Você está na página 1de 12

TEMPORAL STABILITY OF SOIL ELECTRICAL CONDUCTIVITY

IN IRRIGATED SANDY FIELDS IN COLORADO

H. J. Farahani, G. W. Buchleiter

ABSTRACT. The utility of apparent soil electrical conductivity (ECa ) remains elusive because of its complex interactions with
soil properties. Nonetheless, a significant interest is emerging to utilize the spatial variability in ECa to guide direct soil
sampling and develop varying site-specific management. While the spatial variability of ECa is of significant importance,
understanding its temporal variability is equally important. That is particularly true if delineated ECa zones are to be used
to manage agricultural inputs across the field for multiple years. We present multi-year (1998 to 2002) measurements of
shallow (top 0.3 m of soil) and deep (top 0.9 m of soil) ECa from three irrigated sandy fields in eastern Colorado and quantify
their degree of temporal change. During the study period, the fields were managed uniformly, including all farm operations
and the applied inputs. For each field, soil ECa values were highly correlated between measurement days (for periods of a
few days to four years between measurements) with coefficients ranging between 0.71 and 0.92 for shallow and 0.87 and 0.96
for deep ECa , but significant deviations from the 1:1 line (indicative of temporal variability) were exhibited by shallow ECa .
In these non-saline and sandy fields, delineating spatial patterns of ECa into low, medium, and high zones was highly stable
over time, mainly because they reflect the stable soil properties. Grid-by-grid comparison of the low, medium, and high ECa
patterns from different measurement days showed mean matching percentages of 87, 74, and 78 for shallow and 93, 85, and
89 for deep ECa patterns, respectively, for the three fields. When salt concentration and buildup are low (as was the case
herein), results suggest that single ECa mapping should suffice to delineate stable low, medium, and high ECa zones without
a need for remapping.
Keywords. Management zone, Precision agriculture, Soil electrical conductivity, Soil variability, Temporal stability.

A
mong the many advanced sensors recently Buchleiter, 2002; Sudduth et al., 2002), depth of topsoil
introduced in precision agriculture, bulk or appar- above claypans and depth of depositional sand (Doolittle et
ent soil electrical conductivity (ECa ) measuring al., 1994; Sudduth et al., 1995; Kitchen et al., 1996), soil
devices provide the simplest and least expensive water content (Kachanoski el al., 1988, 1990; Sheets and
soil variability measurement. Mapping the spatial variability Hendrickx, 1995; Hartsock et al., 2000; Schmidhalter et al.,
of soil parameters is key to the successful application of pre- 2001), soil organic carbon (Jaynes et al., 1995), and different
cision agriculture (Robert, 1999). The practical utility of exchange cations (McBride et al., 1990). The fact that
ECa , however, remains elusive because of its complex inter- researchers have found ECa to respond to different soil
actions with soil physical and chemical properties. Nonethe- properties at different sites implies that field characterization
less, a significant interest is emerging to utilize the spatial of ECa variability versus soil properties must be evaluated for
variability in ECa for the purposes of guiding direct soil sam- site-specific areas. There are two main methods to measure
pling (as opposed to systematic grid sampling) and develop- soil ECa : non-contact electromagnetic induction (EMI), and
ing management zones to vary agricultural inputs. the contact methods that induce electrical current into the soil
The strong response of ECa to soil ionic concentration through insulated metal electrodes, which penetrate the soil
(salts and exchangeable ions) has been used extensively for surface. Although many of the above researchers used EMI
in-situ soil salinity surveys (Halvorson and Rhoades, 1974; methods to measure ECa , side-by-side measurements of soil
Rhoades et al., 1990). In the absence of salinity, research ECa by contact electrodes and EMI methods has given
shows that ECa responds well to soil texture and particularly comparable values (Sudduth et al., 1999; Buchleiter and
clay content (Buchleiter, 2000; Kitchen et al., 2000; Schmid- Farahani, 2002) and has yielded similar maps (Doolittle et
halter et al., 2001; Nehmdahl and Greve, 2001; Farahani and al., 2002).
While the spatial variability of ECa and its causative
factors are of significant importance, understanding the
Article was submitted for review in December 2002; approved for temporal variability of both location-specific ECa values and
publication by the Soil & Water Division of ASAE in November 2003. the field-scale ECa patterns is equally important. For ECa
Commercial product names are mentioned for the convenience of the maps to have utility in site-specific management, multiple
reader and do not imply endorsement by USDA-ARS. ECa mappings should yield similar delineations at a relative
The authors are Hamid J. Farahani, ASAE Member, Agricultural
Engineer, and Gerald W. Buchleiter, Agricultural Engineer, USDA-ARS scale, regardless of external factors (Hartsock et al., 2000;
Water Management Unit, Fort Collins, Colorado Corresponding author: King et al., 2001; Nehmdahl and Greve, 2001). That is
Hamid J. Farahani, USDA-ARS Water Management Unit, 2150 Centre particularly important if zones developed based on ECa
Avenue, Building D, Suite 320, Fort Collins, CO 80526; phone: patterns are to be used to vary agricultural inputs across the
970 -492 -7413; e-mail: hamid.farahani@ars.usda.gov.
field for multiple years. Lack of temporal stability in ECa

