Vous êtes sur la page 1sur 14

Agricultural Water Management 77 (2005) 96109


Assessment of hydrosaline land degradation

by using a simple approach of
remote sensing indicators
Nasir M. Khan a,*, Victor V. Rastoskuev b,
Y. Sato a, S. Shiozawa a

Graduate School of Agricultural and Life Sciences, The University of Tokyo,

Bunkyo-Ku, Tokyo 113-8657, Japan
Research Center for Ecological Safety, Korpusnaya Str. 18, 197110 St. Petersburg, Russia
Accepted 1 September 2004
Available online 10 May 2005

This research deals with monitoring irrigated saline soils of Faisalabad, Pakistan. The analysis
is based on remote sensing data acquired from the Indian Remote Sensing satellite (IRS-1B) and
using a Geographical Information System (GIS). We have examined how different remote sensing
indicators work for salinity prone lands classification and assessment in part of the Indus basin of
Pakistan, which is facing extremely hydrosalinized land degradation problems. The study has
suggested some new but simple and practical approaches for assessing salinity. We have analyzed
the effectiveness of several indicators for the presence of salts in the area in terms of salinity indices
(SI), especially several combinations of the ratio of the signals received in the third spectral band to
others. As salt-affected soils are also characterized by stressed vegetation, vegetation indices were
also analyzed as concurrent indicators. The probability for obtaining a correct classification of the
satellite images has shown to be strongly dependent on the season for all indicators analyzed. The
best results can be achieved for the dry season (MarchApril), but not in humid or high temperature
periods. The most difficult part in the classification processes was to distinguish between saltaffected areas and rural/village populated areas due to its muddy roofs producing similar reflection
as of patchy saline and dry barren distributed soils. We have come-up with two original schemes of
classification through the analysis of available satellite data for this specific test area. In the first
case, we tried to produce a new set of composite and stretched images out of four channels data
* Corresponding author. Tel.: +92 51 4436432; fax: +92 51 4436431.
E-mail address: kmnasir97@yahoo.com (N.M. Khan).
0378-3774/$ see front matter # 2005 Elsevier B.V. All rights reserved.

N.M. Khan et al. / Agricultural Water Management 77 (2005) 96109


using special digital image processing (DIP) techniques and then analyzing their ratios. In the
second scheme, we analyzed isoclustering functions that perform classification based on specifically created images (through principal component analysis (PCA) and salinity indices) instead of
the common practice of using just satellite sensors reflectance measurements. Both schemes have
shown the ability to perform good classification and assessment for hydrosaline degraded lands in
the study area using IRS-1B data.
# 2005 Elsevier B.V. All rights reserved.
Keywords: Waterlogging and salinity; Remote sensing; GIS; Salinity indices; Indus basin

1. Introduction
Soil salinization is becoming an increasing problem, especially in arid and semi-arid
regions wherever irrigation is practiced. In some parts of the world, like Pakistan, the
population is growing very fast, and therefore, attempts are made to increase the
agricultural production, in many cases by land reclamation, but facing limited water
resources. It is reported recently that about 10% of presently arable lands of the world are
affected by salinity (Tabet et al., 1997). Salinity and sodicity affect an estimated 952 Mha
of land (Szabolcs, 1992). This is the case in Pakistan, where seepage and percolation from
large irrigation systems gave rise to watertable, and thus, gradually and substantially
contributed to the outburst of a twin menace, i.e., waterlogging and salinization. The
magnitude of the salinity/sodicity problem can be gauged from the fact that at one stage in
the country, the area of productive land was being damaged by salinity at a rate of about
40,000 ha/year (WAPDA, 1981). Salt-affected lands in the Indus basin of Pakistan cover an
area of 4.22 Mha, which is 26% of the total irrigated area (Ghassemi et al., 1995). Thus,
monitoring saline degraded lands has always been a primary issue for efficient irrigation
systems management and rehabilitation policies.
The problem of detection, monitoring and mapping salt-affected soils is known to be a
difficult matter because dynamic processes are involved. Recent advances in the
application of remote sensing technology in mapping and monitoring degraded lands,
especially in salt-affected soils, have shown great promise for enhanced speed, accuracy
and cost effectiveness. The approach to the problem of delineating saline soils using remote
sensing data and Geographical Information System (GIS) techniques has been proved
efficient in many recent studies (Sharma and Bhargawa, 1988; Rao et al., 1991; Dwivedi,
1992; Srivastava et al., 1997; Dwivedi and Sreenivas, 1998; Khan and Sato, 2001). The
combination of remote sensing with GIS is very promising, especially for the monitoring of
soil salinization (Goossens et al., 1993; Casas, 1995). This study is devoted for mapping
salt-affected soils in Pakistan. An approach using remote sensing indicators, different
salinity indices (SI), vegetation indices and principal component analysis (PCA), which are
based on spectral characteristics of different kind of surfaces, has been applied here as
digital image processing (DIP) and GIS techniques. The work has been accomplished with
the use of the capacity of GIS IDRISI for Windows (Eastman, 1995) through its special
modules and mathematical operator functions for remote sensing data processing and


