at risk. Agricultural production and forest management contribute to resilience
with opposite trends: the role of investments in agriculture was
positive in the past, but no signicant improvements are expected
in the future. On the contrary forests have been shrinking and are
expected to decrease also in the near future, with projection for
stabilisation and some recovery further on. Trajectories of indicators contributing to coping capacity show positive trends for what
concerns the eradication of poverty, uctuations for the employment and expected increases in terms of energy produced and
consumed in the Assam state. All the indicators for which distinct
projections were available for the two scenarios, showed more
positive trends in the B1 case.
Concerning exposure to ood risk, the selected indicator shows
oscillating values on extreme rainfall events with no signicant
trend for the B1 scenario and an increasing trend (more severe
extremes in the future) for A1B.
In the phases of development and consolidation of the model
and the related materials (in particular the elicitation questionnaire) experts of the research consortium and students were
involved to test the procedures. A class of nine master students
provided the set of Choquet measures for the nal tests of model

Please cite this article in press as: Giupponi, C., et al., A dynamic assessment tool for exploring and communicating vulnerability to oods and
climate change, Environmental Modelling & Software (2012), doi:10.1016/j.envsoft.2012.05.004

C. Giupponi et al. / Environmental Modelling & Software xxx (2012) 1e12


Table 2
Average values of parameters elicited from masters students.
Aggregations of 3 indicators
Values

Measure

Sensitivity

Coping capacity

(0,0,0)
(1,0,0)
(0,1,0)
(0,0,1)
(1,1,0)
(1,0,1)
(0,1,1)
(1,1,1)
Orness

m(0)
m(1)
m(2)
m(3)
m(1,2)
m(1,3)
m(2,3)
m(1,2,3)

0.000
0.187
0.193
0.194
0.448
0.423
0.442
1.000
0.315

0.000
0.206
0.283
0.238
0.508
0.507
0.549
1.000
0.382

Aggregations of 2 indicators

Scenario A1B
Sub-domains

Scenario B1

1980 2000 2020 2050 1980 2000 2020 2050

Sensitivity
basin morphology

0.00

0.00

0.20

1.00

0.00

0.00

0.20

1.00

infrastructure pressure

0.00

0.38

1.00

1.00

0.00

0.38

1.00

1.00

population dynamics

0.00

0.27

0.73

1.00

0.00

0.27

0.63

0.75

agricultural production

0.00

0.87

0.81

0.73

0.00

0.97

0.82

1.00

forest management

1.00

0.38

0.00

0.25

1.00

0.38

0.00

0.25
0.00

Resilience

Coping Capacity
poverty

0.97

1.00

0.19

0.00

0.97

1.00

0.15

employment

1.00

0.41

0.70

0.00

1.00

0.67

0.73

0.71

energy production

0.00

0.97

1.00

1.00

0.00

0.97

1.00

1.00

Fig. 5. Normalised trajectories of climate extremes and values of the indicators selected
as inputs for the vulnerability assessment in the IPCC SRES scenarios A1B and B1.

behaviour and sensitivity analysis. Table 2 reports the average


weights obtained, which were used for a preliminary simulation
without the use of randomly generated uncertainty variables. The
results of the exercise showed an average preference for additive to
synergic effects of aggregation (Orness values oscillating below or
close to 0.5), and as a result of the combination of the indicator
values and the average measures, a very rst projection of vulnerability was calculated on an annual basis (Fig. 6). The rst 30 years
report on the past, while the other 40 years represent the projections into the future. The opposite value of adaptive capacity
(1  AC) was used in the calculations to express the fact that its
effect is making the overall vulnerability to decrease. Values below
zero are therefore possible for the vulnerability, but they dont have
any specic meaning since the tool is intended only for comparative
purposes. According to these results, no remarkable trends are
calculated for the past 30 years, while a rather constant trend
towards increasing vulnerability to oods is projected for the
coming 40 years, with uctuations due to the effect of the climatic
drivers, and this appears to be substantially due to the increase of
the sensitivity of the area to ood risk, combined with a tendency
towards an increase of the extreme climatic events in the future. On
the other hand, the successes expected in terms of poverty eradication in the near future are not adequately supported by the other
socio-economic indicators of coping capacity and by the trends of
resilience indicators.
Given that the values provided by the nine respondents were
remarkably varying (e.g. Orness values ranging from approximately
0.15 to 0.5), a series of parallel model runs were performed by
adopting the preferences of every single respondent, to observe the