Transactions of the ASAE


Vol. 47(1): 79-90 2004 American Society of Agricultural Engineers ISSN 0001-2351 79
maps would dictate a need for repeated mappings that precipitation of 356 to 406 mm, falling mostly in the spring
increases cost and complicates management. and summer months.
Literature on the temporal variability of ECa is relatively The landscape is flat to rolling topography, and soils are
sparse and mostly qualitative, particularly for large-scale formed in a mix of stabilized sand dunes, sand mantled South
time frames. Hartsock et al. (2000) reported the magnitudes Platte River terrace, swales, and drainages. The Wiggins1
of ECa vary in time, being more pronounced for the and Wiggins2 fields are 71 and 52 ha, respectively, and are
near -surface soil layer. Eigenberg et al. (2002) documented a few kilometers apart, with soils including a Bijou loamy
significant seasonal change in weekly ECa values in a field sand (coarse loamy, mixed, mesic Mollic Haplargid),
of corn, with the sequential measurements of ECa identifying Valentine sand (sandy, mixed, mesic Typic Ustipsamment),
the dynamic changes in available soil N as affected by animal and Truckton loamy sand (coarse loamy, mixed, superactive,
manure and N fertilizer treatments. Visual comparisons of mesic Aridic Argiustoll). The Yuma field is 57 ha with soils
ECa maps were found to be similar over time (Lund et al., including Haxtun loamy sand, Albinas loam and Ascalon fine
1999), even with dramatic changes in soil water and sandy loam (fine-loamy, mixed, superactive, mesic Pachic
temperature (Dabas et al., 2001; King et al., 2001). Argiustolls).
Pixel -by-pixel correlation of interpolated ECa maps (Nehm- Measurements of ECa were taken between 1998 and 2002
dahl and Greve, 2001) and ECa classification based on equal using the Veris 3100 Soil EC Mapping System (Veris
numbers of observations (Hartsock et al., 2000) yielded Technologies, Salina, Kansas). The Veris unit has six coulter
largely the same delineations at different measurement days, electrodes mounted on an implement that can be pulled by a
with dissimilarities found at a scale smaller than is important pickup truck. It uses a modified Wenner configuration to
for practical applications. Research in soils with low salt measure ECa by inducing current in the soil through two
concentrations, as summarized above, shows that the spatial coulter electrodes and measuring the voltage drop across the
patterns in ECa are more influenced by the stable soil two pairs of coulters that are spaced to measure ECa for the
properties of clay, sand, and organic matter contents, top 0.3 m and 0.9 m soil (Lund et al., 2000). During field
exchange cations, and subsoil structure than the transient measurements, the coulters penetrate the soil 20 to 50 mm
properties of soil water and temperature. The continual (more penetration for drier and looser soil surface condi-
changes in soil transient properties alter the absolute tions). The Veris unit interfaces with a differential GPS and
magnitudes of ECa values. Field spatial patterns of ECa are provides simultaneous and geo-referenced readings of ECa
thus expected to be less prone to changes over time, just like for the shallow (0 to 0.3 m) and deep (0 to 0.9 m) soil depths.
their surrogate soil stable properties. For simplicity, we will use the terms “shallow” and “deep”
The stability of field-scale ECa over long periods of time to refer to these readings.
(i.e., a few years) on fields that have been exposed to different Field -scale ECa mappings were conducted in the spring
cultural and cropping practices has not been examined in prior to tillage and planting operations and/or in the fall after
detail. Our objective was to quantify the degree and nature of harvest. We experienced no difficulty with the Veris unit in
temporal changes in both ECa magnitudes and field patterns bare soil to low surface residue conditions during pre-plant
using multi-year measurements from three sandy and spring measurements (March to May) and in heavy standing
center -pivot irrigated fields in eastern Colorado. Detailed and flat-lying surface residue conditions after fall harvest
discussion of the causes of the observed ECa variability is not (September to November). For ease of maneuvering, fields
presented herein. For the purposes of comparing ECa patterns were traversed in the direction of crop rows in the fall, but
over time, an unsupervised ECa classification method was direction of travel changed at some spring measurements. On
applied to all data sets. In the future, a more practical average, travel speeds through the field ranged between 7 and
approach for developing production-important zones based 16 km h -1 with measurements taken every second, corre-
on ECa data may require a multivariate approach, for sponding to 2 to 4 m spacing between measurements in the
instance similar to Fraisse et al. (2001), that includes soil direction of travel, respectively. A parallel swather (AgGPS
properties and topographic attributes. Parallel Swathing Option, Trimble Navigation Ltd., Sunny-
vale, Cal.) mounted inside the truck guided parallel passes
through the field at 12 to 18 m swath widths with a GPS unit
(AgGPS 132, Trimble Navigation Ltd., Sunnyvale, Cal.)
METHODS AND MATERIALS with sub-meter accuracy, providing spatial coordinates for
This study is part of a larger multi-disciplinary precision each ECa measurement. Changes in soil water content were
farming research study (Heermann et al., 2002), established
monitored during a few of the ECa measurement days (i.e.,
in 1997 on farmer-owned and operated production fields in
spring and fall of 2001 at all fields, except no fall
eastern Colorado. The overall objective of the broader study measurements at Wiggins1, and spring of 2002 at both
is to evaluate the technical and economic feasibility of
Wiggins fields). A total of 20 to 40 sample locations were
precision farming on three center-pivot fields: two near the
identified in each field (randomly selected from delineated
town of Wiggins, Colorado (called Wiggins1 and Wiggins2), ECa zones), and 0.9 m soil profile cores were collected, with
and one near Yuma, Colorado (called Yuma). During the
each 0.3 m long section subsequently analyzed for water
study period (1998 to 2002 at the Wiggins fields and 2000 and
content and other soil properties (not discussed herein).
2001 at Yuma), farm cooperators conducted all field To evaluate ECa variability over time, three levels of
operations and applied all production inputs uniformly across
spatial data representation were developed for each field-
each field. The fields were planted in corn (Zea mays C.)
scale ECa dataset. Examples are presented in figure 1 for the
every year, except Wiggins2, which was planted in onions in 1999 deep ECa data at Wiggins1, including: a non-interpo-
spring of 2001. The climate is semi-arid with average annual
lated 15 × 15 m grid surface (fig. 1a) with each grid assigned
the mean ECa value of all data points falling within that grid,