N.M. Khan et al. / Agricultural Water Management 77 (2005) 96109

Fig. 1. Location of the study area.

2. Area description and data used

The area under study is located in the center of Punjab province of Pakistan, latitudes
318020 to 318450 N and longitudes 728500 to 738220 E (Fig. 1), which is part of the worlds
largest contiguous Indus basin irrigation system. It is a part of subtropical continental low
land region and is designated as the semi-arid central Punjab. The climate conditions have
marked variations in temperature (mean monthly maximum temperature range from 19.4
to 41.2 8C) and rainfall over the year (<350 mm on average) occurs mostly (75%) during
the monsoon season (JulySeptember). The mean annual potential evaporation is as high as
2100 mm (WAPDA/SMO, 19931994). Physiographically, the area is nearly flat to very
gentle slope (the average topo-gradient is 0.02%) from north-east to south-west direction
with mean elevation of 190 m. The soils are alluvial deposits classified as silt loam, loam
and silt clay loam and loamy sands. The cropping calendar has two seasons, called rabi
(winter) and kharif (summer). The main rabi crops are wheat, sugarcane, pulses, and
fodder, while corn, paddy, cotton, sugarcane and fodder occupy lands in kharif season. The
average yields of wheat, sugarcane, maize, cotton and rice in the whole study area are 1.5,
26.1, 0.85, 0.3 and 1.37 tonnes ha1, respectively. Except for rice, yields of all other crops
are much lower than the average yields in Faisalabad administrative division.
Data were recorded by the sensor Linear Image Self-scanning Spectrometer (LISS-II) of
Indian Remote Sensing satellite (IRS-1B) (launched on August 29, 1991 and declared in
operation from September 16, 1991) having 22 day repeating orbit (857  919 km) at
905 km mean altitude and 99.258 sun-synchronous/inclination working in four spectral
bands: blue (B1: 0.0420.52 mm), green (B2: 0.520.59 mm), red (B3: 0.620.68 mm) and
near infrared range (NIR) (B4: 0.770.86 mm). It exhibits a narrow field-of-view (FOV)
with 74 km  2 swath-width. A spatial resolution of 36  36 m has been used in the
analysis. This data series extended the work potential for groundwater exploration, surface
water and land use problems, like salinity, land/water management, flood monitoring,
forest, etc. The parametric details of IRS LISS-II instrument are shown in Table 1.

N.M. Khan et al. / Agricultural Water Management 77 (2005) 96109


Table 1
Characteristics of IRS-IB satellite, LISS-II Sensor data used in the study
Serial no.


LISS-II sensor

Spectral range (mm)


Number of bands

Spectral bands (mm)



Ground resolution (m)


Swath (km)
B2, B3, B4

74  2

Radiometric resolution (gray levels)


Data rate (Mbps)

10.4  2

Different time periods (pre-monsoon and post-monsoon) have been chosen for DIP
analysis because of considerable variation in soil surface salinity as well as in vegetation
cover (Table 2). Topographic maps of Pakistan at 1:50,000 scale were used as a main
supporting data to register all images. Many other data reports/maps showing extremely
salt-affected areas (manually surveyed) obtained from the International Waterlogging and
Salinity Research Institute (IWASRI) and Scarp Monitoring Organization (SMO) of Water
and Power Development Authority (WAPDA) were also used for identification of results.

