Environmental factors of spatial distribution of soil salinity on flat irrigated terrain

Abstract

Inefficient irrigation and the excessive use of water on agricultural land in the Aral Sea Basin over several decades have led to saline soils. The main objective of this paper is to identify the environmental predictors to model the spatial distribution of soil salinity in a highly irrigated landscape. Soil salinity at farm scale was measured in the topsoil (Total Dissolved Solids, TDS) and down to a depth of 1.5 m by electromagnetic conductivity meter (CMv) over a regular grid covering an area of approximately 15 km2 in Khorezm Province, Uzbekistan. Six nested samplings within selected grids were conducted to reveal short-distance variation. Apart from widely-used terrain indices and those acquired from remote sensing, data on distance to drainage channels and long-term average groundwater observations were used to account for local parameters possibly influencing soil salinity. Topsoil salinity (TDS) was seen to be highly variable even at short distances (40 m) compared to average bul

Full text

Environmental factors of spatial distribution of soil salinity on flat irrigated
terrain
A. Akramkhanov1*, C. Martius1, 2, S. J. Park3, J. M. H. Hendrickx4
1 Center for Development Research (ZEF), University of Bonn, Walter-Flex-Str. 3, 53113 Bonn,
Germany
2 Current address: Inter-American Institute for Global Change Research, Avenida dos Astronautas
1758, 12227-010 São José dos Campos, SP, Brazil
3 Department of Geography, Seoul National University, Shilim-Dong, Kwanak-Gu, Seoul, Korea
4 New Mexico Tech, LeRoy Place 801, Socorro NM 87801, USA
ABSTRACT
Inefficient irrigation and the excessive use of water on agricultural land in the Aral
Sea Basin over several decades have led to saline soils. The main objective of this paper
is to identify the environmental predictors to model the spatial distribution of soil
salinity in a highly irrigated landscape. Soil salinity at farm scale was measured in the
topsoil (Total Dissolved Solids, TDS) and down to a depth of 1.5 m by electromagnetic
conductivity meter (CMv) over a regular grid covering an area of approximately 15 km2
in Khorezm Province, Uzbekistan. Six nested samplings within selected grids were
conducted to reveal short-distance variation. Apart from widely-used terrain indices and
those acquired from remote sensing, data on distance to drainage channels and long-
term average groundwater observations were used to account for local parameters
possibly influencing soil salinity. Topsoil salinity (TDS) was seen to be highly variable
even at short distances (40 m) compared to average bulk soil salinity (CMv). CMv
readings were better correlated with factors obtained from remote sensing and distance
to drains than TDS. This might be attributable to the fact that topsoil salts are dynamic
1* Corresponding author. Fax: +998 62 2243347. E-mail address: akmal@zef.uzpak.uz
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
1

in nature, and land management practices (e.g. leaching, cultivation, irrigation) might
have contributed considerably to spatial variation. The CMv shows the average amount
of salt within a larger soil volume and to greater depth and is less affected by land
management than topsoil salinity, which is reflected in the TDS. Most terrain indices
showed a low correlation with topsoil and bulk salinity. There was a strong indication
that the effects of water management are dominant and tend to outweigh the effects of
environmental factors. The very low R2 for relationship of TDS with environmental
factors is evidence that taking TDS samples close to the soil surface is not a good way
to assess salinity trends in irrigated land. These findings have important implications for
salinity survey methods on flat irrigated terrain: CMv seems to be a more reliable
predictor than environmental proxy factors, even if the latter are easier to determine.
Keywords: soil degradation, Aral Sea Basin, Khorezm, spatial interpolation,
variability
INTRODUCTION
In Central Asia, agriculture is a major source of income for a large part of the
population, and in the arid climate, irrigation is an essential factor. Water supply
problems have been exacerbated by recent geopolitical developments in the Central
Asian region (cf. for Uzbekistan: Djanibekov et al., 2011); conflicts over water and
energy are emerging between up-, mid- and downstream countries, and Afghanistan
may want to claim a share of the water of the Amu Darya River (Glantz, 2002; Martius
et al., 2009). Thus, future water shortages are to be anticipated, and droughts such as the
one the region experienced in 1999-2001 may become more frequent. Salinization
affects an estimated 75% of the irrigated land in the Aral Sea Basin (van Dijk et al.,
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50

1999). One of the adversely affected areas in the basin is the Khorezm region of
Uzbekistan, which is the model area in an ongoing project aimed at providing integrated
solutions for land and water management (Martius et al., 2011). Salt leaching is a
common practice in the region.
The intricate processes of soil salinization in irrigated arid regions are well
understood but difficult to measure, although the latter is becoming easier with the
advent of quick appraisal methods. In the developing world, soil salinity surveys at the
landscape level still remain the major source of information on salinity distribution,
despite their many limitations: conventional soil maps do not delineate all of a field’s
inherent variability, nor do they show specific soil attribute variations (Moore et al.,
1993; Burrough, 1993). They also inadequately represent the dynamics of soil salinity.
The recent development of quantitative methods based on geostatistics and
incorporation of environmental variables partly stems from the practical constraints of
conventional soil survey methods, which can be criticized as being too qualitative and
too focused on soil management and land-use planning (McBratney et al., 2003; Scull et
al., 2003; Grunwald, 2006).
For the purpose of characterizing the spatial variability of soil salinity, a conceptual
framework which links soil characteristics to certain landscape features can be applied.
The theoretical details of this concept can be found in the reviews by Scull et al. (2003)
and McBratney et al. (2003). Similar concepts with additional environmental variables
were used for salinity studies (Odeh et al., 1998). The use of interpolation techniques to
estimate values at unsampled locations offers more opportunities to combine factors
which were previously hard to incorporate. Evans and Caccetta (2000) used remote
sensing data and landform data obtained by Digital Elevation Modeling (DEM) to
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74