80 TRANSACTIONS OF THE ASAE


an interpolated 5 × 5 m grid surface (fig. 1b) using the method as natural breaks) or that have limits that are in direct
of inverse distance weighting (IDW), and a smoothed version mathematical relation to each other, such as equal intervals,
of the 5 × 5 m grid surface (fig. 1c) using neighborhood equal number of observations, or standard deviation. For
analysis that reassigns each grid a new value represented by multivariate data sets, clustering techniques have become the
the mean of the grids within a specified neighborhood (in this standard (Lark and Stafford, 1997). Clustering methods are
case, a moving 7 × 7 grid window). The 15 × 15 m grid surface not immune from the difficulties experienced with univariate
(fig. 1a) was used to examine temporal changes in ECa classification, but they are considered more objective.
values. Interpolating spatially dense point data into a Generally speaking, delineating spatial data involves more
continuous surface is a suggested practice in GIS (Burrough, art than science and thus is subjective in nature.
1986). Because of the high density of measured ECa data In this study, data distributions were plotted for the three
points per unit area (a few hundred points per hectare), the sites (fig. 2). The distribution of the shallow and deep ECa
IDW interpolation technique was adequate for creating a data at Wiggins1 was skewed and truncated to the left while
continuous ECa map (fig. 1b). For pattern comparison, we nearly normally distributed at Yuma. Wiggins2 gave mixed
used the smoothed version of the interpolated 5 × 5 m grid distributions for the shallow and deep data, which were
surface (fig. 1c), which provides better-defined and crisp generally skewed and/or bimodal. Among the univariate
delineations with fewer numbers of isolated and small classification methods examined herein, natural breaks
patches. (Jenks, 1977) in ArcGIS (ArcMap, V8.2. ESRI, Redlands,
Field -scale ECa data, either solely or in combination with CA) was least sensitive to the shape of the frequency
other soil and field attributes, have been used to classify distribution, and thus used to produce three ECa delineations
(delineate) field variability (Fridgen et al., 2000; Johnson et for each dataset represented by the smoothed version of the
al., 2001; Fraisse et al., 2001; King et al., 2001). The two 5 × 5 m grid surface. For diagnostic purposes of ECa and soil
main requirements in any classification scheme are identify- variability studies, more delineations may be required, but
ing the number of desired classes and defining the method by three rankings were selected in this study to resemble low,
which classes are assigned. The number of classes is medium, and high ECa . The method of natural breaks
dependent on the objective of the analysis. For instance, delineates the boundaries statistically by finding adjacent
defining low, medium, and high management zones (Fleming feature pairs, between which there is a relatively large
et al., 2000) seems appropriate if the goal is practicality and difference in data values. The method is a robust scheme of
ease of management. As summarized by Burrough (1986) grouping values on the basis of least-variance within class
based on a thorough study by Evans (1977), class intervals for and maximum variance between the classes. To quantify the
univariate data are best defined using class boundaries that temporal variability of the ECa patterns across each field, we
are either selected based on some aspect of the data set (such defined an ECa index given by:

Figure 1. Spatial representations for the 1999 deep ECa data at Wiggins1: (a) a non-interpolated 15 × 15 m grid surface with each grid assigned the
mean ECa value of all data points falling within that grid, (b) an interpolated 5 × 5 m grid surface using IDW, and (c) a smoothed version of the 5 × 5
m grid surface using neighborhood analysis.

Vol. 47(1): 79-90 81


EC a (actual) − EC a (min)  the shallow ECa values but did not appreciably alter the deep
EC a ( Index) =   × 100 (1) readings. The total number of ECa measurement points (n)
 EC a (max) − EC a (min)  per field varied between 8,000 and 20,000 due to changes in
travel speeds and the width of parallel swaths for different
which was used to normalize measured ECa data (ECa(actual))
measurement days. Whole-field mean ECa values across all
to a range between zero and 100. In equation 1, ECa(max) and
years were 14.3, 19.0, and 27.5 mS m -1 (milli-Siemen per
ECa(min) represent maximum and minimum ECa values,
meter) for shallow and 22.9, 22.7, and 33.3 mS m -1 for deep
respectively, for a given field and measurement day.
at Wiggins1, Wiggins2, and Yuma, respectively. The higher
Equation 1 is particularly helpful when the range of values
ECa values at Yuma are mostly due to higher clay and organic
that define a given ECa delineation change over time. It is
matter contents (Farahani and Buchleiter, 2002). Deep ECa
noted that normalization, using the above index or other
values were mostly higher than shallow values (except the
common approaches such as the standard normal variable, is
fall of 2001 at Wiggins2), largely due to higher exchange
simply a one-to-one mapping of the original (raw) ECa data,
cations and, to some extent, water and clay contents at deeper
with no effect on the shape of the ECa patterns across the
depths (data not shown). The ECa data in these irrigated
field.
fields ranged from 2 to 72 mS m -1. Conductivity values of a
few hundred mS m -1 or greater would be required to indicate
salinity, a condition not encountered in these fields. Spatial
RESULTS AND DISCUSSIONS variability, as quantified by coefficient of variation (CV),
TEMPORAL VARIABILITY OF ABSOLUTE VALUES OF ECa was generally high at all sites, ranging from 27% at Yuma to
MEASUREMENTS 50% at Wiggins1. A positive correlation was observed
Summary statistics of ECa data for measurement days between shallow and deep ECa measurements in each field
between 1998 and 2002 are presented in table 1. Due to and for all measurement days. That suggests lack of
equipment malfunction, deep ECa readings for the spring of heterogeneity of horizons, although the positive correlation
2001 were found unreliable (erratic measurements caused by is partly due to deep ECa integrating the 0.9 m soil that
loose wiring) and thus are not reported. The spring 2002 includes the 0.3 m soil layer represented by shallow readings.
shallow readings at Wiggins2 are also not included in the As given in table 1, whole-field mean ECa values
final analysis (but given in table 1) since part of the field was exhibited variations over time, ranging from 12.3 to
disked as ECa mapping was taking place. Disking lowered 17.9 mS m -1 for shallow and 22.3 to 23.2 mS m -1 for deep

Figure 2. Frequency distribution of shallow and deep ECa measurements at Wiggins1 (1999), Wiggins2 (1999), and Yuma (2000).