3. Remote sensing indicators investigation

3.1. Spectral response patterns
The IRS LISS-II digital data were registered to the topo-sheets for the area using about
15 control points that were easily recognizable on the satellite images. It was presumed that
the topographic maps are the most reliable source of information. After this registration
procedure, satellite data were ready to be used in GIS. The Mercator projection (UTM) was
chosen for data presentation in GIS IDRISI for Windows, which was used throughout for
interpretation and classification procedures.
Table 2
Archived dates of the satellite images used for the analysis of the study area
Image numbers

Archived dates (dd/mm/yr)




N.M. Khan et al. / Agricultural Water Management 77 (2005) 96109

Spectral response patterns of various terrain features are fundamental to a derivation of

information on natural resources and environmental degradation using space-borne
multispectral measurements. For different types of surfaces, the amount of reflected solar
radiation varies with the wavelength, which makes it possible to identify various kinds of
surfaces or classes in a satellite image and distinguish them from each other by the
differences in reflectance. To generate spectral response patterns of the various selected
major classes, LISS-II data were displayed and spectrally homogenous areas representing
various categories were then identified. In this study, five major classes were separated:
vegetation (crops), populated areas (towns and settlements), salt-affected areas (saline
soils), water bodies like lakes/ponds and irrigation-drainage channels (canal/drain) for
relative comparison (Fig. 2). Spectral responses in different spectral bands in the form of
digital number (DN) values were generated for different classes. A similar exercise was
carried out for all other LISS-II images to evaluate the results.
The ground truth data in terms of survey topo-maps were used to help in picking out
these training sites for the crop, town, lakes, airport and canal classes (Fig. 2). The training

Fig. 2. Red band image of LISS-II sensor showing the training sites for the different land classes: (1) cropped area,
vegetation, (2) town area, urban area, (3) airport, (4) salt prone lands, (5) water body, lake and (6) water channel.

N.M. Khan et al. / Agricultural Water Management 77 (2005) 96109


Fig. 3. Spectral response pattern for various land cover features using LISS-II data.

sites for salt-affected areas, which have a relatively high reflectance coefficient in the first
three spectral bands of satellite data, were picked out from satellite images taking into
account all surface collected information. Examples of spectral response patterns for all
these classes at selected points are presented in Fig. 3. These results showed that the
spectral response pattern of salt-affected lands is higher than the other classes in all bands
and in all images whereas vegetation reflects maximum in NIR range, i.e., band 4 (0.77
0.86 mm). The salt-affected soils have been found to reflect more incident light energy in
visible spectrum (0.450.68 mm) than those of other land cover features. This response for
saline soils is extremely useful as it helps the segregation of salt-affected soils from normal
soils and vegetation. Metternichit and Zinck (1997) also found that salt-affected lands had
high spectral reflectance in the visible window, particularly in the blue and red range of the
spectrum at low moisture content. There is also possibility that the spectral data would have
been affected due to certain external effects such as soil moisture and atmospheric factors.
In terms of DN values, the spectral pattern variation for the training class features of saltaffected, waterlogged and cropped area were noted for all images and for all bands and then
normalized to band 4 (NIR) to have visual comparison of variation with time and each band
for that particular class. These results showed that there is a noticeable variation with time