predict areas at risk from dryland salinity. Searle and Baillie (1997) developed a salinity
hazard map based on topographic indices, soils, geology, climate and vegetation.
There are numerous studies on soil salinity (Roberts et al., 1997; McKenzie and
Austin, 1993); however, most of them were conducted on landscapes that had not been
extensively altered by irrigation projects, and on large catchments. For the present study
an intensive investigation of various aspects of soil salinity in an irrigated terrain setting
was undertaken. The main objective of the study was to find environmental predictors
to model the spatial distribution of soil salinity in a highly irrigated landscape. The
following specific objectives were addressed: (i) to characterize soil salinity at the study
site; (ii) to characterize the spatial distribution of salinity; (iii) to identify the best
predictors for soil salinity distribution. The study provides baseline data for monitoring
soil salinity over the area and environmental parameters to improve salinity estimates.
MATERIALS AND METHODS
Site description and sampling design
The survey was conducted on two farms: the research farm (41°35’N, 60°31’E) of the
Urgench State University, and the Pahlavon Mahmud private shareholder farm (shirkat)
(41°37’N, 60°31’E), south-west of the city of Khiva (Figure 1).
The topography of the land is flat, with elevation points normally distributed and
ranging from 88 to 97 m above sea level (mean 92 m). Due to the flat topography the
groundwater flow is limited, resulting in a very shallow groundwater table depth which
contributes to soil salinity (Ibrakhimov et al., 2007). The salinity type is mainly
chloride-sulphate, which is typical for soils in this region (Akramkhanov, 2005). The
landscape is dissected by an extensive network of drains and collectors. The soils in the
district are classified by the local classification system as desert zone meadow-oasis
soils and fit the description of Fluvisols and gleyic calcaric Arenosols.
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99

The sampling campaign was carried out from June to August 2002. Core sampling
and electrical conductivity (EC) measurements were done systematically in a
150×200 m regular grid. They were complemented by nested grid sampling with a finer
40×40 m grid in order to reveal short-distance variation. Nested fields were randomly
selected within the grid area totaling 6 nests (cf. Figure 1). This design provided 580
grid nodes, of which 452 were sampled; the rest included housing settlements and were
therefore omitted.
Field survey and environmental factors
The locations of the grid nodes (in the World Geodetic System 1984) were laid out
prior to the sampling campaign and uploaded to a Global Positioning System (GPS)
receiver (GPS 12, Garmin International Inc., USA). At each grid node, soil core
samples were taken at a depth of 0-30 cm using a split tube sampler with an inner
diameter of 53 mm. Samples were analyzed for total dissolved solids (TDS) and texture
(pipette method).
The apparent electrical conductivity of the bulk soil was measured in situ using an
electromagnetic conductivity meter (CM-138, GF Instruments, Czech Republic). The
device allows two measuring modes: the vertical mode (CMv), which provides
estimates to a depth of 1.5 m, and the horizontal mode (CMh), which works to a depth
of 0.75 m. Since CMv and CMh data are closely correlated (r=0.84), only the CMv
results are presented here.
On the basis of existing literature on similar environments and our own local
knowledge, we selected and analyzed several environmental variables from the many
factors influencing soil salinity (Table 1). The initial approach for this study was the
assumption that local terrain serves as a simplified surrogate integrating the numerous
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123

landscape processes that influence the total amount of soil salinity. For the arid and
semi-arid areas, Salama et al. (1999) related the spatial distribution of saline land and
water to its hydro-geomorphology (e.g., topography and hydro-stratigraphy). Florinsky
et al. (2000) integrated the concept of accumulation, transition and dissipation zones
together with a digital terrain model to map small-scale salinity risk from the large scale
maps. Topographic indices are not only important, they are also easy to calculate with
modern computers, thus facilitating their use in soil salinity prediction. Additionally,
other local environmental factors such as the water network, soil texture, land cover,
groundwater table depth and salinity are thought to further improve estimates of soil
salinity distribution.
Land cover (cotton, alfalfa, bare (fallow), maize, melon, wheat, no-crop (sandy) and
other (i.e. abandoned or marginal)) was recorded during the survey. Proxy data on
groundwater table depth (GWT) and its salinity (GWS) were obtained from the
observation wells installed in the area by the hydrogeological-melioration expedition
(HGME) of the Khorezm Department of Land and Water Resources and the ZEF
project (ZEF, 2003). Groundwater table depth and salinity data collected in the month
of July from 1990 to 2002 at 45 groundwater observation wells covering a larger area
(ca. 90 km2) were averaged and interpolated with ordinary kriging. It was assumed that
long-term average data represent the spatial distribution of groundwater table depth and
salinity better than single-year observations. In July and August 2002, measurements
were obtained for eight observation wells in conservation agriculture trial fields (O.
Egamberdiev, personal communication, 2002) and six from irrigation efficiency study
fields (I. Forkutsa, personal communication, 2002). It must be stressed that the
correlations between salinity and GWS and GWT are for interpolated GWS and GWT.
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147

