Characterizing soil salinity in irrigated agriculture using a remote sensing approach Akhtar Abbas a,⇑,1 , Shahbaz Khan b , Nisar Hussain c , Munir A. Hanjra d , Saud Akbar e a Cooperative Research Centre for Irrigation Futures, Darling Heights Qld, Australia b UNESCO Division of Water Sciences, Paris, France c College of Agriculture, D.G. Khan, Pakistan d Charles Sturt University, Wagga Wagga NSW, Australia e Industry & Investment NSW, Richmond NSW, Australia article info Article history: Received 27 February 2008 Received in revised form 16 September 2010 Accepted 7 December 2010 Available online 14 December 2010 Keywords: Irrigation Soil salinity Spatial analysis Waterlogging Reclamation abstract Managing salinity in irrigated agriculture is crucial for minimising its negative environmental impacts and for ensuring the long-term sustainability of irrigated agriculture. It demands establishing rapid mon- itoring systems that help develop sustainable management plans. Remote sensing offers several advan- tages over the conventional proximal methods to map and predict areas at salinity risk. This paper presents an integrated approach to characterize soil salinity using remotely-sensed data in the District Faisalabad, Punjab, Pakistan. The IRS-1B LISS-II digital data was acquired and analysed in combination with field data and topographical maps. Remotely-sensed data based salinity indices or band combina- tions were developed to monitor the occurrence pattern of salt-affected soils. Using supervised maximum likelihood classification, the images were classified into eight land use classes with an overall accuracy of around 90%. The classified images showed that 22.2% of the total area was under salt-affected soils in 1992. The occurrence pattern of salt-affected soils varied with positive and negative trends during 1992–1995 to a minimum of 10.6%. The delineation analysis into levels of saline soils revealed three types based on USDA classification (USDA, 1954). The slightly saline, moderately saline and strongly sal- ine soils during 1992 were in the order of 15%, 3%, and 1% respectively. The interactive behaviour of salin- ity and sodicity and their combinations showed that saline-sodic soils occurred predominantly ranging from 6.9% to 17.3% of the salt-affected soils. The shallow watertable was found to be of hazardous quality in 28% of the study area. The relationship between salt-affected soils, waterlogged soils and groundwater quality revealed that 60–70% of the salt-affected soils occurred in shallow watertable areas during 1992– 1995. The reuse of poor quality groundwater for irrigation and the failure of tile drainage system in the area are likely to further increase the risk of salinisation in the Indus Basin of Pakistan. Ó 2010 Elsevier Ltd. All rights reserved. 1. Introduction Soil salinity in semi-arid regions where crop water require- ments are augmented by irrigation supplies is a major concern for the sustainability of irrigated agricultural systems. It causes se- vere environmental degradation that impedes crop growth and overall regional production. Irrigation-induced salinity occurs in large and small irrigation systems alike. Salinisation is a worldwide problem, particularly in irrigated lands, which are extensively irrigated and are poorly drained. FAO (2007) reported that at the global scale more than 77 mha (million hectares) of land are salt- affected and about 43 mha are attributed to secondary salinisation. Estimates indicate that about one-third of the irrigated lands in the major countries with irrigated agriculture is badly affected by salinity or is expected to be salinised in the near future. Current estimates of the salt-affected soils as percent of the irrigated lands for different countries are: 27% for India, 28% for Pakistan, 13% for Israel, 20% for Australia, 15% for China, 50% for Iraq, and 30% for Egypt (Stockle, 2007). In only New South Wales (NSW) state of Australia, salinity is estimated to affect 15% of the irrigated land and, in recent years, many farmers have abandoned their rice fields due to the incidence of soil salinity. At global scale, soil salinisation is spreading at a rate of up to 2 mha per year, which offsets a sig- nificant portion of the crop production that is otherwise achievable by using the best management practices at a system level. Pakistan has a total area of 79.6 mha, and 22 mha of cultivated land (GOP, 1999) with approximately 6 mha affected by the salin- ity within the irrigation areas (Mahmood et al., 2003). This envi- ronmental degradation is of staggering proportions and is one of the factors preventing the achievable potential of agricultural pro- ductivity. Geologically, the Indus plain, which is the major land for irrigated agriculture in Pakistan, was formed from alluvium depos- ited by rivers into shallow sea. The receding sea left residues of 1474-7065/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.pce.2010.12.004 ⇑ Corresponding author. E-mail address: [email protected](A. Abbas). 1 Formerly with CSIRO Land and Water, Canberra ACT, Australia. Physics and Chemistry of the Earth 55-57 (2013) 43–52 Contents lists available at ScienceDirect Physics and Chemistry of the Earth journal homepage: www.elsevier.com/locate/pce
Transcript
Physics and Chemistry of the Earth 55-57 (2013) 43–52
Contents lists available at ScienceDirect
Physics and Chemistry of the Earth
journal homepage: www.elsevier .com/locate /pce