Values

Measure

Resilience

Potential
impacts

Adaptive
capacity

Vulnerability

(0,0)
(1,0)
(0,1)
(1,1)
Orness

m(0)
m(1)
m(2)
m(1,2)

0.000
0.344
0.433
1.000
0.389

0.000
0.411
0.361
1.000
0.386

0.000
0.356
0.417
1.000
0.386

0.000
0.411
0.389
1.000
0.400

Note: The sequence of values are as follows: Sensitivity: Basin Morphology;


Infrastructures; and Population. Coping capacity: Employment; Energy; and
Poverty. Resilience: Agricultural production; and Forest Management. Potential
Impacts: Exposure; and Sensitivity. Adaptive Capacity: Resilience; and Coping
Capacity. Vulnerability: Potential Impacts; and Adaptive Capacity.

sensitivity of the model to weights provided by students. The


results reported in Fig. 7 show that the majority of the trajectories
(5 out of 9) vary within an interval usually within 0.1; one of them
(SH5) shows a similar trend, but with lower average values; all of
them provide the same message about vulnerability in the test case
study: a rather constant trend towards higher vulnerability over the
whole period. Also the weights provided by SH1, SH3 and SH5
provide trajectories of increasing vulnerability in the future, but
simulate rather stable or decreasing values until now. The consistency of past trajectories could be theoretically validated by
comparison with existing records of indicators that could be used
as proxies of vulnerability. In this case, for instance, statistics about
damages and casualties due to past oods could be used, but this is
beyond the aims and limits of this work.
3.2. Final model runs and uncertainty
After the preliminary tests of the model reported above, a new
set of weights has been discussed by the authors and a nal set of
model runs has been performed.
The objectives of the new set of runs were to analyse how the
proposed model may allow us to express our understanding of the
conceptual model and of the behaviour of the system and to
observe their effects on the vulnerability trajectories. The preference was to opt for measure close to additiveness for the rst layer
of aggregation, but with different weights assigned to the various
indicators, thus assigning a higher relevance to the indicator
quantifying population at risk for the calculation of exposure, and
the same for the effect of poverty on coping capacity, and forest
management on resilience. At the second level of aggregation we
expressed logical conjunction (low Orness) for the combination of
exposure and sensitivity, or in other words the fact that high
potential impacts arise only when both sensitivity and exposure are
high (the multiplicative operator is here common in the literature).
On the contrary we considered that adaptive capacity can derive
both from good coping capacity and high resilience of the system,
with the rst having a greater role, in that a good social coping
capacity can surrogate in many cases an intrinsic limited resilience
of socio-ecological systems. Finally, we opted again for a synergic

Please cite this article in press as: Giupponi, C., et al., A dynamic assessment tool for exploring and communicating vulnerability to oods and
climate change, Environmental Modelling & Software (2012), doi:10.1016/j.envsoft.2012.05.004

C. Giupponi et al. / Environmental Modelling & Software xxx (2012) 1e12

Fig. 6. Trajectories of the main variables of the model for the IPCC SRES scenario A1B and with average weights.