Elevation data from a 1:10,000 topographic map were obtained to create a Digital
Elevation Model (DEM) of the study area. Because a DEM is of primary importance to
derive terrain indices, data were obtained from elevation points and contour lines
covering an area of approximately 27 km2. The accuracy of interpolation was assessed
by using cross validation error statistics including root mean square error (RMSE).
Terrain indices were calculated from a 30×30 m raster-based DEM. The grid size of
30 m was chosen because it has proven to be the most suitable for soil-landscape
analyses (Park et al., 2009). The following terrain indices were calculated using the
software DIGEM 2.0 (Olaf Conrad, Göttingen, Germany): aspect (AS), slope (SL),
profile curvature (PROFC), plan curvature (PLANC) (Zevenbergen and Thorne, 1987),
divergence/convergence (DC) indices, solar radiation (Solar), flow accumulation
(upslope contributing area, UA), wetness (WT), and erosivity (based on universal soil
loss equation, LS) (Moore et al., 1993). Additionally, curvature (CURV7) and the
terrain characterization indices (TCI) were calculated according to Park et al. (2001).
A 1:10,000-scale agricultural map was used to obtain information on the water
network infrastructure, which consisted of irrigation and drainage channels within the
sampling area. The layers were digitized and the shortest distance from the sampling
points to drains was calculated using ArcView 3.2 (ESRI Inc., USA). The influence of
the lakes that can be seen in Figure 1 is difficult to assess; during sampling some of
them were in fact dry. For the present analysis, the lakes were not considered as a
factor.
Remote sensing parameters were obtained from a Landsat 7 satellite image acquired
on July 12, 2002. We correlated salinity data with the normalized difference vegetation
indices (NDVI) as well as the transformed normalized difference vegetation indices
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171

(TNDVI). Additional indices calculated from these bands are: soil-adjusted vegetation
index (SAVI) and ratio vegetation index (RVI), known to delineate reduced reflectance
due to salinity (Wang et al., 2002), and soil band ratio (RS57), which is sensitive to clay
minerals and is the combination most likely to indicate saline soils (Goossens et al.,
1996). Altogether, remote sensing provided 12 variables (including raw band signals;
Bands 1-5, 7, and 8) that were included in the analysis.
Data analysis and visualization
The geostatistical analyses were performed in ArcMap 8.3 (ESRI Inc., USA) using
the Geostatistical Analyst 8.3 extension. Spatial continuity was modeled by
semivariograms. A spherical model was fitted to the experimental variograms. Ordinary
kriging was applied as an interpolation method, as it minimizes the influence of outliers
on prediction performance and has been widely used (Odeh et al., 1994; Triantafilis et
al., 2001). Positively skewed data were log-transformed according to Saito and
Goovaerts (2000), who found that log-normal kriging consistently yields the best results
compared to other kriging methods.
Correlation analyses were performed in SPSS 11.0 (SPSS Inc., USA). To identify the
main factors to determine soil salinity at the study sites, stepwise multiple regression
was conducted with 150×200 m grid nodes. To investigate the effect of soil texture,
crop type, and field or management unit on soil salinity we used ANOVA analyses for
comparing means among clustered locations.
RESULTS AND ANALYSES
Summary statistics of soil salinity
The average total dissolved solids (TDS) of 3200 ± 3500 ppm (Table 2) indicate that
the topsoil is of low salinity (cf. also the classification of Kaurichev (1989), where 4000
and 7000 ppm correspond to moderately and highly saline soils), but peaks of >30000
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196

ppm are found. Overall, about 43% of all soil samples range between slightly and
moderately saline. The coefficient of variation for TDS (109) is high.
The average CMv value 0.50 dS m-1 indicates that the salt concentration over the
profile is low to moderate (Bennett et al., 1995). Overall, about 53% of sampling
locations showed slight to moderate soil salinity. The coefficient of variation for CMv
(66) is small compared to TDS; this is very likely due to the larger volume of soil
measured by the device, whereas TDS was measured in a smaller soil sample taken
from the top 30 cm layer only.
Electromagnetic conductivity readings are influenced by many factors such as soil
salinity, texture, temperature and moisture content. In the local conditions soil salinity
has a dominant effect on the CM-138 readings and thus the latter can be used as a proxy
for soil salinity. This effect is explained by the fact that the clay content in soils in the
study area rarely exceeds 20%. The difference between clay content of sandy soils and
silt loams contributes little to CM-138 readings. The temperature effect is also
minimized by conducting the survey only during the summer period when average soil
temperatures of the 1 meter profile are around 25° Celsius (Hendrickx et al., 2002).
Although most of the study area was irrigated, soil moisture content in some sampling
locations was below 20%, the effect of which on CM-138 readings is unclear (Bennett
et al., 1995). This factor could not be quantified, but previous studies indicate that it
might have had a minor effect on the analyses (Hendrickx et al., 1992; Rhoades et al.,
1990).
The coefficients of variation for ancillary variables such as clay, groundwater table
depth and salinity show a moderate to high magnitude of variability (Table 2), with clay
content above the ranges typical for published soil studies as summarized by Mulla and
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220

McBratney (2000). Although a relatively high clay content is typical for the region,
which belongs to the so-called meadow group of soils according to the Russian
classification described in detail in Tursunov and Abdullaev (1987), the prevailing low
clay content in the study area is explained by the interspersed sandy desert soils. Soil
texture is roughly characterized by silt loam (34%), sand (30%), and sandy loam (28%).
Groundwater table depth and salinity were clearly not normally distributed. Locally
large values were found. The distribution of clay content was more close to normal.
Spatial distribution of soil salinity
Soil salinity as measured by TDS and CMv were autocorrelated to a distance of
448 m and 571 m respectively (Figure 2). On a small scale however, the variograms
show that TDS had a considerable nugget effect, which suggests that. The grid sampling
design with nested grids (vs. coarse selective sampling) implemented in this study
provided a good way of exploring the spatial structure of soil salinity distribution. The
large nugget variance in the TDS variogram could be caused by the small sample
support, the soil cores having a diameter of 53 mm only.
The interpolated maps for CMv and TDS are given in Figure 3. Salinity measured by
CMv shows two distinct areas of high and low readings. The study area consists of old
deposits from the ancient Daudan river (which later became the Amu Darya), with clay
and loamy texture (Tursunov and Abdullaev, 1987) bordering Karakum desert sands to
the south. Areas with low readings coincide with interspersed sands and higher readings
with relatively heavier soils. TDS (Figure 3) did not show these trends so clearly, the
highest salinities being measured in the centre of the southern part (Figure 4).
The spatial trend analyses revealed similar patterns for clay and groundwater table
depth; they gradually decreased from the north-east to the south-west. The same
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244