Characterizing soil salinity in irrigated agriculture using a remote sensing approach
Akhtar Abbas a,⇑,1, Shahbaz Khan b, Nisar Hussain c, Munir A. Hanjra d, Saud Akbar e
a Cooperative Research Centre for Irrigation Futures, Darling Heights Qld, Australiab UNESCO Division of Water Sciences, Paris, Francec College of Agriculture, D.G. Khan, Pakistand Charles Sturt University, Wagga Wagga NSW, Australiae Industry & Investment NSW, Richmond NSW, Australia
a r t i c l e i n f o a b s t r a c t
Article history:Received 27 February 2008Received in revised form 16 September 2010Accepted 7 December 2010Available online 14 December 2010
Managing salinity in irrigated agriculture is crucial for minimising its negative environmental impactsand for ensuring the long-term sustainability of irrigated agriculture. It demands establishing rapid mon-itoring systems that help develop sustainable management plans. Remote sensing offers several advan-tages over the conventional proximal methods to map and predict areas at salinity risk. This paperpresents an integrated approach to characterize soil salinity using remotely-sensed data in the DistrictFaisalabad, Punjab, Pakistan. The IRS-1B LISS-II digital data was acquired and analysed in combinationwith field data and topographical maps. Remotely-sensed data based salinity indices or band combina-tions were developed to monitor the occurrence pattern of salt-affected soils. Using supervised maximumlikelihood classification, the images were classified into eight land use classes with an overall accuracy ofaround 90%. The classified images showed that 22.2% of the total area was under salt-affected soils in1992. The occurrence pattern of salt-affected soils varied with positive and negative trends during1992–1995 to a minimum of 10.6%. The delineation analysis into levels of saline soils revealed threetypes based on USDA classification (USDA, 1954). The slightly saline, moderately saline and strongly sal-ine soils during 1992 were in the order of 15%, 3%, and 1% respectively. The interactive behaviour of salin-ity and sodicity and their combinations showed that saline-sodic soils occurred predominantly rangingfrom 6.9% to 17.3% of the salt-affected soils. The shallow watertable was found to be of hazardous qualityin 28% of the study area. The relationship between salt-affected soils, waterlogged soils and groundwaterquality revealed that 60–70% of the salt-affected soils occurred in shallow watertable areas during 1992–1995. The reuse of poor quality groundwater for irrigation and the failure of tile drainage system in thearea are likely to further increase the risk of salinisation in the Indus Basin of Pakistan.
� 2010 Elsevier Ltd. All rights reserved.
1. Introduction
Soil salinity in semi-arid regions where crop water require-ments are augmented by irrigation supplies is a major concernfor the sustainability of irrigated agricultural systems. It causes se-vere environmental degradation that impedes crop growth andoverall regional production. Irrigation-induced salinity occurs inlarge and small irrigation systems alike. Salinisation is a worldwideproblem, particularly in irrigated lands, which are extensivelyirrigated and are poorly drained. FAO (2007) reported that at theglobal scale more than 77 mha (million hectares) of land are salt-affected and about 43 mha are attributed to secondary salinisation.Estimates indicate that about one-third of the irrigated lands in themajor countries with irrigated agriculture is badly affected bysalinity or is expected to be salinised in the near future. Current
ll rights reserved.
T, Australia.
estimates of the salt-affected soils as percent of the irrigated landsfor different countries are: 27% for India, 28% for Pakistan, 13% forIsrael, 20% for Australia, 15% for China, 50% for Iraq, and 30% forEgypt (Stockle, 2007). In only New South Wales (NSW) state ofAustralia, salinity is estimated to affect 15% of the irrigated landand, in recent years, many farmers have abandoned their rice fieldsdue to the incidence of soil salinity. At global scale, soil salinisationis spreading at a rate of up to 2 mha per year, which offsets a sig-nificant portion of the crop production that is otherwise achievableby using the best management practices at a system level.
Pakistan has a total area of 79.6 mha, and 22 mha of cultivatedland (GOP, 1999) with approximately 6 mha affected by the salin-ity within the irrigation areas (Mahmood et al., 2003). This envi-ronmental degradation is of staggering proportions and is one ofthe factors preventing the achievable potential of agricultural pro-ductivity. Geologically, the Indus plain, which is the major land forirrigated agriculture in Pakistan, was formed from alluvium depos-ited by rivers into shallow sea. The receding sea left residues of
44 A. Abbas et al. / Physics and Chemistry of the Earth 55-57 (2013) 43–52
salts in soil profiles and aquifers. Minerals in parent rocks also re-leased significant quantities of salts into the soil through weather-ing. Under historic climatic conditions, salts released throughweathering were not leached out and consequently, accumulatedwithin the soil profiles. Moreover, the irrigation systems (the mostcontiguous irrigation system in the world) introduced during thesecond half of the nineteenth century caused salt accumulationin the soil profile leading to the secondary salinisation (Abbas,1999). A schematic representation of the development of the sec-ondary salinisation is shown in Fig. 1.