aggregation of impacts and adaptive capacity to determine the level


of vulnerability, keeping its value low when at least one of the two
components has very low values and making the value jump
whenever both have a relatively bad status. Table 3 reports the
parameters adopted for the model and Fig. 8 shows the results of
100 simulations run in parallel for the two scenarios, thus presenting the effects of the uncertainty variables on the spread of
vulnerability trajectories.
A broad literature analyses the management and communication of uncertainty and it is out of the scope of this work to go into it
in detail. In the present work the inclusion of random variables
associated to every indicator, whose uctuations can be fully parameterised, allows for demonstrating the consideration of such an
important dimension of the problem and presenting its effects on
the outputs (see the spread of the curves in Fig. 8).
The envisaged use of the results of the proposed model for
communicating projections on the vulnerability of a portion of the
Brahmaputra river basin to ood risk in two climate change
scenarios can be described by the following statements:

 Given the lack of a broadly accepted quantitative model for the


vulnerability of socio-ecosystems to climate change,
 Given the availability of a dynamic vulnerability assessment
model for comparative scenario analysis, developed with the
contribution of relevant local actors and experts, upon
a conceptual model widely adopted by policy makers,
 Given that a selected group of experts and local stakeholders
had selected a series of relevant issues for the Upper
Brahmaputra River Basin and identied sub-domains of
knowledge as being the most relevant for the assessment of
vulnerability,
 Given that a research consortium has identied adequate
indicators to quantitatively assess the past and current state
and future trends of the selected sub-domains,
 Given the raw data currently available for the selected indicators and the time frame of interest,
 Given the current knowledge of the possible future trends of
extreme rainfall events in the study area, according to two
climate change scenarios,

Fig. 7. Multiple vulnerability trajectories reporting on the sensitivity of the model to changing weights.

Please cite this article in press as: Giupponi, C., et al., A dynamic assessment tool for exploring and communicating vulnerability to oods and
climate change, Environmental Modelling & Software (2012), doi:10.1016/j.envsoft.2012.05.004

10

C. Giupponi et al. / Environmental Modelling & Software xxx (2012) 1e12

Table 3
Values of parameters agreed by the authors.
Aggregations of 3 indicators
Values

Measure

Sensitivity

Coping capacity

(0,0,0)
(1,0,0)
(0,1,0)
(0,0,1)
(1,1,0)
(1,0,1)
(0,1,1)
(1,1,1)
Orness

m(0)
m(1)
m(2)
m(3)
m(1,2)
m(1,3)
m(2,3)
m(1,2,3)

0.000
0.200
0.250
0.500
0.350
0.650
0.800
1.000
0.458

0.000
0.300
0.150
0.550
0.450
0.750
0.850
1.000
0.508

Aggregations of 2 indicators
Values

Measure

Resilience

Potential
impacts

Adaptive
capacity

Vulnerability

(0,0)
(1,0)
(0,1)
(1,1)
Orness

m(0)
m(1)
m(2)
m(1,2)

0.000
0.350
0.650
1.000
0.500

0.000
0.250
0.050
1.000
0.150

0.000
0.500
0.900
1.000
0.700

0.000
0.200
0.100
1.000
0.150

Note: The sequence of values are as follows: Sensitivity: Basin Morphology; Infrastructures; and Population. Coping capacity: Employment; Energy; and Poverty.
Resilience: Agricultural production; and Forest Management. Potential Impacts:
Exposure; and Sensitivity. Adaptive Capacity: Resilience; and Coping Capacity.
Vulnerability: Potential Impacts; and Adaptive Capacity.

2. The tendency towards increasing vulnerability to oods in the


years to come is evident in particular for scenario A1B, characterised by higher climate change impacts. Such trends depend
on the combination of a series of factors, and in particular: an
increase of risks deriving from expected demographic and
economic developments in areas exposed to oods; possible
problems in the maintenance of defence structures; insufcient
compensation by the expected positive effects of socioeconomic development (in particular the decrease of the
number of people in extreme poverty), and of the maintenance
of the territory provided by agriculture and forestry.
3. Scenario B1 instead does not show an evident trend of the 10year moving averages, but the uctuations of the projected
vulnerabilities remarkably increase in amplitude in the future,
as a consequence of the greater variability and magnitude of
extreme events projected by climatic models, having amplied
effects due to the negative synergies with increasing sensitivity
estimated for future social and ecological systems.
4. Even with the inclusion of a random noise in the data to
represent the uncertainty associated to our future projections
(up to 20% from present to 2050), the spread of the 100
simulations (solid grey lines in Fig. 8) does not affect the
interpretation of the results.