directional trend was displayed by groundwater salinity, the only difference being that it
gradually increased along the north-east to south-west gradient.
Similar spatial trends for measured soil properties and the elevation indicate that one
factor is most likely responsible for this effect, i.e. the former course of the Amu Darya
river, which formed the existing relief (Tsvetsinskaya et al. 2002) and influenced the
soil texture.
Correlation between soil salinity and environmental factors
Table 3 shows the correlation coefficients between soil salinity and environmental
factors. In most cases, the correlations were low (<0.5). Of the remote sensing factors
only Band 7 and of the terrain indices only UA are included in the table.
TDS had low and non-significant correlation with most terrain indices. This might be
attributable to the fact that topsoil salts are dynamic in nature, and land management
practices (e.g., leaching, cultivation, irrigation) might have contributed considerably to
spatial distribution. On the other hand, terrain indices had a low but significant
influence on bulk soil salinity CMv. Moore et al. (1993) emphasized that surface soil
properties are most modified by land management and that as a result features of lower
horizons in the profile may display greater response to topographic attributes.
Remotely sensed data correlated significantly with both salinity parameters (Table 3),
due to the contribution of vegetation cover to reflectance values and the subsequent
vegetation indices calculated from band combinations.
Distance to drains (DCOLL) correlated slightly but significantly with TDS. Since the
existing drainage network was built to lower the groundwater table, its functionality
should be indirectly reflected by changes in soil salinity. Close to the drains, soil
salinity should be lower and increase with distance from the drain. However, the spatial
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268

distribution of salinity indicates that drainage networks have little influence on topsoil
salinity, which can be attributed to high variability of salinity at small scales, irrigation
events, or the irregular layout of the drainage network.
However, drain proximity was significantly related to CM-138 readings, and was
similar to the effect of terrain attributes. The CM-138 shows the average amount of salt
within a larger soil volume and to greater depth, and is less affected by land
management than topsoil salinity which is reflected in the TDS.
As anticipated, interpolated groundwater table depth (GWT) and salinity (GWS) had
high correlation coefficients with soil salinity CMv, but the direction of the influence is
somewhat contradictory. The positive correlation of groundwater table depth with soil
salinity suggests that salinity was higher when the groundwater table was deeper. This
result is counter intuitive and will be explained below.
Tables 4, 5 and 6 show that the salinity estimates (CMv, TDS, and GWS) have the
strongest correlation with the field or management unit (Table 4). The upper case letters
in the three tables indicate which means are significantly different. For example, in
Table 4 there is no significant difference between the mean CMv of field F3 and F4, but
these fields are significantly different from the other fields. There is a rather weak
correlation with crop type (Table 5) and a fairly good correlation with soil texture
(Table 6). This is a strong indication that the effects of water management are dominant
and tend to outweigh the effects of environmental factors. This finding has important
implications on how to conduct salinity surveys in flat irrigated terrain: the field or
management unit seems to be a more important predictor than environmental proxy
factors - even if the latter are easier to determine (Hendrickx et al., 1992).
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291

Salt accumulation in the soil depends to a large extent on the capillary flux that
carries water and salts from the groundwater table towards the soil surface. Capillary
rise is a critical factor for soil salinity that not only depends on groundwater table depth
but also on the hydraulic properties of the soil profile (Hendrickx et al., 2003). Table 6
shows that the groundwater table depths of the coarser textures (sand and loamy sand)
are significantly shallower than those of the finer textures (silt loam and loam). Yet, the
soil salinity of the sand is significantly lower than that of the finer textured soils. This
can be explained by the fact that capillary rise in finer textured soil is greater than in a
coarse textured soil. For example, the data in Table 6 show that the soil salinities (CMv)
of loam and sand are, respectively, 0.69 and 0.50 dS m-1 while their groundwater table
depths are, respectively, 131 and 114 cm. Table 5.5 of Hendrickx et al. (2003) implies
that in a loam a capillary flux of 0.2 cm/day can occur from groundwater table depths
65 to 130 cm but only 0.04 cm/day from 31 to 115 cm in the sand. This large difference
of capillary fluxes explains the unexpected negative correlation between groundwater
table depth and soil salinity in this study.
In the local conditions soil salinity has a dominant effect on the CM-138 readings and
these can therefore be used as a proxy for soil salinity. This is because the clay content
in soils in the study area rarely exceeds 20%. The environmental attributes considered to
be the possible controlling factors were selected on the assumption that they are
representative of the area and could be easily extracted for the rest of the Khorezm
region. The only variable that was discarded was ‘distance from main collector’ (the
collector which carries drainage water out of the region), which was initially considered
in the analyses. This had the highest correlation coefficient with CMv (0.75), which was
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314