Among all categories of salt-affected soils in Pakistan between 2and 3 mha is understood as completely wasteland due to high levelof salinity and sodicity (Qureshi et al., 1993). Estimates furthershow that 25% and 40% of the irrigated lands in the Punjab andSindh are salt-affected respectively, and the livelihoods of about10–20 million people are affected due to these salt-affected landswith impoverished productivity and increased threats to ecosys-tem (Barrett-Lennard and Hollington, 2006). Seepage from supplycanals, extensive network of on-farm and near-farm watercourses;and flooded irrigated fields are known as primary causes of second-ary soil salinity (Abbas et al., 2005; Kijne, 2006). Waterlogging andsalinity have distressing social and economic effects on farmingcommunities (Khan and Hanjra, 2008). Moreover, the increasingshortage of fresh water resources is likely to trigger increasedenvironmental damage (Khan and Hanjra, 2009). The falls in riverflows and reduced irrigation supplies due to persistent droughthave disturbed the water ecosystems resulting into food insecurityand non sustainable farming systems (Hanjra and Qureshi, 2010;Thenkabail et al., 2010).
Conserving fertile lands and environmental stewardship havebeen indeed complementary goals for any land reclamation pro-gram in Pakistan agriculture. However, the solutions were neither
Fig. 1. Process of secondary salinisation in irrigated areas (modified from FAO,2002).
cheap nor easy. For example, the suite of Salinity Control and Rec-lamation Projects (SCARPs) designed and implemented by Waterand Power Development Authority (WAPDA) Pakistan failed tocontrol salinity due to exorbitant costs and institutional failure.The key lessons learnt from this experience demanded for strongcommitment in all jurisdictions plus the development of local landand water management plans (LWMPs) through effective stake-holders’ engagement. However, such actions remain a dream andhighly productive lands would further continue to retire unlessreliable and updated information on the soil salinity and the linkedimpediments is not made available. Thus, monitoring the status ofirrigated soil salinity in temporal and spatial extent is a primaryconcern for saving the productive irrigation systems.
The salinisation is inevitable and likely to expand becausearound 70% of the available surface water with varying degree ofquality used for irrigation worldwide is continuously adding mil-lions of tons of salts to productive lands. In future, to cope withthe rapid population growth and increasing food demand, moredry lands will be put into agricultural production or croppingintensity must increase manifolds. Such extensive cropping willbe mainly achieved through irrigation and thus accelerates thesalinisation risks (Metternichet and Zinck, 2003). Monitoring andevaluation using visual interpretations, field surveys, and labora-tory analyses have been in place since decades. However the ques-tion is, ‘‘do we have a simple tool for keeping track of continuouschanges that help anticipate further degradation’’? This paperbuilds on previous work (Abbas and Khan, 2007) and uses remotesensing data for characterizing salinity in a major irrigation districtin the upper Indus Basin of Pakistan as a case example. The specificobjectives of the paper are: (1) to develop a robust method forcharacterizing irrigated salinity based on remote sensing data;(2) to determine the temporal and spatial changes in salt-affectedsoils; (3) and to identify the association of occurrence of salt-af-fected soils with shallow watertable of varying quality.
2. Remote sensing for soil salinity – past perspective
What kind of information can remote sensing provide for man-aging salinity in irrigated agriculture? Remote sensing has the pos-sibility to predict soil salinity, performance diagnosis, and impactassessment (Bastiaanssen et al., 2000). It saves labour, time and ef-forts when compared to conventional field measurements. Theinformation derived from remote sensing tools is significantlyobjective, collected in a systematic way, and covers a wide area.Thus, remote sensing data has the ability to capture informationon salt-affected soils in both spatial and temporal extents (Abbas,1999). However, remote sensing based approaches for salinitymapping has met with limited success compared with the otherapplications in irrigated agriculture (Bastiaanssen et al., 2000; Met-ternichet and Zinck, 2003; Lagerloef, 2009). The development ofmethods to map salinity using remote sensing data in combinationwith field data have been the objective of several studies duringthe past two decades. Several algorithms and models have beendeveloped to process satellite remote sensing data. Past researchshows mapping and assessing of soil salinity using remote sensingtools like aerial photography, videography, infrared thermometryand multispectral scanners, RADAR and LIDAR imagery. For exam-ple, Dwivedi and Rao (1992) adopted a three-band combination ofLandsat TM data for identifying salt-affected soils and the bands 1,3, and 5 were found to be more informative. Tripathi et al. (1997)developed two indices using remote sensing data to monitor soilalkalinity in the Indo-Gangnetic plains. Dwivedi and Sreenivas(1998) demonstrated the potential of image transformations anddescribed that the principal component analysis, image differenc-ing and ratioing of the bands provided substantial information
Table 1Characteristics of the images used.