Fig. 8. Vulnerability trajectories for scenario A1B (left) and B1 (right), reporting on the spread of 100 simulation runs as affected by increasing uncertainty on future projections, and
the 10-year moving averages (dashed lines).

 Given the notion we have of the uncertainty associated to the


future projections:
1. According to the conceptual framework reported above and
modelling results depicted in the graphs of Fig. 8, the future
trend of vulnerability to ood risk in the area shows a tendency
to increase over the 70-year period (dashed lines with 5-year
moving averages), with remarkable uctuations related to
past and projected extreme events (solid lines, spreading as
consequence of the uncertainty associated to input indicators).
The trajectories of the two scenarios obviously coincide in the
past and show divergences when projected on to the future.10

10
The vertical axes of the two graphs have the same scale and thus values
calculated for the vulnerability composite index are comparable. They are left
intentionally blank, because this approach is intended only for explorative and
comparative purposes, and the absolute values calculated for vulnerability are not
of any interest. But the scale.

4. Discussion and conclusions


The assessment of vulnerability to various risks related to
climate change is still an open issue, not yet consolidated in its
conceptualisation and implementation. The notion of vulnerability
has signicant social and subjective dimensions, which have to be
carefully managed and integrated with the needed bases in various
scientic elds. Specic policy document are available, but they
provide only generic conceptual frameworks, which require
signicant efforts to become operational means for supporting the
development of strategies and only rarely scientically sound
methods are consistent with them, and designed for facilitating
communication and integration.
Simplied models with adequate visualization interfaces, made
consistent with the conceptual models adopted by reference policy
documents can signicantly facilitate the communication among
the different communities of actors in the eld of climate change
adaptation (CCA). Simplied models can also contribute to

Please cite this article in press as: Giupponi, C., et al., A dynamic assessment tool for exploring and communicating vulnerability to oods and
climate change, Environmental Modelling & Software (2012), doi:10.1016/j.envsoft.2012.05.004

C. Giupponi et al. / Environmental Modelling & Software xxx (2012) 1e12

overcome the quite common reluctance of decision makers to


uptake the results of extremely complex and opaque integrated
assessment models, while making use of available raw data of
projections for relevant indicators, and can facilitate informed
debates, knowledge transfer and the improvement of common
practices in local planning.
The approach proposed herein can provide an operational
solution in the eld of participatory modelling for facilitating local
dialogues among different actors: scientists of different disciplines,
policy/decision makers, local experts, and stakeholders in general.
It is not a proposal for another vulnerability index, because, as
compared to the usual static additive and fully compensatory
approaches for comparisons amongst countries, it moves the focus
to the temporal dynamics, which are much more relevant for
supporting the debates about CCA and local development. In that
the proposed approach should be intended as a generalized
modelling framework with rather innovative aggregation techniques offering efcient solutions to the formalisation of inherent
subjective judgments and new opportunities for exploiting available information with focus on the most relevant local issues.
The dynamic assessment tool proposed is not at all a mechanistic approach for vulnerability modelling, in that there is no
physically based functional relationships between the various
indicators utilised as proxies for very complex concepts such as the
coping capacity of local society, but simply meaningful aggregation
procedures to express cascades of relationships among relevant
indicators with reference to a specic vulnerability problem (in this
case ood risk). For example, given the relevance of forestry
management for ood risk mitigation expressed by local stakeholders, the fact that better management would contribute to the
resilience of the local SESs is rather evident, and so is the fact that in
turn resilience contributes to reduce vulnerability. Sensitivity
analysis applied to parallel implementations provided by multiple
actors contribute to explore the robustness and reliability of
vulnerability trajectories as affected by exogenous drivers (climate
change in particular) and the issues of local relevance expressed by
means of the proxies provided by the available indicators.
Several methodological issues are to be considered with care:
normalisation is crucial to provide meaningful information for the
aggregation procedures. The independence of indicators projections and the assumptions and scenarios behind their future
projection is also crucial. Model validation is as usual another
challenging task, which could be approached thanks to the
dynamic features of the model allowing us to explore both the past
and the future. Past trends of indicators related to vulnerability,
such as in this case direct and indirect damages deriving from ood
events, could thus be used to compare and to some extent validate
the simulated trajectories. Communication of results should always
be maintained within the scheme of statements proposed in the
previous section.
The application case is a rst approximation implementation
carried out for testing the proposed approach, taking advantage of
the nal results of a recently concluded research project and the
outcomes of the exercise cannot be discussed beyond their
demonstration purposes. Future implementations would require
much more accurate work in the identication of the knowledge
sub-domains and the related indicators. The process of identifying
adequate quantitative indicators, acquiring the required historical
time series and future projections is far from being an easy task. For
instance for the exploratory and methodological purposes of this
work and for the sake of simplication we decided to use only one
indicator per sub-domain, and this can be considered as oversimplifying the complexity of the socio-ecosystem to be analysed.
Nevertheless, the objective of the proposed approach is not to
provide an overall picture of vulnerability, but to analyse the