82 TRANSACTIONS OF THE ASAE


Table 1. Summary statistics of ECa measurements at Wiggins1, Wiggins2, and Yuma center-pivot fields in eastern Colorado.[a]
Measurement Date Shallow ECa (mS m -1) Deep ECa (mS m -1)
Previous
Year Season Crop[b] Mean Std Min Max CV Mean Std Min] Max CV n
Wiggins1
1998 Spring Maize 12.3 4.0 5.6 34.9 33 22.3 9.3 7.6 68.8 42 11290
1999 Spring Maize 13.3 6.0 4.0 44.6 45 23.2 10.9 7.1 70.9 47 9948
2001 Spring Maize 13.5 6.4 3.1 51.8 47 - - - - - 14499
2002 Spring Maize 17.9 9.0 3.0 69.6 50 23.2 9.4 7.2 71.7 41 14989
Mean 14.3 6.4 3.9 50.2 44 22.9 9.9 7.3 70.5 43 12682
Wiggins1
1998 Spring Maize 10.0 2.4 4.0 21.8 24 18.4 6.3 0.9 47.9 34 8934
1999 Spring Maize 16.5 4.8 6.6 37.7 29 21.2 7.6 7.2 49.9 36 8417
2001 Spring Maize 16.4 5.3 4.0 41.9 32 - - - - - 11124
2001 Fall Onions 35.9 8.6 10.3 70.9 24 30.4 9.2 3.4 68.6 30 11874
2002 Spring Onions 16.4 4.0 3.1 32.8 24 20.8 6.1 1.5 49.6 29 19620
Mean 19.0 5.0 5.6 41.0 27 22.7 7.3 3.3 54.0 32 11994
Yuma
2000 Spring Maize 28.9 7.9 9.5 57.7 27 34.5 9.5 12.3 67.7 28 12732
2001 Spring Maize 26.2 7.1 9.1 48.5 27 - - - - - 11798
2001 Fall Maize 27.5 7.3 8.7 50.9 27 32.0 9.0 10.5 60.7 28 17169
Mean 27.5 7.4 9.1 52.4 27 33.3 9.3 11.4 64.2 28 13900
[a] Std = standard deviation, Min = minimum, Max = maximum, CV = coefficient of variation (%), and n = number of observations (grids).
[b] Previous crop = crop grown at the last growing season and prior to ECa .

at Wiggins1, 10.0 to 35.9 mS m -1 for shallow and 18.4 to spring rain (i.e., low in solutes) offers a plausible explanation
30.4 mS m -1 for deep at Wiggins2, and 26.2 to 28.9 mS m -1 for the reduced ECa measurements in spring of 2002.
for shallow and 32.0 to 34.5 mS m -1 for deep at Yuma. The The 15 × 15 m grid data was used to evaluate variability
variations over time were generally small, except that the fall of ECa values across measurement days. A complete
2001 ECa measurements at Wiggins2 were found to be un- summary of grid-by-grid correlation coefficients and regres-
usually much higher (fig. 3) for both shallow and deep ECa sion slopes between all measurement days is given in table 2,
than on any other measurement date. For instance, mean shal- with a few graphical examples presented in figure 4.
low ECa measurements more than doubled (from 16.4 to 35.9 Correlation was generally high, but lower for shallow than for
mS m -1) between the spring and fall of 2001, during which deep (particularly at Wiggins2), ranging between 0.71 and
time a crop of onions was grown. Although the exact reasons 0.92 for shallow and 0.87 and 0.96 for deep ECa . There were,
for such increases are not known, laboratory-measured soil however, significant deviations from the 1:1 line, with slope
saturated paste extract conductivity (ECe ) for soil samples values in table 2 ranging from 0.69 to 2.75 for shallow and
collected on 27 September 2001 revealed increased levels in 0.64 to 1.29 for deep ECa . That is indicative of temporal
ECe (by a factor of two, from 70 mS m -1 in spring of 1999 to variability of absolute values of ECa being particularly
140 mS m -1 in fall of 2001). The elevated soil solution con- pronounced for shallow ECa . That is explained by the fact
centration could have been caused by solutes in the nitrogen that the surface soil layer is subjected to significantly more
fertilizer (UAN) and irrigation water (groundwater from a disturbances within the soil by prevailing climate, agricultur-
depth of 20 m) applications to onions during the 2001 grow- al implements, chemical applications, and microbial activity
ing season. From literature, the effect of N fertilizer and sol- than the subsurface soil.
utes on ECa has been pronounced (Halvorson and Rhoades, Date of measurement changed spatial statistics of field
1974; Eigenberg et al., 2002). For the fall 2001 mapping day, ECa data in Kentucky (Hartsock et al., 2000) when the soil
measured soil water content values were significantly lower water was high. For instance, the portion of the field with
in the fall (0.105 m3 m -3) than in the spring (0.168 m3 m -3) Vertrees soil series (high clay content) had higher ECa values
of 2001, and thus not a causative factor of the elevated ECa
values. Equipment malfunction was also ruled out as a caus-
ative factor for the elevated ECa values in the fall of 2001
since the integrity and accuracy of the Veris unit was verified
(using the procedure outlined in the Veris manual) prior to
each mapping. Furthermore, ECa values similar to the Veris
data were measured on the same day (27 September 2001) us-
ing EM-38 equipment (Geonics Limited, Ontario, Canada)
at 38 locations across Wiggins2 (Buchleiter and Farahani,
2002). Interestingly, the causes of the elevated ECa measure-
ments in the fall of 2001 apparently disappeared by the spring
of 2002, as ECa values decreased significantly (fig. 3).
Leaching of potential solutes (presumably added during the
2001 growing season) below 0.9 m depth by snow melt and Figure 3. Mean shallow and deep ECa values and error bars at Wiggins2
for measurement days between 1998 and 2002 (the 2002 shallow ECa ex-
cludes the field areas where the farmer disked during ECa mapping).