N.M. Khan et al. / Agricultural Water Management 77 (2005) 96109

in each band response for that class which could be attributed to a change in land cover,
canopy of the vegetation and atmospheric parameters like temperature, etc. Therefore, the
selection of the image for a certain time and for detection of a certain class has a main role
in any classification procedure.
3.2. Remote sensing indices
Indices are usually designed by combining two or more spectral bands and have been
widely acknowledged as powerful tools in identifying features of interest. Many studies
describe a wide range of band combination/indices developed for vegetation vigor, crop
assessment and land use change. However, there is paucity of information about the band
combinations meant for waterlogged and salt-affected lands. In the present study, the
proposed band combinations are hoped to discriminate salt-affected and water bodies/
waterlogged areas. Bands used were selected after examining the spectral reflectance
pattern of salt-affected soils. These indices were primarily related to the spectral
reflectance pattern of salts present in soil strata responsible for salinity/alkalinity. The term
alkalinity or sodicity is used for sodic soils that are dominant in the test area and also
considered as the principal water quality concerns in irrigated areas receiving such waters
(Ayars and Tanji, 1999).
The overall trend pattern of the spectral reflectance had been observed similar in all
scenes as shown in Fig. 3, indicating quite clearly the relative higher reflectance for saltaffected soils compared to that of all other classes. In order to evidence the saline zones and
suppressing those with vigorous vegetation, the investigation of various indices (for
salinity and vegetation) were carried out (Khan and Sato, 2001) with the aim to compare
their effectiveness for the study area and the satellite sensor used. In view of spectral
reflectance of individual bands approximated for salt-affected soils, various combinations
were tested for LISS-II sensor (Khan, 2002) but the followings main indices, including the
water index (explained in the next section of classification), performed relatively better
were as:
salinity index SI

B1  B3

normalized differential salinity index NDSI

brightness index BI

B32 B42

B3  B4
B3 B4


where in all above expressions (1)(3) B1, B3 and B4 are first, third and fourth spectral
bands, respectively.
The data recorded in the third spectral band were also used in the analysis as an index
based on its especial reflectance characteristics for salt class, to check an idea to use just the
spectral satellite data in this band for salt-affected areas delineation. To play a role of an
index this data should be normalized; for example, that can be normalized to the average
value of an image for salt class. Salt-affected soils are usually characterized by poorly
developed vegetation areas, thus, such state of stressed vegetation could be an indirect sign

N.M. Khan et al. / Agricultural Water Management 77 (2005) 96109


of the presence of salts in the soils. Two well-known vegetation indices were, therefore,
also included in the analysis as:
normalized differential vegetation index NDVI

B4  B3
B3 B4


and the ratio of two spectral bands as:




In the process of choosing the index that produces the best results in determining saltaffected areas, it was supposed that the values of that index in two other adjacent points
belonged to two different classes (salt versus any non-salt class) should be different enough
in the scale of values of that index. To simplify the comparison of different indices, we
operated with the ratio of the indices for two adjacent classes, one of them being a non-salt
class. It should be pointed out that such ratio takes different values from one satellite scene
to another, which is caused by non-concurrent changes in spectral characteristics of
different surfaces. We concluded finally to mention here two most suitable solutions that
gave the best results for our data analysis approach. Using the third channel data of the
instrument as an index (SI3) for the presence of salts assessment in the area could be one
good solution, and other is normalized differential salinity index (NDSI) or normalized
differential vegetation index (NDVI), either could be used as the absolute values (NDVI or
NDSI) are equal, as explained below.
Fig. 4 shows the ratio of the signals measured in the third channel (SI3) for different
classes: Salt, Town and Crop to the third channel signals of the Salt class for every
processed image. It can be concluded that in most cases the signal in the third channel of the
instrument received from salt-affected areas is substantially higher than the signal received
from the other surfaces (the water classes are not shown in the figure because they have

Fig. 4. Variation of the ratio of signatures in the third spectral band (SI3) for different classes to the B3 signature of
the Salt class.


N.M. Khan et al. / Agricultural Water Management 77 (2005) 96109

very low signal in this spectral band and can not stand in the way of delineating salt
surfaces). The only one class that could mimic the Salt class was the Town class
observed in this case study. Some training sites belonged to the class Town had the same
spectral signature as the class Salt in June, as shown in Fig. 4, in case of image no. 3. This
problem might be due to the fact that villages have muddy roofs similar to dry barren soils
along with patchy saline areas within the village, therefore, showing almost similar
signatures to saline soils.
The next index that gave satisfactory results in retrieving Salt class areas was NDSI
(Eq. (2)). One may also use NDVI (Eq. (4)), since these two indices are equal in absolute
values. Fig. 5(a) shows for six different images that the NDVI values for the four classes:
Crop, Canal Town and Salt range from the maximum to the lowest, respectively. It
may be easy to decide which classes are lying closer to Salt class in terms of NDVI-index
values if one observes results in Fig. 5(b) where is shown the ratio of the NDVI indices
relative to those classes to the NDVI index of the Salt class. It can be concluded again that
similarly in using measurements of the third channel as mentioned in above paragraph, the

Fig. 5. (a) Variation of NDVI for different land surface classes, (b) ratio of NDVI calculated for different classes to
the NDVI derived for the Salt class at various times: (1) 31/03, (2) 22/04, (3) 05/06, (4) 15/10, (5) 06/11, (6) 20/12.