Satellite Band Spectral range (lm) Spatial resolution (m)
A. Abbas et al. / Physics and Chemistry of the Earth 55-57 (2013) 43–52 45
about salt-affected soils. The index transformation made the salin-ity information more prominent while decreasing the effect ofother features of land use/land cover to a minimum. Verma et al.(1994) combined the TM false-colour composite (FCC) with ther-mal data to solve the problem of spectral similarity where thedull-white tones of salt-affected and sandy soils had been difficultto distinguish. Metternichet and Zinck (1997) applied a synergisticapproach to map salt-affected surfaces, combining image classifi-cation with field observations and laboratory results. Also vanLynden and Mantel (2001), Metternichet and Zinck (2003), andFarifteh et al. (2006) presented the reviews of assessment of salt-affected soils using remote sensing, solute modelling, geophysics,and data integration techniques. Ambast et al. (1999) tested anew approach to quantify the physical appearances of salinitythrough biophysical parameters (e.g., surface albedo, fractionalvegetation coverage, leaf area index (LAI), surface resistance, evap-orative fraction of salt-afflicted crops).
Some worldwide examples of remote sensing applications tosalinity issues include: extracting soil salinity from the remote sens-ing data using Kauth–Thomas (K–T) transformation in the Yanggaobasin of China (Peng, 1998); remote sensing-based salinity predic-tion model for a large estuarine lake in the US Gulf of Mexico (Wangand Xu, 2008); mapping salinity within the environmental contextof Lower Cheliff Plain, Algeria (Douaoui et al., 2006); temporal andspatial changes of salinity using fuzzy logic, remote sensing andGIS in a South American case study (Metternicht, 2001); analysisof salinisation dynamics in Hetao irrigation district of North China(Yu et al., 2010); assessment of surface and sub-surface waterloggedirrigation areas in Bihar state, India (Chowdary et al., 2008); assess-ing the spatial extent of dryland salinity in the Wheatbelt of WesternAustralia (Malins and Metternicht, 2006); quantitative analysis ofsalt-affected soil reflectance spectra using two modelling tools ofpartial least square regression (PLSR) and artificial neural network(ANN) in the Netherlands and Hungary (Farifteh et al., 2007); anintegrated approach of using remote sensing, GIS and geostatisticaltechniques to spatially predict the soil salinity and sodicity usingindices such as salinity index, normalised salinity index andbrightness index in semi-arid region of NSW, Australia (Odeh andOnus, 2008); hyper-spectral remote sensing of salt marsh vegetationin Venice lagoon, Italy (Silvestri et al., 2003); mapping soil salinity inthe lake Texcoco, Mexico (Fernández-Buces et al., 2006); andprediction of the soil-depth salinity-trend in a vineyard in SouthAfrica (de Clercq et al., 2009).
In Pakistan, the applications of the state-of-the-art remote sens-ing technologies for monitoring salt-affected areas have been verylimited. Few studies like Farooq and Din (1980), Khan and Siddiqui(1987), and WAPDA (1981, 1984) are reported in the literature. Themain focus of these studies had been using visual interpretation ofthe photography and the ‘car-stuck formula’ of the field visits thatprovided rough assessments of only highly saline areas – showingpatchy vegetation and salt efflorescence at the soil surface. A majorattempt to classify salinity using remote sensing was carried out byTabet (1995, 1999) and Vidal et al. (1998). They used SPOT XSsatellite data and developed vegetation and brightness indices.These indices were used to classify salinity for vegetative andnon-vegetative areas. The resulting classification allows identifica-tion of highly saline and non-saline areas, but areas with low tomedium salinity levels were not clearly distinguished. Paucity ofinformation about salinity classes remained because of the puzzle– ‘‘uneven crop growth was from impact of salinity or poor agricul-tural practices’’. Also Khan et al. (2005) conducted hydrosaline landdegradation assessment in part of the Indus Basin of Pakistan usingremote sensing data. This paper builds on our previous work(Abbas and Khan, 2007) and uses spatial analysis to characterizethe spatial and temporal changes in soil salinity in irrigatedagriculture. The present paper is an attempt to overcome these
constraints and a step forward to use remote sensing data forsalinity assessment and the associated processes in the upperIndus Basin of Pakistan.