Vol. 47(1): 79-90 83


Table 2. Grid-by-grid correlation coefficients and slopes (in parentheses) of simple linear regressions of ECa values between
measurement days and percent matching of delineated ECa grids between individual delineations of low,
medium, and high ECa rankings between measurement days at the three fields in Colorado.
Shallow ECa Deep ECa
Percent Matching Percent Matching
Correlation Correlation
Coefficient Whole Coefficient Whole
Measurement Date (Slope) Low Medium High Field (Slope) Low Medium High Field
Wiggins1
Spring 98 vs. Spring 99 0.92 (1.30) 95 82 84 88 0.96 (1.07) 99 89 93 94
Spring 98 vs. Spring 01 0.88 (1.29) 94 83 77 87 -- -- -- -- --
Spring 98 vs. Spring 02 0.83 (1.82) 98 73 69 84 0.95 (0.92) 93 94 94 93
Spring 99 vs. Spring 01 0.89 (0.92) 91 86 80 88 -- -- -- -- --
Spring 99 vs. Spring 02 0.82 (1.27) 93 74 70 83 0.96 (0.83) 88 95 90 92
Spring 01 vs. Spring 02 0.82 (1.23) 95 76 78 85 -- -- -- -- --
Mean 94 79 76 86 93 93 92 93
Wiggins2
Spring 98 vs. Spring 99 0.84 (1.76) 97 61 81 74 0.89 (1.15) 93 82 92 89
Spring 98 vs. Spring 01 0.71 (1.48) 94 52 56 61 -- -- -- -- --
Spring 98 vs. Fall 01 0.77 (2.75) 89 65 84 76 0.87 (1.29) 91 82 82 86
Spring 98 vs. Spring 02 -- -- -- -- -- 0.87 (0.87) 93 77 87 86
Spring 99 vs. Spring 01 0.83 (0.82) 93 76 67 79 -- -- -- -- --
Spring 99 vs. Fall 01 0.86 (1.45) 74 76 91 80 0.95 (1.10) 94 92 86 91
Spring 99 vs. Spring 02 -- -- -- -- -- 0.95 (0.74) 96 87 91 92
Spring 01 vs. Fall 01 0.72 (1.24) 63 56 87 66 -- -- -- -- --
Fall 01 vs. Spring 02 -- -- -- -- -- 0.95 (0.64) 97 83 93 91
Mean 85 64 78 73 94 84 89 89
Yuma
Spring 00 vs. Spring 01 0.83 (0.72) 77 75 83 77 -- -- -- -- --
Spring 00 vs. Fall 01 0.80 (0.69) 81 75 80 78 0.89 (0.82) 92 77 87 84
Spring 01 vs. Fall 01 0.80 (0.83) 91 83 81 84 -- -- -- -- --
Mean 83 77 81 80 92 77 87 84

at high water contents than the rest of the field with lower clay ranging from zero to 35 for low, 35 to 50 for medium, and
content. This anisotropic behavior was not apparent in the greater than 50 for the high ECa delineations across all fields.
measurements taken during the drought of 1999 and was It is important to note that the maps shown in figure 6 are
much less pronounced in the deep than in the shallow ECa . interpolated surfaces, and thus the patterns (or the boundaries
Anisotropic (or directional) behavior was not apparent in our of the low, medium, and high ECa areas) are affected by
data, but non-stationarity or varying degree of relative differences in the density of the ECa data per unit area at
change was observed in location-specific ECa values be- different mapping days (due to varying travel speeds and the
tween measurement days (fig. 5). The ratio of location-spe- width of the parallel passes). This contributes to the observed
cific shallow ECa values between two measurement days had dissimilarities in ECa maps across time. To quantify the
a spatial component, as highlighted by the gray-scale maps degree of similarities between ECa patterns, each 5 × 5 m grid
in figure 5, implying that the causes of the mostly increased was assigned the appropriate delineation ranking of low,
ECa values during the second mappings were more pro- medium, and high. Grid-by-grid comparison of maps
nounced in the darker grid areas than in the lighter areas. The between two measurement days was used to quantify the total
darker grid areas correspond to higher clay and organic mat- number of grids that matched (overlaid) and did not match
ter contents. The non-stationarity was much less pronounced (i.e., changed rank and thus did not overlay) each other’s
for deep ECa , which exhibited more uniform shift in whole- delineation ranking. A tabular summary of the pattern
field ECa , if any, between measurement days (see fig. 4 for comparison results is given table 2, showing whole-field
an example shift of deep ECa at Wiggins2 during the fall of matching percentages of 86 and 93 at Wiggins1, 73 and 89 at
2001 measurements). Wiggins2, and 80 and 84 at Yuma for the shallow and deep
ECa patterns, respectively. It is obvious that while both the
TEMPORAL VARIABILITY OF ECa PATTERNS shallow and deep ECa exhibit large-scale temporal stability
Examples of temporal variation in ECa patterns between in patterns, the patterns for the 0.9 m soil profile (or deep
the measurement days in 1998 and 2002 (Wiggins1 and ECa ) remained more stable over time and thus are better
Wiggins2) and 2000 and 2001 (Yuma) are depicted in suited for soil management zone delineation. When salt
figure 6. Visual examination reveals a striking similarity in concentration and buildup is low (as was the case herein), the
the delineated patterns, particularly for deep ECa , over a span results suggest that low, medium, and high ECa zones
of five years. Although there exist dissimilarities between developed based on deep ECa are very similar and indepen-
maps, the overall shapes of the three delineations are dent of time of measurement. Single ECa mapping should
preserved over time. Figure 6 also shows the ranges of the thus suffice to define stable zones without a need for
normalized ECa (or ECa(Index)) values for the delineation remapping.
boundaries of low, medium, and high ECa , with index values

84 TRANSACTIONS OF THE ASAE


Figure 4. Temporal variability of shallow and deep ECa values (each point represents average of all ECa sample points within a 15 × 15 m grid) between
measurement days at the three fields.