11

vulnerability to a specic hazard, within a well-known conceptual


model and a well-dened set of drivers, and with the participation
of relevant actors for the development of the model and for
enhanced impact and communication of the results.
Policy measures are not included in the current version of the
model, but ongoing research is considering them in a new prototype,
and this will allow using the tool to support the participatory
processes aimed at exploring possible adaptation strategies. Other
methodological developments are planned in different directions:
rstly, the exploration of potentials for bringing the assessment
model into a spatial context to produce dynamic vulnerability maps;
secondly, the improvement of the procedure and the questionnaire
for the elicitation of weights; thirdly, a distinction between exogenous and internal variables, with consideration of feedbacks
amongst the latter and cumulative effects; nally, the identication
of the main stocks and ows of resources, to explore the possibility
of providing quantications expressed in physical units.
Acknowledgements
The authors gratefully thank the contributions of the partners of
the Brahmatwinn consortium coordinated by Prof. W. Flugel, the
students of the courses on Natural Resources management and
integrated modelling taught by C. Giupponi over the last two year. A
special thank to J. Massheder (Simulistics) and S. Bhandari for their
support for software development. The authors share the responsibility of the paper, and in particular C. Giupponi for what concerns
the design and development of the model and the writing of the
paper, S. Giove for the development of the aggregation procedure,
and V. Giannini for the management of the case study and data
collection.
References
Belton, V., Stewart, T.J., 2002. Multiple Criteria Decision Analysis: an Integrated
Approach. Kluwer Academic Publishers, Boston.
Berkes, F., Folke, C., 1998. Linking Social and Ecological Systems: Management
Practices and Social Mechanisms for Building Resilience. Cambridge University
Press, New York. 476 p.
Ceccato, L., Giannini, V., Giupponi, C., 2011. Participatory assessment of adaptation
strategies to ood risk in the Upper Brahmaputra and Danube river basins.
Environmental Science and Policy 14, 1163e1174.
Coletti, G., Vantaggi, B., 2006. Possibility theory: conditional independence. Fuzzy
Sets and Systems 157, 1491e1513.
Commission of the European Communities, C.E.C, 2009a. Commission Staff Working
Document Accompanying the White Paper Adapting to Climate Change:
Towards a European Framework for Action. Impact Assessment. SEC(2009).
387 p.
Commission of the European Communities, C.E.C, 2009b. White Paper Adapting to
Climate Change: Towards a European Framework for Action. COM(2009). 147
Final.
Dennison, W.C., 2008. Environmental problem solving in coastal ecosystems:
a paradigm shift to sustainability. Estuarine. Coastal and Shelf Science 77 (2),
185e196.
Despic, O., Simonovic, S.P., 2000. Aggregation operators for soft decision making in
water resources. Fuzzy Sets and Systems 115 (1), 11e33.
Dobler, A., Yaoming, M., Sharma, N., Kienberger, S., Ahrens, B., 2011. Regional
climate projections in two alpine river basins: upper Danube and upper
Brahmaputra. Advances in Science and Research 7, 11e20.
Ebert, U., Welsch, H., 2004. Meaningful environmental indices: a social choice
approach. Journal of Environmental Economics and Management 47, 270e283.
EC (European Commission), 2002. Common Strategy on the Implementation of the
Water Framework Directive (2000/60). Guidance on Public Participation in
Relation to the Water Framework Directive. Active Involvement, Consultation
and Public Access Information. Ofce for the Ofcial Publications of the European Communities, Luxembourg.
Field, C.B., Barros, V., Stocker, T.F., Qin, D., Dokken, D.J., Ebi, K.L., Mastrandrea, M.D.,
Mach, K.J., Plattner, G.-K., Allen, S.K., Tignor, M., Midgley, P.M., 2012. Managing
the Risks of Extreme Events and Disasters to Advance Climate Change
Adaptation. A Special Report of Working Groups I and II of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, UK,
and New York, NY, USA, 582 pp.
Fssel, H.-M., 2007. Vulnerability: a generally applicable conceptual framework for
climate change research. Global Environmental Change 17 (2), 155e167.