unexpected and not representative of other areas of the Amu Darya delta, and led to the
decision to drop the variable from further analyses.
The negligible effect of terrain attributes on soil salinity expressed by low correlation
coefficients can be considered an artifact of the traditional correlation tools. It is a well-
known principle that landform has a significant effect on salinity distribution, and this
was confirmed by our own observations in the study area. Indeed, the ordinary
correlation coefficient cannot account for the presence of both significant negative and
positive correlations existing between two variables at different frequencies (Nielsen et
al., 1983).
Best predictors of soil salinity distribution
The data were analyzed by stepwise multiple regression, using a set of remote
sensing parameters and terrain indices, groundwater table depth and salinity, distance to
collectors, soil texture, and land cover as regressors. The regression with CMv (Table 7)
as a dependent variable had rather good fit (R2adj = 48%), and the overall relationship
was significant (F=41.5, p<0.01). The regression with TDS (Table 8) as dependent
variable was a poor fit (R2adj = 21%), but the overall relationship was significant
(F=12.5, p<0.01). Cotton, GWT and no-crop (sandy) land cover were best predictors for
both salinity estimates, TDS and CMv. The loamy sand and the rest of the variables did
not have significant influence on salinity estimates.
DISCUSSION
The characterization of the spatial distribution of soil salinity by geostatistical
analyses shows that topsoil salinity is highly variable, whereas bulk soil salinity to a
depth of 1.5 m is less so. Therefore, in studies similar to this one, when interpreting
analysis results more emphasis should be given to the CM-138 measurements, because
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338

they provide the average of the soil profile salt content and seem to be less affected by
disturbances or other human activities.
As soil texture largely influences salt content, it could be expected that terrain exerts
a similarly strong influence on salinity. The soil texture was determined at each
sampling site and was used for statistical analyses; however, such information might be
not available at other locations. Nevertheless, there is a large amount of information
available on soils, especially where irrigation projects have been established, and this
information could be used to improve soil salinity prediction. The lack of variance in
soil salinity explained by topography using the correlation technique suggests that there
are some other constraints at work that need to be taken into account when considering
the results from this analysis. As mentioned before, terrain indices are mostly well
pronounced for the topography, following catenary development. The area of the
present study is mainly flat, and delineating it into landscape units according to a catena
was not possible. However, some catenary distribution of the topsoil salinity was
discernable when isolated sections of the study area were considered individually.
Figure 4 shows that the lowest points on sandy soils had high TDS values. In contrast,
the same sandy soils on a slope (although slopes here are gentle, they have a marked
effect) had low TDS values, while heavier soils on the lower slopes had higher TDS
values. Nonetheless, confirmation of this micro-catenary effect would require a greater
number of samples and study slopes.
Since the research site represents an area of recent land development where various
transformations are still under way (e.g., the cropping of areas that were previously
unused, the filling in of depressions and converting them into fields), the soil properties
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361

are not in the steady state statistical approaches assume, making it difficult to relate soil
salinity to landform geometry.
Similarly, estimates of groundwater table depth and salinity are affected by the
interpolation technique. This probably increased the error of the groundwater data, the
reliability of which on this scale was doubtful from the start. The authors are aware of
the errors peculiar to the monitoring procedure of observation wells. Also, the range of
change of groundwater table depth and salinity is not very high and thus does not allow
inferences to be made. Perhaps groundwater observation data taken for July only was
not adequate to quantify the processes that influence salinity. Therefore, no conclusive
statements can be made on the direction of the correlation coefficients of soil salinity
and groundwater table depth and salinity.
Larger study areas generally show poorer environmental correlation due to the
additional heterogeneity of the environmental factors (Park et al., 2009). The use of a
large number of variables obtained or calculated from remote sensing or topography
would improve the ability of the model to predict soil salinity. However, the danger of
multicollinearity exists, and a minor change in one variable could then have a
considerable influence on model output.Nevertheless, our study shows that (historic or
current) water management overrides the eventually existing correlations between
environmental parameters and salinity. Hence, the best advice that can be given at
present is that direct estimates of salinity are preferable to indirect estimates based on
environmental proxies. Having said this, a second important lesson is that the use of
bulk estimation methods, such as the electromagnetic conductivity meter here used in
vertical mode (CMv), is to be preferred to direct sampling and laboratory analysis of
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384

topsoil salinity, which is not only costly and time-consuming but also provides values
that are too localized and therefore of limited use for practical management purposes.
CONCLUSIONS
Spatial distribution of soil salinity and influencing factors in the landscape of the
Khorezm region were investigated by linking environmental variables to soil salinity.
Topsoil salinity was seen to be highly variable even over short distances (40 m)
compared to average bulk soil salinity measured by CM-138.
Soil salinity was poorly correlated with terrain attributes. Likely reasons for this poor
correlation between terrain attributes and soil salinity are flat topography, land
management practices that were difficult to incorporate into this study, or the failure of
the correlation tool used to study soil salinity relationship with terrain attributes.
Factors obtained from remote sensing (listed in Table 1) had low but significant
correlation coefficients with both salinity of topsoil and measured by the CM-138. Since
band signals and calculated indices are mainly an indication of vegetation, the
correlation suggests that salinity affects crop growth significantly, band signals and
indices can be used as a remote sensing indicator.
Distance to drains is an important factor, especially for the bulk soil salinity of the
profile. Correlation was lower for topsoil, which might be due to higher spatial variation
of the topsoil salinity.
The very low R2 for relationship of TDS with environmental factors is evidence that
it is not possible to predict TDS close to the soil surface from environmental factors in
irrigated lands. This study provides evidence that the use of electromagnetic
conductivity meters that can be used to provide maps of soil salinity is currently the best
option for the management of salinity in flat irrigated terrain.
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408