Percent matching between individual ECa delineations of 15 m grids, changed appreciably over a two-day period
low, medium, and high is also presented in table 2, suggesting (summarized in table 3), with mean ECa grid values of 38.2
that the low ECa ranking had higher stability over time. For and 36.0 mS m -1 for 25 and 27 September, respectively. The
instance, mean percentages of matching grids among the low measurements between the two days were highly correlated
shallow ECa rankings were 94, 85, and 83, as compared to 79, (coefficient of 0.89) but unexpectedly deviated from the 1:1
64, and 77 for the medium ranking and 76, 78, and 81 for the line (the slope and intercept of the simple regression line were
high ranking in Wiggins1, Wiggins2, and Yuma, respectively. 0.88 and 2.3 mS m -1, respectively), as shown in figure 7. On
The delineated low ECa areas correspond to higher sand both days, the field was mapped by driving over exact passes
contents, suggesting that sandy soils bear a greater stability or tracks that were parallel to the pivot road, which runs
in ECa patterns than clay soils. Due to greater interactions through the center of field from southwest to northeast. There
with soil water and chemistry, soil ECa is expected to be more were no precipitation events between these measurement
dynamic and varying in the presence of clay than sand days, but September 27 was hotter and drier than September
particles. 25, as measured with an on-site weather station. Average
Small -scale temporal variability was examined by ex- daily air temperature was 16.5°C and 21°C, near-surface soil
ploring changes in shallow ECa measured two days apart on temperature was 16.4°C and 17.7°C, and humidity was 22%
25 and 27 September 2001 at Wiggins2. Deep ECa readings and 11% on 25 and 27 September, respectively.
were not available on 25 September for comparison. The The increase in soil temperature does not explain the
magnitudes of shallow ECa , even when averaged for 15 × decrease in ECa between the measurement days since

Vol. 47(1): 79-90 85


Figure 5. Maps of the ratio of location-specific shallow ECa values between two measurement days at each field. The three delineations of the ECa ratio
are based on the method of natural breaks.

literature attributes a 2% increase in ECa per 1°C increase in black marks the non-matching field areas. The majority of
soil temperature (Rhoades et al., 1990). Differences in soil the non-matching grids (black areas) were near the bound-
water are not known since soil water content was only mea- aries of the delineated zones in each field, as was the case for
sured on 27 September. The increased air temperature large-scale variability in ECa patterns. In this study, the
coupled with the decreased humidity on 27 September could highest matching between delineated ECa patterns was
have enhanced evaporation (and thus near-surface soil water expected for the small-scale time frame between 25 and
and solute redistribution), which may explain the 2.2 mS m -1 27 September. However, we only obtained 80% matching
decrease in mean ECa . Small-scale variability in ECa mea- between the ECa patterns, reflecting the difficulties of
surements between two consecutive days was previously comparing interpolated surfaces of temporally varying ECa
documented by Hartsock et al. (2000), who attributed a major values and possibly the realities of a measurement system
portion to measurement errors. Differences in soil water and that disturbed the soil.
temperature between the two days were not discussed. In our The dissimilarities in patterns near their boundaries is
study, it is possible that the coulter electrodes opened up the additionally complicated by the varying upper and lower
soil, allowing drying and increasing resistivity between coul- boundary values of ECa delineations, as dictated by the
ters on the second pass (27 September), which retraced the unsupervised classification method of natural breaks. The
first pass through the field. While it is difficult to quantify the classification method of natural breaks derives such bound-
contribution of the many possible factors to the observed ary limits from the data set itself, and thus is free from
changes in ECa , 75% of the changes in ECa values between subjectivity but biased toward the data set. Since the
25 and 27 September were within ±5 mS m -1. The important boundaries of ECa patterns are influenced by data density,
point is that the level of changes in shallow ECa measured interpolation, smoothing, and classification techniques, and
two days apart is comparable to changes observed in deep thus not unique, it is doubtful that other similar geographical
ECa over a span of few years (i.e., between 1998 and 2002, methods would identify much greater similarities between
as given in table 3). ECa maps. That is supported by past research showing similar
In spite of the varying ECa values between the two dissimilarities among ECa maps using other classification
measurement days, the general patterns of ECa (shown in techniques (Dabas et al., 2001; King et al., 2001; Nehmdahl
figures 8a and 8b as ECa(Index) of the 15 × 15 m grid data) and Greve, 2001; Hartsock et al., 2000). The choice of abrupt
were not significantly distorted. Comparing the delineated ECa pattern boundaries and discrete number of classes
ECa patterns between the two measurement days of 25 and (as commonly used in practice) limits the accuracy of pattern
27 September yielded a whole-field matching of 80% (with delineation and comparison, particularly since true patterns
79%, 71%, and 92% matching for the low, medium, and high of ECa are continuous, and not discrete, surfaces and are
ECa rankings). A spatial view of the matching patterns is defined by subtle transitions from one soil type and condition
presented in figure 8c, where white marks the matching and to the next (as depicted in fig. 8d).

86 TRANSACTIONS OF THE ASAE


Figure 6. Delineated field-scale deep ECa(Index) maps at the three fields for measurement days in 1998 and 2002 at Wiggins fields and 2000 and 2001
at Yuma.
Table 3. Summary statistics (using 15 × 15 m grid values) of temporal
small-scale (shallow ECa on 25 and 27 September 2001) and
large -scale variability (deep ECa on 30 April 1998
and 15 April 2002) at Wiggins2.[a]
Measurement Date Mean Std Min Max CV n
Shallow ECa (mS m -1)
25 Sept. 2001 38.2 8.1 19.8 70.3 21 2073
27 Sept. 2001 36.0 8.1 15.5 64.0 23 2073
Difference[b] 2.2 3.8 -14.2 16.5 173 2073
Deep ECa (mS m -1)
30 April 1998 18.5 5.9 1.2 45.5 32 2073
15 April 2002 20.3 5.9 9.2 43.3 29 2073
Difference[b] -1.8 3.1 -23.8 17.8 172 2073
[a] Std = standard deviation, Min = minimum, Max = maximum, CV = coef-
ficient of variation (%), and n = number of observations (grids).
[b] Statistics for dataset of grid-by -grid differences in EC values between
a
the two measurement days.

Figure 7. Comparison of shallow ECa measured two days apart (25 and
27 September 2001) at Wiggins2 (thick line is the best-fit linear regression
line).