3. Study area
The climate of Pakistan is categorised as subtropical, semi-aridwith rainfall varying from about 150 mm in the south to 1200 mmin the north over the Karakoram Range. The study area is located inthe District Faisalabad, the Central Punjab of Pakistan and lies inthe south-western part of ‘‘Rachena Doab’’. ‘Doab’ in local dialectdefines the region lying between two rivers, and ‘Rachena’ is de-rived from two rivers – the Ravi and the Chenab, bounding thisarea in the upper Indus Basin of Pakistan. The study area hasgeo-coordinates of 31�020 to 31�450N and 72�500 to 73�220E andconsists of two subregions: S-I and S-II (‘S’ stands for schedule inline with the implementation of the project). The climate of thearea is characterised by large seasonal fluctuations of ambienttemperatures and rainfall. The summer is long and hot, from Aprilto September; with maximum temperature soaring to 49 �C. Win-ter maximum is 27 �C during the day and the minimum falls nearzero at night. The annual average rainfall is about 350 mm. Thisindicates that the region is semi-arid such that agriculture reliesheavily on irrigation, causing salinity in the flat plains.
The area is mostly coved by extensive irrigated lands and hasbeen well known for its potential agricultural productivity. Thesoils are mainly alluvial deposits with wide range of moderatelycoarser to moderately fine medium textured material. Physio-graphically, the area is nearly level to very gently sloping landswith mean elevation of 190 m above sea level. The average landgradient is 0.02% from north–south to south–west directions. Theaquifer depth varies from 300 m to 500 m. The phreatic surfacewas 20–30 m deep before the gravity irrigation system was intro-duced in 1856 during the British regime (Abbas, 1999). By 1960s,the watertable rose close to the surface creating waterlogged con-ditions that resulted in secondary salinisation. Several hundredhectares have undergone land degradation by salinity and the agri-culture-based regional economy had been badly affected.
4. Materials and methods
Two types of data are required to complete any sort of mappingwith remotely-sensed data. The first is the remotely-sensed data it-self, and the second (and equally important) is the ground truthdata. Without ground truth data, remotely-sensed data, indeed, isof limited value. In the present study, the satellite images fromIRS-1B LISS-II (Indian Remote Sensing Linear Imaging Self Scan-ning) were used. The details on bands with spatial resolution aregiven in Table 1.
The following images (path 033, row 045) for 1992–1995 wereused in this study and are given in Table 2. These images have areasonable temporal coverage including both summer and wintercropping seasons of the Faisalabad District.
The ground truth data consisting of soil samples for ECe andSAR analysis was taken from 70 locations within the study area.This comprised of two groups: (1) normal plots, and (2) affected
Table 2Dates of satellite images acquired for the study.
18 June 1992 28 October 1992 5 June 199315 October 1993 14 June 1994 15 November 19941 June 1995 2 November 1995
46 A. Abbas et al. / Physics and Chemistry of the Earth 55-57 (2013) 43–52
plots. Geo-coordinates were recorded for each location, furtherused in the analysis. Topographic maps from the Soil Survey ofPakistan documented as 44 E/2, 44 E/3, 44 A/14, and 44 A/16 atscale 1:50,000 were collected and used in the analysis. Layoutmaps of irrigation canals and surface drains were collected fromWAPDA and were also used. ILWIS 2.2 for Windows system wasused for the remote sensing and GIS analyses. Other informationon ground truthing consisting of water and soil quality, soil types,crop rotations, crop yield, and irrigation facilities were collectedand used in the image registration and subsequent comparisons.Around 30 well defined and uniformly distributed ground controlpoints (GCPs) were selected from the topographic maps. Layoutmap of canals and surface drains were used in locating GCPs. Topo-graphic sheets were used to register satellite images and for basemap preparation. The GCPs (road intersections, canal and drainnetwork, railway crossings, rivers, bridges, and the airport) wereeasily identified both on the image and the topographic maps.The planimetric coordinates were measured from the topographicmaps and transferred to the image. The satellite images were geo-referenced to UTM projection using affine transformation method.The transformation matrix resulted in root mean square (RMS) er-ror or gamma (r) less than 0.5 pixel. After registering the images,the band data were corrected for geometric distortion by resam-pling using cubic convolution interpolation method. The flow chartof the various steps involved in the analysis is shown in Fig. 2.
The registered and corrected images were further used for landuse/land cover classification and spatial analysis. Supervised classi-fication, using maximum likelihood classification (MLC) algorithm,was performed for major land uses and salt-affected lands for bothwinter and summer cropping seasons. The false-colour composite(FCC) image was used as background image for selection of trainingdata sets. Separation of land use classes and their correct
Fig. 2. Flow chart of the processes used in the analysis.
classification is significant for the quality of results. To evaluatethe classification performance, two indicators, called separabilityindex and accuracy matrix, were examined. The separability indexwas calculated in terms of the normalised difference (ND) betweenany two land use classes. The normalised difference is the ratio ofthe difference between means of the two land use classes and thesum of their standard deviations. A normalised difference with avalue of 1 is considered as the threshold value to indicate the goodseparability between the land use classes (Abbas, 1999). It is givenby the equation:
ND ¼ x1 � x2
SD1 þ SD2ð1Þ
where ND is the normalised difference, x the mean of a class, and SDis the standard deviation of a class.