ACKNOWLEDGEMENTS
This study was funded by the German Ministry for Education and Research (BMBF;
Project Number 0339970A). We would like to thank the anonymous reviewers of an
earlier version of this paper for their valuable comments and suggestions that improved
the paper.
REFERENCES
Akramkhanov, A., 2005. Spatial distribution of soil salinity: detection and prediction.
Ph.D. Thesis, Bonn University, Bonn, Germany.
Bennett, D.L., George, R.J. and Ryder, A., 1995. Soil salinity assessment using the
EM38: field operating instructions and data interpretation. Miscellaneous Publication
4/95, Department of Agriculture, Western Australia.
Burrough, P.A., 1993. Soil variability: a late 20th century view. Soils and Fertilizers,
56: 529-562.
Djanibekov, N., Bobojonov, I. and Lamers, J. P.A., 2011. Farm reform in Uzbekistan.
In: C. Martius, I. Rudenko, J.P.A. Lamers and P.L.G. Vlek (Editors), Cotton, Water,
Salts and Soums – Economic and Ecological Restructuring in Khorezm, Uzbekistan.
Springer, Berlin, Heidelberg, New York. (in press)
Evans, F.H. and Caccetta, P.A., 2000. Broad-scale spatial prediction of areas at risk
from dryland salinity. Cartography, 29: 33-40.
Florinsky, I.V., Eilers, R.G. and Lelyk, G.W., 2000. Prediction of soil salinity risk by
digital terrain modelling in the Canadian prairies. Canadian Journal of Soil Science, 80:
455-463.
Glantz, M., 2002. Water, climate, and development issues in the Amudarya Basin.
Report of informal planning meeting held 18-19 June 2002 in Philadelphia,
Pennsylvania. Environmental and Societal Impacts Group, NCAR, Boulder, CO.
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433

Goossens, R., De Vliegher, B.M., Ghabour, T.K. and Gad, A., 1996. Monitoring,
simulation and prediction of soil salinity and waterlogging risk using remote sensing
and GIS on provincial level (Ismailia, Egypt). In: A. Spiteri (Editor), Remote
Sensing'96. Integrated applications for risk assessment and disaster prevention for the
Mediterranean. 16th EARSeL Symposium, Malta.
Grunwald, S., 2006. Environmental Soil-Landscape Modeling – Geographic
Information Technologies and Pedometrics. CRC Press, New York.
Hendrickx, J.M.H., Phillips F.M., and Harrison J.B.J., 2003. Water flow processes in
arid and semi-arid vadose zones. In: I. Simmers (Editor), Understanding Water in a Dry
Environment. Hydrological Processes in Arid and Semi-arid Zones, Vol. 23. A.A.
Balkema Publishers, Lisse, The Netherlands, pp. 151-210.
Hendrickx, J.M.H., Wraith, J., Kachanoski, R.J., and Corwin, D.L., 2002. Solute
content and concentration, p. 1253-1322, In: J. Dane and C. Topp (Editors), Methods of
Soil Analysis. Part 1. Soil Science Society of America Madison, Wisconsin.
Hendrickx, J.M.H., Baerends, B., Raza, Z.I., Sadiq, M. and Chaudhry, M.A., 1992.
Soil salinity assessment by electromagnetic induction on irrigated land. Soil Sci. Soc.
Am. J., 56: 1933-1941.
Ibrakhimov, M., Khamzina, A., Forkutsa, I., Paluasheva, G., Lamers, J.P.A.,
Tischbein, B., Vlek, P.L.G., and Martius, C., 2007. Groundwater table and salinity:
Spatial and temporal distribution and influence on soil salinization in Khorezm region
(Uzbekistan, Aral Sea Basin). Irrigation and Drainage Systems, 21: 219-236.
Kaurichev, I.S., 1989. Pochvovedenie. Agropromizdat, Moscow (in Russian).
Martius, C., Froebrich, J., Nuppenau, E.-A., 2009. Water resource management for
improving environmental security and rural livelihoods in the irrigated Amu Darya
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457

Lowlands. In: H. G. Brauch, Ú. O. Spring, J. Grin, C. Mesjasz, P. Kameri-Mbote, N. C.
Behera, B. Chourou, H. Krummenacher (Editors), Facing Global Environmental
Change: Environmental, Human, Energy, Food, Health and Water Security Concepts.
Hexagon Series on Human and Environmental Security and Peace, Vol. 4, Springer,
Berlin, Heidelberg, New York, pp. 749-762.
Martius, C., Rudenko, I., Lamers, J.P.A., Vlek, P.L.G., 2011. Cotton, Water, Salts
and Soums – Economic and Ecological Restructuring in Khorezm, Uzbekistan.
Springer, Berlin, Heidelberg, New York.
McBratney, A.B., Mendonça Santos M.L. and Minasny B., 2003. On digital soil
mapping. Geoderma, 117: 3-52.
McKenzie, N., and Austin, M., 1993. A quantitative Australian approach to medium
and small scale surveys based on soil stratigraphy and environmental correlation.
Geoderma, 57: 329-355.
Moore, I.D., Gessler, P.E., Nielsen, G.A. and Peterson, G.A., 1993. Soil attribute
prediction using terrain analysis. Soil Science Society of America Journal, 57: 443-452.
Mulla, D.J., and McBratney, A.B., 2000. Soil spatial variability. In: M. E. Sumner
(Editor), Handbook of Soil Science. CRC Press LLC, Boca Raton.
Nielsen, D.R., Tillotson, P.M. and Vieira, S.R., 1983. Analyzing field-measured soil-
water properties. Agricultural Water Management, 6: 93-109.
Odeh, I.O.A., McBratney, A.B. and Chittleborough, D.J., 1994. Spatial prediction of
soil properties from landform attributes derived from a digital elevation model.
Geoderma, 63: 197-214.
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479