N.M. Khan et al. / Agricultural Water Management 77 (2005) 96109


main difficulty in retrieval of salt-affected areas from satellite data lies in distinguishing
Salt and Town classes. To overcome this difficulty one can use up-to-date topographic
maps to get information about the settlement/urban area boundaries. Such areas can be
vectorized and excluded from the calculation process while monitoring salt-affected soils.
If topo-sheets are not available, then a possibility is to avoid using that specific period when
satellite data do not provide for appropriate distinction between these two classes.
According to this study, the optimal time period for taking satellite data to assess saltaffected soils using the NDVI (or NDSI) index is March (Fig. 5(b); Table 2). Then it is
possible to avoid any mix-pixels confusion with urban areas, which are expected during
peak summer. Each index image was generated using GIS IDRISI through mathematical
operator functions. The respective histograms were then created using stretch module
functions for better understanding the distribution of the pixels. A reclassification was then
performed after visualizing the histogram for water, salt and vegetation classes to create a
new image for that particular class.

4. Classification approach
In this last part of the paper, new possibilities of classification on the base of using indices
are discussed. This part of the work has been accomplished using special functions of GIS,
called image enhancement procedures. In GIS IDRISI, they are FILTER (different kinds of
filtering), STRETCH (rescaling image values to fall within a given range) and COMPOSIT
(producing new images with new characteristics on the base of received satellite data). These
GIS functions increase the ratio of signal to noise in satellite data and suggest the most
efficient use of important information contained in different spectral channels.
The problem of delineation of salt-affected areas is often accompanied by the problem
of determination of regions under risk of waterlogging. The latter task is usually solved
using the fourth band (NIR) data processing (Dwivedi and Sreenivas, 1998; Sujatha et al.,
2000). The main obstacle for obtaining a good classification is the low contrast between
areas to be identified relative to other surface classes. To overcome this difficulty, the
following water index was suggested for the study area (Khan, 2002):



where, COMP124 is a composite image produced on the basis of B1, B2 and B4 spectral
data, i.e.:
COMP124 B1 6  B2 36  B4


It is obtained by using the COMPOSIT module of IDRISI for Windows with the type of
contrast stretch specified to be linear with a saturation level of 2.5%. The parameter
B4_STR concerns stretched satellite data received in the fourth spectral band (NIR) using
the stretch module in the regime of histogram equalization.
The analysis of the image and its histogram shows that the values of water index for
irrigation channel pixels are close to 0, being more than 3.0 for other water objects (lakes/


N.M. Khan et al. / Agricultural Water Management 77 (2005) 96109

ponds and waterlogged areas). The main achievement of the described procedure is the act
of getting an image where water objects (including waterlogged areas) are definitely
separated from other surface classes. This has become possible due to stretching the B4
satellite image and using it in the way that is described in expression (6). This index image
was used in further classification procedure as described below.
To get a classified image of the surface under study one can use various classification
procedures of GIS. Here, we have tried to describe how to get satisfactory results using
unsupervised classification criteria. In IDRISI for Windows, unsupervised classification is
usually applied to a composite image. The routine procedure to get that composite is
through the use of satellite data sets of different spectral channels. It could be B1, B2 and
B4 data of the LISS-II sensor. Obviously, a composite image can be produced on the other
basis and composition according to different spectral ranges available in various
commercial satellites. In our case, we have calculated a specialized composite image using
three indices: NDSI (Eq. (2)), water index (Eq. (6)) and NDVI (Eq. (4)). Then we have
applied to that image an ISOCLUST module of IDRISI for Windows under unsupervised
classification procedure. The result is shown in Fig. 6 where salt-affected areas are
identified in white tones in the resulted image. The comparison with topographic maps

Fig. 6. Unsupervised classification (using the ISOCLUST Module) relative to salinity (white), vegetation (gray)
and water bodies (dark).