The combinations of multi-spectral bands (hereafter, the sali-nity indices) were developed (Abbas and Khan, 2007). They werebased on the assumption that this transformation would makethe required salinity information more prominent while suppress-ing the effects of other land use/land cover features. Similar ap-proach is reported in the literature and used in the salt-affectedregions (Dwivedi and Rao, 1992; Tripathi et al., 1997; Khan et al.,2005). Band selection was dependent on the spectral signals ex-pressed as the digital number (DN) pertaining to the land use clas-ses. The salt-affected soils had relatively higher reflectancecompared with other land use. Correlation analysis between bandsand ground truth data of ECe and SAR was carried out. It showedthat first three bands (B1–B3) had good relationship in terms ofthe correlation coefficient. It was found that first three bands (blue,green, and red represented as B1, B2, and B3, respectively) weresuitable in presenting salinity information by using these salinityrelationships. The infrared band (B4) had very low value of the cor-relation coefficient and was not included in the indices structure.The salinity indices, thus, developed were further analysed usingthe data of ECe and SAR. The following indices (represented byS1–S4) were developed.
where S1–S4 is the salinity indices, 1–4, B1–B4 is the four bands ofIRS data.
The indices data corresponding to geo-referenced locations ofECe and SAR were retrieved by overlay analysis. The empirical rela-tionships or models were established using the indices and groundtruth data. The models were tested and validated using the twosets of ground truth data from normal and affected plots respec-tively. The exponential function and polynomial function of thethird order were found promising with significant results.
To find the occurrence pattern of salt-affected lands over thestudy period of 4 years, the spatial distribution was determinedbased on the probability density. This transformation enhancedthe salinity features and classified the saline areas based on thepercent occurrence. This was accomplished by assigning binarynumbers to the classified images. Two groups of land use classeswere selected. The binary numbers of 0 and 1 were assigned as:‘1’ for the salt-affected soils and ‘0’ for the other classes (assumedthey belong to normal soils). All processed images, thus obtained,from 4 years satellite data were added. The occurrence probabilityfor salt-affected soils was assumed to be equal or more than 40%
A. Abbas et al. / Physics and Chemistry of the Earth 55-57 (2013) 43–52 47
and the image was finally processed into two classes: normal soilsand salt-affected soils.
Fig. 3. Spectral signals of land use/land cover classes for the scene of 28 October1992.
Fig. 4. False-colour composite of the Faisalabad region. White patches show salt-affected soils in the study area. (For interpretation of the references to colour in thisfigure legend, the reader is referred to the web version of this article.)
5. Results and discussion
5.1. Spectral behaviour of salt-affected soils
Normal and salt-affected soils demonstrated strong contrast inground surface conditions as reported previously (Abbas and Khan,2007). Salt-affected soils were characterised by the salt efflores-cence that had been accumulated over the soil surface by the cap-illary rise of low quality water. This distinctive feature of saltappearance over the soil surface was prominent and easily cap-tured in the satellite data. The spectral response of the salt-affectedsoils was higher than those of normal soils. It shows that salty soilsreflected more incident light energy in visible spectrum and thisresponse was extremely useful in the segregation of saline soils.Metternichet and Zinck (1997) found that salt-affected soils hadhigh spectral reflectance in the visible window particularly in theblue band. The soil salinity affects vegetation density and cropgrowth, explained by normalised differential vegetation index(NDVI) and the false-colour composite. The corrected images wereinterpreted using various interpretation keys like shape, tone, tex-ture, location, and the association. The increased brightness that iseasily detectable from the visible part of the spectrum was a clearindicator of the salts present in the soils. Hence, the presence ofsalts showed higher spectral signals (in terms of digital numbers)than that of other land features. The corresponding pixels werecross-validated by the results of soil analyses and field surveys.This showed that salinity could be interpreted and assessed usingremote sensing data. The spectral signals in the form of digitalnumber (DN) from four bands were plotted and shown in Fig. 3.
False-colour composite (FCC) image were generated using bandcombination of R:G:B = B3:B2:B1 (red, green, blue). A representativeselection for the scene taken on 18 June 1992 is as shown in Fig. 4.Pre-monsoon (before July) FCC images showed better contrast andprovided better visual interpretation of the salt-affected soils ascompared to post monsoon (after September) images. The FCCdemonstrated white patches followed by light bluish tone of differ-ent dimensions showing salt appearance at the soil surface, favour-ing the use of false-colour composite for the visual interpretationof salt-affected soils. The FCC clearly delineated salt-affected areasamong the cropland, patchy vegetations, and the land-not-under-crops (all other land features).
Table 3Accuracy measures for classification with spectral data (in percent) of 31 March 1993.