Vol. 47(1): 79-90 87


Figure 8. Maps representing (a and b) 15 × 15 m grid shallow ECa(Index) at Wiggins2 on 25 and 27 September 2001, (c) matching and non-matching
patterns of ECa between the measurement days of 25 and 27 September, and (d) gray-scale representation of shallow ECa on 27 September highlighting
the continuity of ECa patterns.

CONCLUDING REMARKS Results show that in the sandy and non-saline fields
In this study, ECa data were collected from three examined in this study, delineated ECa patterns into low,
center -pivot irrigated, sandy fields in eastern Colorado from medium, and high zones remained largely unchanged
1998 to 2002. During the study period, farm cooperators between 1998 and 2002, even though the absolute values of
conducted all field operations and applied all production ECa varied. The stability was stronger for the 0.9 m soil than
for the shallower 0.3 m topsoil, suggesting care in using soil
inputs uniformly across each field. Apparent soil electrical
management zones based on near-surface (or shallow) ECa
conductivity values and maps were compared over several data. Results suggest that deep ECa is better suited for
days and years to determine if values and patterns were
long-term soil management zone delineation. The higher
temporally stable. Both small- and large-scale temporal
temporal variability of shallow ECa values and patterns is
variability (i.e., for periods of a few days to years between
explained by the surface soil layer (represented by the
measurements) were found in the absolute magnitudes of
shallow ECa ) being subjected to significantly more distur-
ECa . For each field, grid-by-grid comparison of ECa values
bances within the soil by prevailing climate, agricultural
across time yielded high correlations, but correlations were
implements, chemical applications, and microbial activity
lower for shallow ECa than deep ECa , with coefficients
than the subsurface soil. When salt concentration and buildup
ranging between 0.71 and 0.92 for shallow and 0.87 and 0.96
are low (as was the case herein), results show that zones
for deep. The comparison of ECa from one measurement day
developed based on deep ECa are very similar and indepen-
to the next showed deviations from the 1:1 line (indicative of
dent of time of measurement. Single ECa mapping thus
temporal variability), particularly for shallow ECa measure-
should suffice to define stable zones without need for
ments.
remapping.

88 TRANSACTIONS OF THE ASAE


The temporal stability of ECa patterns is attributed to the Agriculture, CD-ROM. P. C. Robert et al., eds. Madison, Wisc.:
underlying stability in the soil factors, such as clay, sand, and ASA-CSSA-SSSA.
organic matter contents and major exchange ions. While the Halvorson, A. D., and J. D. Rhoades. 1974. Assessing soil salinity
transient soil properties of solution concentration, water and identifying potential saline-seep areas with field soil
resistance measurements. SSSA Proc. 38(4): 576-581.
content, and temperature were not examined in detail in this
Hartsock, N. J., T. G. Mueller, G. W. Thomas, R. I. Barnhisel, K. L.
study, others have found equally stable ECa maps over time, Wells, and S. A. Shearer. 2000. Soil electrical conductivity
even with dramatic changes in soil water and temperature variability. In Proc. 5th International Conference on Precision
(Dabas et al., 2001; King et al., 2001). These findings justify Agriculture, CD-ROM. P. C. Robert et al., eds. Madison, Wisc.:
ECa measurements as an economical tool to create potential ASA-CSSA-SSSA.
field areas that may benefit from varying management inputs Heermann, D. F., J. Hoeting, S. E. Thompson, H. R. Duke, D. G.
and practices. In this study, we did not explore in detail the Westfall, G. W. Buchleiter, P. Westra, F. B. Peairs, and K.
causes of the observed temporal variability of ECa , but we Fleming. 2002. Interdisciplinary irrigated precision farming
recognize its importance in using delineated ECa maps to research. Precision Agric. 3(1): 47-61.
develop site-specific management. Results from this study Jaynes, B. D., J. M. Novak, T. B. Moorman, and C. M.
Cambardella. 1995. Estimating herbicide partition coefficients
apply to fields with similar non-saline and sandy soils. Other
from electromagnetic induction measurements. J. Environ.
soils and salinity levels may demonstrate different temporal Quality 24(1): 26-41.
characteristics. Jenks, G. 1977. Optimal data classification for choropleth maps.
Occasional Paper 2. Lawrence, Kansas: University of Kansas,
Department of Geography.
Johnson, C. K., J. W. Doran, H. R. Duke, B. J. Wienhold, K. M.
REFERENCES Eskridge, and J. F. Shanahan. 2001. Field-scale electrical
Buchleiter, G. W. 2000. Advances in soil mapping for improved conductivity mapping for delineating soil condition. SSSA J. 65:
irrigation management. In Central Plains Irrigation Short 1829 -1837.
Course and Exposition Proceeding, 48-57. Colby, Kansas: Kachanoski, R. G., E. G. Gregorich, and I. J. van Wesenbeeck.
Central Plains Irrigation Association. 1988. Estimating spatial variations of soil water content using
Buchleiter, G. W., and H. J. Farahani. 2002. Comparison of noncontacting electromagnetic inductive methods. Canadian J.
electrical conductivity measurements from two different sensing Soil Science 68(4): 715-722.
technologies. ASAE Paper No. 021056. St. Joseph, Mich.: Kachanoski, R. G., E. deJong, and I. J. van Wesenbeeck. 1990.
ASAE. Field-scale patterns of soil water storage from non-contacting
Burrough, P. A. 1986. Principles of Geographical Information measurements of bulk electrical conductivity. Canadian J. Soil
Systems for Land Resources Assessment. New York, N.Y.: Science 70(3): 537-541.
Oxford University Press. King, J. A., P. M. R. Dampney, M. Lark, T. R. Mayr, and R. I.
Dabas, M., J. Tabbagh, and D. Boisgontier. 2001. Multi-depth Bradley. 2001. Sensing soil spatial variability by
continuous electrical profiling (MuCEP) for characterization of electro-magnetic induction (EMI): Its potential in precision
in-field variability. In Third European Conference on Precision agriculture. In Third European Conference on Precision
Agriculture, 1: 361-366. Agro Montpellier, France: Agro Agriculture, 1: 419-424. Montpellier, France: Agro Montpellier.
Montpellier. Kitchen, N. R., K. A. Sudduth, and S. T. Drummond. 1996.
Doolittle, J. A., K. A. Sudduth, N. R. Kitchen, and S. J. Indorante. Mapping of sand deposition from 1993 midwest floods with
1994. Estimating depth to claypans using electromagnetic electromagnetic induction measurements. J. Soil and Water
induction methods. J. Soil and Water Cons. 49(6): 572-575. Cons. 51(4): 336-340.
Doolittle, J. A., S. J. Indorante, D. K. Potter, S. G. Hefner, and W. Kitchen, N. R., K. A. Sudduth, and S. T. Drummond. 2000.
M. McCauley. 2002. Comparing three geophysical tools for Characterizing soil physical and chemical properties influencing
locating sand blows in alluvial soils of southeast Missouri. J. crop yield using soil electrical conductivity. In Second
Soil and Water Cons. 57(3): 175-182 International Geospatial Information in Agriculture and
Eigenberg, R. A., J. W. Doran, J. A. Nienaber, R. B. Ferguson, and Forestry Conference, 2: 122-131. Ann Arbor, Mich.: Altarum.
B. L. Woodbury. 2002. Electrical conductivity monitoring of Lark, R. M., and J. V. Stafford. 1997. Classification as a first step in
soil condition and available N with animal manure and a cover the interpretation of temporal and spatial variation of crop yield.
crop. Agric., Ecosystems, and Environ. 88(2): 183-193. Ann. Applied Biology 130(1): 111-121.
Evans, I. S. 1977. The selection of class intervals. Trans. Inst. Lund, E. D., C. D. Christy, and P. E. Drummond. 1999. Practical
British Geographers (N.S.) 2(1): 98-124. applications of soil electrical conductivity mapping. In Precision
Farahani, H. J., and G. W. Buchleiter. 2002. Practical utility of bulk Agriculture 99: Proc. 2nd European Conference on Precision
soil electrical conductivity. In Proc. 23rd Annual International Agriculture, 771-779. J. V. Stafford, ed. Oxford, U.K.: BIOS
Irrigation Show and Conference: Understanding and Scientific Publishers.
Addressing Conservation and Recycled Water Irrigation, Lund, E. D., C. D. Christy, and P. E. Drummond. 2000. Using yield
CD-ROM. Falls Church, Va.: The Irrigation Association. and soil electrical conductivity (EC) maps to derive crop
Fleming, K. L., D. G. Westfall, D. W. Wiens, and M. C. Brodahl. production performance information. In Proc. 5th International
2000. Evaluating farmer defined management zone maps for Conference on Precision Agriculture, CD-ROM. P. C. Robert et
variable rate fertilizer application. In Precision Agriculture, 2: al., eds. Madison, Wisc.: ASA-CSSA-SSSA.
201 -215. Dordrecht, The Netherlands: Kluwer Academic McBride, R. A., A. M. Gordon, and S. C. Shrive. 1990. Estimating
Publishers. forest soil quality from terrain measurements of apparent
Fraisse, C. W., K. A. Sudduth, and N. R. Kitchen. 2001. Delineation electrical conductivity. SSSA J. 54: 290-293.
of site-specific management zones by unsupervised Nehmdahl, H., and M. H. Greve. 2001. Using soil electrical
classification of topographic attributes and soil electrical conductivity measurements for delineating management zones
conductivity. Trans. ASAE. 44(1): 155-166. on highly variable soils in Denmark. In Third European
Fridgen, J. J., N. R. Kitchen, and K. A. Sudduth. 2000. Variability Conference on Precision Agriculture, 1:461-464. Montpellier,
of soil and landscape attributes within sub-field management France: Agro Montpellier.
zones. In Proc. 5th International Conference on Precision