N.M. Khan et al. / Agricultural Water Management 77 (2005) 96109


proves reliable results of the classification procedure adopted. One of the main advantages
of the described procedure is that water objects can be classified correctly in this case with a
high probability of success.
The next example of classification of satellite data described below uses a different
approach based on the principal component analysis. This sort of analysis of a set of images
produces a new set of images, i.e., components that are uncorrelated with each other and
explain progressively less of the variance found in the original set of spectral bands. In our
case, we have built principal components (PCACMPi) with the use of all satellite
measurements: B1, B2, B3 and B4 spectral bands. Results show that the first component
explains 92.52% of the variance in the original set of band and the second one explains
6.81% of variance, giving together more than 99%. The percentage of the first component
(PCACMP1) was not enough to use it alone for classification. We have, therefore, taken the
two first components and produced a composite image using as the following:


where, PCACMPi is the ith component in the principal component analysis.

The described composite image has been used as a base for unsupervised classification
(ISOCLUST module). The analysis of the results (Fig. 7) proves also a good quality of
classification. The classes Salt, Crop, Canal and Lake are delineated correctly in
the displayed part of the satellite image (Fig. 7), which was observed as the most difficult

Fig. 7. Classified results after principal components analysis of LISS-II data.


N.M. Khan et al. / Agricultural Water Management 77 (2005) 96109

land-use/cover area in the classification procedure. We have applied this procedure to

different scenes taken at different times in the year (Table 2) to check the potential of PCA
approach. It always produced good results for classification of salinity prone areas when
compared with the available topographic sheets and field reports/maps of WAPDA.
However, some other classes like Town, Bare land and some Crop types have not
come-up with clear difference in PCA analysis.

5. Conclusions
The two most suitable solutions producing the best results for assessing saline lands
using LISS-II sensor data were the selection of third channel due to its substantial higher
spectral reflectance compared to other wavelength ranges, and the NDVI or NDSI indices,
that are same in their absolute values.
The main difficulty in retrieval of salt-affected areas from satellite data is to distinguish
between Salt and Town classes. This can be overcome using up-to-date topographic
maps to get information about the settlement/urban area boundaries and excluding it after
vectorizing from the calculation process while monitoring salt-affected soils. The other
possibility when that information is not available is to avoid using satellite scenes where
such distinction cannot be made but other images taken at different times. The optimal time
period for taking satellite data (IRS LISS-II) in order to assess salt-affected soils in the test
area with the NDVI (NDSI) index is March to avoid spectral confusion of mixing with
urban areas that is occurring in peek summers times.
To identify various types of water objects (channels, rivers and waterlogged areas) the
newly developed water index may be used, which is capable to correctly differentiate water
objects; it can be used with confidence for the classification of waterlogged areas. This
became possible because of using a B4 stretched image (NIR) of the LISS-II sensor in the
model used for classification procedure.
It is concluded that a simple but practical approach based on remote sensing
indicators as well as the specialized classification procedures through PCA has shown
promising potential in delineating the hydrosalinized degraded lands for this specific
area using LISS-II data. This approach is the first study adopted for the selected area of
Indus basin (Pakistan) using the LISS-II sensor of IRS-1B satellite. The approach stated
above has also been carried out in further studies to monitor and assess the status and
cause-effect relationships of waterlogging, groundwater quality indicators and salinization through an integrated methodology of remotely sensed and field data as input for
GIS analysis.