Please cite this article in press as: Giupponi, C., et al., A dynamic assessment tool for exploring and communicating vulnerability to oods and
climate change, Environmental Modelling & Software (2012), doi:10.1016/j.envsoft.2012.05.004

12

C. Giupponi et al. / Environmental Modelling & Software xxx (2012) 1e12

Fssel, H.-M., 2010. How inequitable is the global distribution of responsibility,


capability, and vulnerability to climate change: a comprehensive
indicator-based assessment. Global Environmental Change 20 (4), 597e611.
Fssel, H.-M., Klein, R., 2006. Climate change vulnerability assessments: an
evolution of conceptual thinking. Climatic Change 75 (3), 301e329.
Giannini, V., Giupponi, C., 2011. Improving Water Governance Through Science and
Stakeholder Dialogue: Experience from Assam (Northeast India), CMCC
Research papers RP-0115, 16 pp.
Giupponi, C., 2007. Decision support systems for implementing the European water
framework directive: the MULINO approach. Environmental Modelling and
Software 22 (2), 248e258.
Giupponi, C., Sgobbi, A., 2007. Models and decision support systems for
participatory decision making in integrated water resource management. In:
Koundouri, P. (Ed.), Coping with Water Deciency. From Research to Policy
Making. Springer, pp. 165e186.
Grabish, M., Marichal, J.L., Mesiar, R., Pap, E., 2009. Aggregation Functions.
Cambridge University Press, Cambridge (UK).
GWP (Global Water Partnership), 2000. Integrated Water Resources Management.
GWP.
Hutton, C.W., Kienberger, S., Amoako Johnson, F., Allan, A., Giannini, V., Allen, R.,
2011. Vulnerability to climate change: people, place and exposure to hazard.
Advances in Science and Research 7, 37e45.
Jeffrey, P., 2003. Smoothing the waters: observations on the process of crossdisciplinary research collaboration. Social Studies of Science 33 (4), 539e562.
Kelkar, U., Narula, K.K., Sharma, V.P., Chandna, U., 2008. Vulnerability and adaptation to climate variability and water stress in Uttarakhand State, India. Global
Environmental Change 18 (4), 564e574.
Meyer, P., Pointhire, G., 2011. Eliciting preferences on multiattribute societies with
a Choquet integral. Computational Economics 37, 133e168.
Mirza, M., 2010. Climate change, ooding in South Asia and implications. Regional
Environmental Change, 1e13.
Nakicenovic, N., Swart, R.E., 2000. Emission Scenarios. Cambridge University Press, UK.
Ostrom, E., 2009. A general framework for analyzing sustainability of
social-ecological systems. Science 325, 419e422.
Otto-Banaszak, I., Matczak, P., Wesseler, J., Wechsung, F., 2010. Different perceptions
of adaptation to climate change: a mental model approach applied to the
evidence from expert interviews. Regional Environmental Change, 1e12.