Land use Error ofomission
Error ofcommission
Classificationaccuracy (%)
Classificationreliability (%)
Bare land 2.79 2.39 0.97 0.98Fallow 7.78 8.89 0.92 0.91Cropland 0 0 1.0 1.0Salt-affected
soils0.2 0.2 1.0 1.0
Urban 0 0.31 1.0 1.0Water
bodies3.21 3.21 0.97 0.97
Waterlogged 2.91 2.91 0.97 0.98
5.2. Assessment of salt-affected soils
The classification process included the land use/land cover clas-ses that were specifically focussed to the study objectives discern-ible from data sources. The supervised classification, usingmaximum likelihood algorithm, was carried out. Initially, onlytwo major classes of ‘land-under-crops’ and ‘land-not-under-crops’were considered but they did not serve the purpose of discriminat-ing salt-affected soils very well from other features belonging tothis category. Particularly it mixed up with subclasses of bareand fallow. Thus, the category ‘land-not-under-crops’ was furthersplit into six classes; bare land, fallow, urban, water bodies, water-logged, and finally salt-affected soils. As described in Section 4, thenormalise difference (ND) was used as the indicator of separabilityof how classes would behave and were correctly separated. Land-under-crops (crop land) class had more uniform canopy cover inwinter cropping season due to widely grown wheat crop comparedwith summer cropping season. This is because of post-harvest fal-low lands previously under use by winter crops (e.g. cereals), earlygrowth stage of summer crops and likely lack of natural
vegetations caused by hot and dry summer season. The calcula-tions on normalised difference showed that the ND between ‘‘bareand salt-affected soils’’ was 3.05 while it was 0.53 between ‘‘waterbodies and waterlogged soils’’. This means that ‘‘bare andsalt-affected soils’’ were classified as two distinctive classes. How-ever, the separation as two distinct classes between ‘‘water bodiesand waterlogged soils’’ was not clear and was cumbersome.
48 A. Abbas et al. / Physics and Chemistry of the Earth 55-57 (2013) 43–52
The confusion matrix showed the overall classification perfor-mance in terms of the accuracy measures and is presented inTable 3. It demonstrated how many pixels were correctly classifiedto each land use class and its site-specific errors. However, thenumber of pixels classified under each class is not presented herefor brevity. The accuracy of all classes is quite promising as theaccuracy assessment is quite reasonable. The average accuracy,average reliability, and the overall accuracy for the image of 31March 1993 were 97.6%, 97.7% and 98.8% respectively. The landuse classification map is shown in Fig. 5.
Having successfully completed the classification, the salt-affected soils were found to be 10.1%, 14.2%, 9.4% and 5.7% of thetotal image area during pre-monsoon period for 18 June 1992, 05June 1993, 14 June 1994, 01 June 1995 respectively, and 8.5%,8.3%, 6.7% and 4.8% of the total image area during post monsoonperiod for 28 October 1992, 15 October 1993, 15 November1994, 02 November 1995 respectively.
Using the salinity indices, the geo-referenced data of electricalconductivity of soil (ECe) and sodium absorption ratio (SAR) ob-tained from the ground truthing was assigned to the respectivepixels. The analysis was carried out to validate the four salinityindices (S1–S4). The salt-affected soils were assessed and quanti-fied. They were further categorised into three levels of soil salinityaccording to the classification described by Ghassemi et al. (1995)and WAPDA (1997). These categories are: (1) slightly saline soils,(2) moderately saline soils, (3) and strongly saline soils. The objec-tive cut point was to find how severely the continuum of salinityhad turned the agricultural lands into waste land and what wasthe quantum of degradation? The S4 developed from B1 to B3,which showed higher reflectance than that of B2 and B4, was foundto be the best option among the suite of indices. Statistically, the
Fig. 5. Land use classification map of 18 June 1992.
correlation coefficients for all salinity indices (S1–S4) varied from0.64 to 0.82. The predicted ECe and SAR using the salinity index‘‘S4’’ showed the highest correlation of 0.82 and 0.76 with mea-sured values of ECe and SAR respectively. The scattergram of mea-sured and predicted values of ECe and SAR using salinity index ‘‘S4’’are shown in Figs. 6 and 7. The correlation coefficients and thescattergram showed that the assessment of salinity is statisticallysignificant. The detailed results into three levels of soil salinityand the overall percent assessment are shown in Table 4. It isfound that the assessment using S4 was promising and had goodagreement with the results obtained from land use classificationshown in Fig. 8 and Table 5.
5.3. Temporal and spatial patterns of salt-affected soils
The 4 year data was processed into three levels of soil salinityusing the same criterion as described in Section 5.2. The definitionof the salinity classes is based on the visual interpretations com-bined with various levels of electrical conductivity of soil (ECe)(Ghassemi et al., 1995; WAPDA, 1997). Fig. 8 characterizes the
Fig. 6. Scattergram of measured and predicted ECe using salinity index – S4.
Fig. 7. Scattergram of measured and predicted SAR using salinity index – S4.
Table 4Assessment of salt-affected soils using salinity indices.