Vol. 47(1): 79-90 89


Rhoades, J. D., P. J. Shouse, W. J. Alves, N. A. Manteghi, and S. M. Sudduth, K. A., N. R. Kitchen, D. F. Hughes, and S. T. Drummond.
Lesch. 1990. Determining soil salinity from electrical 1995. Electromagnetic induction sensing as an indicator of
conductivity using different models and methods. SSSA J. 54: productivity on claypan soils. In Proc. Site-Specific
46-54. Management for Agricultural Systems, 671-681. P. C. Robert,
Robert, P. C. 1999. Status and research needs. In Precision R. H. Rust, and W. E. Larson, eds. Madison, Wisc.:
Agriculture 99: Proc. 2nd European Conference on Precision ASA-CSSA-SSSA.
Agriculture, 12-15. J. V. Stafford, ed. Oxford, U.K.: BIOS Sudduth, K. A., N. R. Kitchen, and S. T. Drummond. 1999. Soil
Scientific Publishers. conductivity sensing on claypan soils: Comparison of
Schmidhalter, U., A. Zintel, and E. Neudecker. 2001. Calibration of electromagnetic induction and direct methods. In Proc. 4th
electromagnetic induction measurements to survey the spatial International Conference on Precision Agriculture, 979-990. P.
variability of soils. In Third European Conference on Precision C. Robert, R. H. Rust, and W. E. Larson, eds. Madison, Wisc.:
Agriculture, 1: 479-484. Montpellier, France: Agro Montpellier. ASA-CSSA-SSSA.
Sheets, K. R., and J. M. H. Hendrickx. 1995. Noninvasive soil Sudduth K. A., N. R. Kitchen, W. D. Batchelor, G. A. Bollero, D.
water measurement using electromagnetic induction. Water G. Bullock, D. E. Clay, H. L. Palm, F. J. Pierce, R. T. Schuler, K.
Resources Research 31(10): 2401-2409. Thelen, and W. J. Wiebold. 2002. Characterizing field-scale soil
variability across the midwest with electrical conductivity. In
Proc. 6th International Conference on Precision Agriculture and
Other Precision Resource Management, 54-69. Madison, Wisc.:
ASA-CSSA-SSSA.

90 TRANSACTIONS OF THE ASAE

Você também pode gostar