The authors are grateful to the Japan Society for the Promotion of Sciences (JSPS) for
the fellowship provided for this research study. Thanks are due to Dr. S. Uchida of Japan
International Research Center for Agricultural Sciences (JIRCAS) for providing the IRS1B satellite data, and the International Waterlogging and Salinity Research Institute

N.M. Khan et al. / Agricultural Water Management 77 (2005) 96109


(IWASRI), Lahore, Pakistan, for providing the necessary field data, maps and reports used
in the study analysis. We also acknowledge the contributions of the reviewers of this

Ayars, J.E., Tanji, K.K., 1999. Effects of drainage on water quality in arid and semiarid lands. In: Skaggs, R.W.,
van Schilfgaarde, J. (Eds.), Agricultural Drainage. ASA-CSSA-SSSA, Madison, WI, USA, pp. 831867.
Casas, S., 1995. Salinity assessment based on combined use of remote sensing and GIS. In: Use of Remote Sensing
Techniques in Irrigation and Drainageunknown:book, FAO, Rome, Italy, pp. 141150.
Dwivedi, R.S., 1992. Monitoring and the study of effects of image scale on delineation of salt affected soils in the
Indo-Gangetic plains. Int. J. Remote Sens. 13, 15271536.
Dwivedi, R.S., Sreenivas, K., 1998. Delineation of salt-affected soils and waterlogged areas in the Indo-Gangetic
plains using IRS-1C LISS-III data. Int. J. Remote Sens. 19, 27392751.
Eastman, R.J., 1995. IDRISI for Windows, Users Guide. Clark University, Worcester, MA, USA.
Goossens, R., De Dapper, M., Gad, A., Ghabour, Th., 1993. A model for monitoring and prediction of soil salinity
and waterlogging in the Ismailia area (Egypt) based on remote sensing and GIS. In: Proceedings of the
International Symposium on Operationalization of Remote Sensing, vol. 6, ITC Enschede, The Netherlands,
pp. 97107.
Khan, N.M., Sato, Y., 2001. Monitoring hydro-salinity status and its impact in irrigated semi-arid areas using IRS1B LISS-II data. Asian J. Geoinform. 1 (3), 6373.
Khan, N.M., 2002. Analysis on Environmental Land Degradation Caused by Hydrosalinity in Semi-arid Irrigated
Regions. Ph.D. Thesis, The University of Tokyo, Japan.
Ghassemi, F., Jakeman, A.J., Nix, H.A., 1995. Salinization of Land and Water Resources: Human Causes, Extent,
Manegment and Case Studies. CAB International, p. 526.
Metternichit, G., Zinck, J.A., 1997. Spatial discrimination of salt- and sodium-affected soil surfaces. Int. J. Remote
Sens. 18 (12), 25712586.
Rao, B.R.M., Dwivedi, R.S., Venkataratnam, L., Ravishankar, Thammappa, S.S., Bhargawa, G.P., Singh, A.N.,
1991. Mapping the magnitude of sodicity in part of the Indo-Gangetic plains of Uttar Pradesh, northern India
using Landsat-TM data. Int. J. Remote Sens. 12 (3), 419425.
Sharma, R.C., Bhargawa, G.P., 1988. Landsat imagery for mapping saline soils and wetlands in north-west India.
Int. J. Remote Sens. 9, 6984.
Srivastava, A., Tripathi, N.K., Gokhale, K.V.G.K., 1997. Mapping groundwater salinity using IRS-1B LISS II data
and GIS techniques. Int. J. Remote Sens. 18, 28532862.
Sujatha, G., Dwivedi, R.S., Sreenivas, K., Venkataratnam, L., 2000. Mapping and monitoring of degraded lands in
part of Jaunpur district of Uttar Pradesh using temporal spaceborne multispectral data. Int. J. Remote Sens. 21
(3), 519531.
Szabolcs, I., 1992. Salinization of soil and water and its relation to desertification. Desertificat. Control Bull. 21,
Tabet, D., Vidal, A., Zimmer, D., Asif, S., Aslam, M., Kuper, M., Strosser, P., 1997. Soil salinity characterization in
SOPT imagesa case study in one irrigation system of the Punjab, Pakistan. In: Physical Measurements and
Signatures in Remote Sensingunknown:book, Guyot & Phulpin Eds-Balkema, pp. 795800.
WAPDA, 1981. Soil salinity Survey, vol. 1. Soil Salinity and Watertable Survey Directorate, Planning Div.,
Lahore, Pakistan.
WAPDA/SMO, 19931994. Monitoring and Evaluation Fourth Drainage Project, Annual Report, vol. 1, 2. SM
publications. Monitoring and Evaluation Directorate, SCARP Monitoring Organization (SMO), WAPDA,
Lahore, Pakistan.