Odeh, I.O.A., Todd, A.J., Triantafilis, J. and McBratney, A.B., 1998. Status and
trends of soil salinity at different scales: the case for the irrigated cotton growing region
of eastern Australia. Nutrient Cycling in Agroecosystems, 50: 99-107.
Park, S.J., McSweeney, K. and Lowery, B., 2001. Identification of the spatial
distribution of soils using a process-based terrain characterization. Geoderma, 103: 249-
272.
Park, S.J., Ruecker, G.R., Agyare, W.A., Akramkhanov, A., Kim, D. and Vlek,
P.L.G., 2009. Influence of grid cell size and flow routing algorithm on soil-landform
modeling. Journal of the Korean Geographical Society, 44: 122-145.
Rhoades, J.D., Shouse, P.J., Alves, W.J., Manteghi, N.A., and Lesch, S.M., 1990.
Determining soil salinity from soil electrical conductivity using different models and
estimates. Soil Sci. Soc. Am. J., 54:46-54.
Roberts, D.W., Dowling, T.I. and Walker, J. 1997. FLAG: A fuzzy landscape
analysis GIS method for dryland salinity assessment Technical Report No 8/97. CSIRO
Land and Water, Canberra, Australia.
Saito, H., and Goovaerts, P., 2000. Geostatistical interpolation of positiviely skewed
and censored data in dioxin-contaminated site. Environ. Sci. Technol., 34: 4228-4235.
Salama, R.B., Otto, C.J. and Fitzpatrick, R.W., 1999. Contributions of groundwater
conditions to soil and water salinization. Hydrogeology Journal, 7: 46-64.
Scull, P., Franklin, J., Chadwick, O.A. and McArthur, D., 2003. Predictive soil
mapping: a review. Progress in Physical Geography, 27: 171-197.
Searle, R. and Baillie, J. 1997. Prediction of landscape salinity hazard using a
geographic information system. Department of Natural Resources, The State of
Queensland.
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503

Triantafilis, J., Odeh, I.O.A. and McBratney, A.B., 2001. Five geostatistical models
to predict soil salinity from electromagnetic induction data across irrigated cotton. Soil
Science Society of America Journal, 65: 869-878.
Tsvetsinskaya, E.A., Vainberg, B.I., Glushko, E.V., 2002. An integrated assessment
of landscape evolution, long-term climate variability, and land use in the Amudarya
Prisarykamysh delta. Journal of Arid Environments, 51: 363–381.
Tursunov, L.T. and Abdullaev, S., 1987. Soil-physical characteristics of the lower
Amudarya. 117 pp. FAN, Tashkent (in Russian).
van Dijk, A., den Haan, R., van den Hoven, P. and Bos, R., 1999. Can WARMAP
save the Aral Sea? Land and Water International, 93 (2).
Wang, D., Wilson, C. and Shannon, M.C., 2002. Interpretation of salinity and
irrigation effects on soybean canopy reflectance in visible and near-infrared spectrum
domain. International Journal of Remote Sensing, 23: 811-824.
ZEF, 2003. Economic and ecological restructuring of land- and water use in the
region Khorezm (Uzbekistan). A Pilot Project in Development Research. Project Phase
II: Field Research and Development of a Restructuring Concept (2004-2006). ZEF
Bonn.
Zevenbergen, L.W. and Thorne, C.R., 1987. Quantitative analysis of land surface
topography. Earth Surface Processes and Landforms, 12: 47-56.
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522

Table 1. Sources, types and methods of analysis of environmental parameters for soil
salinity in the Khorezm region
Sources Description Types of available
information
Analysis
Field survey Samples, laboratory
analysis data, field records
and CM-138 readings
Soil salinity (TDS, CMv)
Soil texture
Land cover
Analysis of variance
(ANOVA)
Hydrogeological-
melioration expedition
(HGME)
Network of groundwater
observation wells
GW table depth
GW salinity
GW well location
Correlation
Map, 1:10,000 scale Agriculture, topographic,
geological profile maps
Drainage canals
Elevation points, DEM,
Terrain indices
Correlation
Landsat 7 satellite
sensors
Detect earth scene radiation
in 3 bands: visible and near
(VNIR), short wavelength
infrared (SWIR)
Bands 1-5, 7, 8
Indexes calculated from
band ratios: NDVI,
TNDVI, SAVI, RVI, RS57
Correlation
523
524

Table 2. Summary statistics of variable parameters that were used for interpolation
Units Mean Std. Dev. CV Min Max N
Elevation m 92 1.3 1 88 97 4122
Clay % 8 5 60 0 22 448
Total dissolved solids* g 100g-1 0.32 0.35 109 0.06 3.46 448
CMv in vertical mode dS m-1 0.50 0.33 66 0.01 1.92 445
Groundwater table depth cm 121 31 25 81 262 57
Groundwater salinity dS m-1 2.9 1.3 45 1.3 6.7 59
* 0.32 g 100g-1 of dry soil = 3200 ppm ≈ 3200 ppm x 640-1 = 5 dS m-1
525
526
527
528
529
530
531
532
533

Table 3. Pearson’s correlation matrix between soil salinity and environmental variables:
Upslope contributing Area (UA), Distance to drainage (DCOLL), Groundwater Table
depth (GWT) and Groundwater Salinity (GWS). TDS = total dissolved solids, CMv =
bulk soil salinity measured with CM-138 in vertical mode (1.5 m deep)
Band 7cUAcDCOLL GWT GWSc
TDScg 100g-1 -0.30a-0.04 0.10b0.03 -0.12a
CMv dS m-1 -0.30a-0.15 0.46a0.37a-0.37a
a correlation is significant at the 0.01 level (2-tailed)
b correlation is significant at the 0.05 level (2-tailed)
c log10 transformed
534
535
536
537
538
539
540
541
542
543