Pahl-Wostl, C., Holtz, G., Kastens, B., Knieper, C., 2010. Analyzing complex water
governance regimes: the Management and Transition Framework.
Environmental Science and Policy 13 (7), 571e581.
Parker, P., Letcher, R., Jakeman, A., Beck, M.B., Harris, G., Argent, R.M., Hare, M.,
Pahl-Wostl, C., Voinov, A., Janssen, M., Sullivan, P., Scoccimarro, M., Friend, A.,
Sonnenshein, M., Barker, D., Matejicek, L., Odulaja, D., Deadman, P., Lim, K.,
Larocque, G., Tarikhi, P., Fletcher, C., Put, A., Maxwell, T., Charles, A., Breeze, H.,
Nakatani, N., Mudgal, S., Naito, W., Osidele, O., Eriksson, I., Kautsky, U.,
Kautsky, E., Naeslund, B., Kumblad, L., Park, R., Maltagliati, S., Girardin, P.,
Rizzoli, A., Mauriello, D., Hoch, R., Pelletier, D., Reilly, J., Olafsdottir, J., Bin, S.,
2002. Progress in integrated assessment and modelling. Environmental
Modelling and Software 17, 209e217.
Parry, M.L., Canziani, O.F., Palutikof, J.P., Linden, P.J.v.d., Hanson, C.E., 2007. Climate
change 2007: impacts, adaptation and vulnerability. In: Contribution of
Working Group II to the Fourth Assessment Report of the Intergovernmental
Panel on Climate Change. Cambridge University Press, Cambridge, UK, p. 976.
Rammel, C., Stagl, S., Wilng, H., 2007. Managing complex adaptive systems e a coevolutionary perspective on natural resource management. Ecological
Economics 63 (1), 9e21.
Shrestha, A., Aryal, R., 2010. Climate change in Nepal and its impact on Himalayan
glaciers. Regional Environmental Change, 1e13.
Siebenhuner, B., Barth, V., 2005. The role of computer modelling in participatory
integrated assessments. Environmental Impact Assessment Review 25 (4), 367.
Sterman, J., 2000. Business Dynamics: Systems Thinking and Modeling for
a Complex World. McGraw-Hill/Irwin.
Tabara, J.D., Pahl-Wostl, C., 2007. Sustainability learning in natural resource use and
management. Ecology and Society 12 (2), 3.
Teegavarapu, R.S.V., 2010. Modeling climate change uncertainties in water
resources management models. Environmental Modelling and Software 25 (10),
1261e1265.
van Delden, H., Seppelt, R., White, R., Jakeman, A.J., 2011. A methodology for the
design and development of integrated models for policy support.
Environmental Modelling and Software 26 (3), 266e279.
Vogel, C., Moser, S.C., Kasperson, R.E., Dabelko, G.D., 2007. Linking vulnerability,
adaptation, and resilience science to practice: pathways, players, and
partnerships. Global Environmental Change 17 (3e4), 349e364.
Yager, R.R., 1993. Families of OWA operators. Fuzzy Sets and Systems 59, 125e148.

Please cite this article in press as: Giupponi, C., et al., A dynamic assessment tool for exploring and communicating vulnerability to oods and
climate change, Environmental Modelling & Software (2012), doi:10.1016/j.envsoft.2012.05.004

Você também pode gostar