Table 4. Comparisons of means of CMv, TDS, GWS, and GWT among the six fields.
CMv = bulk soil salinity measured with CM-138 in vertical mode (1.5 m deep), TDS =
total dissolved solids, GWS = groundwater salinity, GWT = groundwater table depth.
The upper case letters indicate which means are significantly different.
Field CMv (dS m-1):
R2=0.62; Pr<0.0001
TDS (g 100g-1):
R2=0.18; Pr<0.0001
GWS (dS m-1):
R2=0.97; Pr<0.0001
GWT (cm):
R2=0.97; Pr<0.0001
ID N AVG STD AVG STD AVG STD AVG STD
F3 29 A 0.81 0.17 A 0.30 0.16 D 2.84 0.07 B 134 0.8
F4 30 A 0.83 0.09 A 0.32 0.16 E 2.02 0.05 A 148 3.3
F5 29 D 0.46 0.10 C 0.13 0.11 B 3.62 0.27 D 130 2.0
F6 30 B 0.62 0.12 B 0.21 0.08 D 2.86 0.08 C 132 0.6
IL 32 C 0.54 0.09 A B 0.28 0.16 A 4.65 0.07 E 120 6.1
IS 31 C 0.55 0.05 A 0.33 0.22 C 3.27 0.14 F 95 1.9
544
545
546
547

Table 5. Comparisons of means of CMv, TDS, GWS, and GWT among the four crops.
CMv = bulk soil salinity measured with CM-138 in vertical mode (1.5 m deep), TDS =
total dissolved solids, GWS = groundwater salinity, GWT = groundwater table depth.
The upper case letters indicate which means are significantly different.
Crop CMv (dS m-1):
R2=0.30; Pr<0.0001
TDS (g 100g-1):
R2=0.09; Pr=0.001
GWS (dS m-1):
R2=0.04; Pr=NS
GWT (cm): R2=0.03;
Pr=NS
ID N AVG STD AVG STD AVG STD AVG STD
alfalfa 11 A 0.94 0.18 A B 0.20 0.13 B 2.90 0.05 A 135 0.3
cotton 150 B 0.63 0.15 A 0.28 0.17 A B 3.22 0.89 A 125 18.0
maize 7 B 0.58 0.18 A B 0.21 0.20 B 2.92 0.25 A 132 0.7
melon 13 C 0.42 0.08 B 0.10 0.03 A 3.73 0.18 A 129 2.8
548
549
550
551

Table 6. Comparisons of means of CMv, TDS, GWS, and GWT among the five soil
textures. CMv = bulk soil salinity measured with CM-138 in vertical mode (1.5 m
deep), TDS = total dissolved solids, GWS = groundwater salinity, GWT = groundwater
table depth. The upper case letters indicate which means are significantly different.
Texture CMv (dS m-1):
R2=0.48; Pr<0.0001
TDS (g 100g-1):
R2=0.19; Pr<0.0001
GWS (dS m-1):
R2=0.30; Pr<0.0001
GWT (cm):
R2=0.48; Pr<0.0001
ID N AVG STD AVG STD AVG STD AVG STD
silt loam 45 A 0.82 0.15 B C 0.31 0.17 B 2.45 0.53 A 141 8.7
loam 3 B 0.69 0.21 A B 0.42 0.18 A 3.12 1.32 A B 131 17.9
sandy loam 73 B 0.62 0.14 C 0.26 0.13 A 3.53 0.97 B 129 8.8
loamy sand 8 B C 0.57 0.11 A 0.47 0.23 A 3.28 0.19 D 100 12.8
Sand 52 C 0.50 0.08 C 0.20 0.17 A 3.46 0.28 C 114 17.8
552
553
554
555

Table 7. Results of stepwise multiple regression analyses with CMv as dependent
variable. Distance to drainage (DCOLL), Groundwater Table depth (GWT),
Groundwater Salinity (GWS), Profile Curvature (PROFC), Slope (SL).
Beta SE b Standardised beta
Constant 130.02 28.63
Silt loam (texture) 0.21 0.03 0.29**
DCOLL 0.0004 0.0001 0.17**
GWT 0.002 0.001 0.09*
GWSc-0.68 0.14 -0.22**
Solar -18.15 4.01 -0.18**
PROFC -1112.98 431.09 -0.09**
SL_rad -23.03 8.66 -0.09**
Cotton (cover) 0.12 0.03 0.18**
Alfalfa (cover) 0.17 0.06 0.11**
No-crop (cover) -0.20 0.05 -0.15**
** correlation is significant at the 0.01 level (2-tailed)
* correlation is significant at the 0.05 level (2-tailed)
c log10 transformed
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574

Table 8. Results of stepwise multiple regression analyses with TDS as dependent
variable. Groundwater Table depth (GWT), Erosivity (LS).
Beta SE b Standardised beta
Constant 1.95 0.78
Band 5c-1.53 0.34 -0.24**
Sand (texture) -0.57 0.10 -0.36**
Sandy loam (texture) -0.24 0.08 -0.15**
GWT -0.01 0.00 -0.11*
LSc-0.12 0.05 -0.11*
Wheat (cover) 0.49 0.14 0.17**
Cotton (cover) 0.19 0.08 0.12*
No-crop (cover) 0.52 0.15 0.17**
Bare (cover) 0.50 0.19 0.12**
Other (cover) 0.90 0.20 0.21**
** correlation is significant at the 0.01 level (2-tailed)
* correlation is significant at the 0.05 level (2-tailed)
c log10 transformed
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592

Figure Captions
Figure 1 Farm scale sampling layout with irrigation and drain network, lakes, main
collector and settlements overlaid on Landsat image (July 12, 2002)
Figure 2 Semivariograms of CM-138 vertical mode reading measured at the soil
surface and TDS
Figure 3 Interpolated maps of CM-138 in vertical position (CMv) and total dissolved
solids (TDS) overlaid over an elevation of the study area. Note that the elevation is
exaggerated compared to the other two dimensions of the graph. Area represents 3×4
km.
Figure 4 Extruded values of total dissolved solids (blue columns; g 100g-1)
differentiated by clay content (darker areas = more clay; %) both overlaid on elevation.
Extreme values of TDS (extruded columns) coincide with local depressions within
lighter areas (sandy). Note that elevation is exaggerated in comparison to the other two
dimensions of the graph. Area represents 3×4 km.
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611