Salt-affected soils Area in hectares determined by salinityindices
Fig. 8. Temporal pattern of saline soils from 1992 to 1995.
Fig. 9. Probability density of salt-affected soils.
A. Abbas et al. / Physics and Chemistry of the Earth 55-57 (2013) 43–52 49
severity of the salinisation and also represents the temporalchange pattern of the surface salinity in the study area. The slightlysaline soils were mainly occurring type of saline soils during the4 years (Abbas and Khan, 2007). During the study period, thevariable patterns in the temporal behaviour of the salt-affectedsoils are noticed and are shown in Fig. 8 and Table 5. The slightlysaline soils increased from 18,000 ha to 25,822 ha during 1992and 1993. There was a major positive change that occurred during1994 and 1995 when slightly saline showed a decreasing trend. Itwas not only in slightly saline soils but overall improvement wasnoticed in soil salinity. However, part of slightly saline soils againreversed to moderately saline soils. The results were consistentwith other studies as reported by WAPDA (1997).
The positive and negative trends in the salinity occurrencecould be traced back to the operation of the reclamation program,which is strongly associated with local and regional impediments.The installation of the tile drainage system and sumps under theFourth Drainage Project (the SCARP) was completed by June1994. The sumps were operational even before the overall comple-tion of the project and obviously improvements were expected inthe area. However, the increase in the saline area during 1993and reversal of the degradation showed the poor performance ofthe project. The full-scale benefits from the project were notachieved and could be linked with the following possible reasons:firstly, a number of sumps (tube wells) had not been working dueto inadequate operation and maintenance of the overall drainagesystem; secondly, reduced canal water supplies and resulting re-duced water allocations of fresh water to the area that helps saltleaching; thirdly, reduced rainfall due to drought; and finally, man-ifold increase in cropping intensity and the lack of economic incen-tives to mitigate salinity (Khan and Hanjra, 2008). The availablecanal supplies hardly meet about 50% of the water demand againstthe total crop water requirements in the area. Due to this shortfall,farmers irrigate their crops with poor quality drainage water (orrecycled water) directly pumped out from the sumps (water qual-ity in ECw units is 2.5–3.2 dS/m; Kelleners and Chaudhry, 1998).Obviously, this poor quality water has higher salinity level thanthat of the canal supply water (water quality in ECw units is0.23–0.31 dS/m; Aslam and Prathapar, 2006) from irrigation ca-nals. Gradually, salts have been further added to the root zone.
Table 5Temporal extent of salinity and sodicity of the salt-affected soils.
Although this study showed comparative reduction in salt-affectedsoils over the period of 4 years (1992–1995), the insignificantreduction meant that these minor gains might be reversed in fu-ture (Abbas, 1999).
The satellite images were analysed to characterize the interac-tive behaviour of salinity and sodicity and their combinations.The known types are: (1) saline-nonsodic soils, (2) saline-sodicsoils, and (3) nonsaline-sodic soils. These types of salt-affectedsoils were assessed according to the benchmark defined by USDA(1954) and WAPDA (1997). The temporal analysis was performedby an overlay procedure through pixel to pixel comparison of thechanges that occurred between consecutive years. The analysisshows that the saline-sodic soils prevailed dominantly. Table 5shows the temporal extent of salinity and sodicity in the region.It is found that saline-sodic soils substantially contributed to theland degradation and was the main issue to the agricultural pro-ductivity. The saline-sodic soils increased by 4.2% from 1992 to1993. Afterward, it had decreasing trend of 10.4% with an overallimprovement in all categories of salt-affected soils.
Fig. 9 shows the probability density of salt-affected soils withintensity ranging from zero to unity. The normal soils were
50 A. Abbas et al. / Physics and Chemistry of the Earth 55-57 (2013) 43–52
represented by zero. The salt-affected soils showed their occur-rence probability ranging from more than zero to unity. Based onthis analysis, the probability occurrence of salt-affected soils was7.5% and 3.8% of the total image area in S-I and S-II regions, respec-tively. The image showing the probability density of the salt-affected soils is not a temporal behaviour of the salinity occurrencebut a perspective of the salinity distribution in the spatial extent.The spatial variability based on the probability occurrence ofsalt-affected soils in the study area is shown in Fig. 10.
5.4. Responses of salt-affected soils to watertable fluctuations
Further analysis to groundwater depths and the occurrence pat-tern of salt-affected soils showed that the shallow watertable(62 m depth from the soil surface) with hazardous quality mainlycontributed to the appearance of salt-affected soils. GIS analysisshowed that 28.4% and 54.4% of the study area (S-I and S-II respec-tively) had shallow watertable. The shallow watertable was foundwith varying water quality standards like good, marginal andhazardous. The complementary analysis of groundwater quality re-vealed that 97.2% of S-I and 89.7% of S-II had ‘hazardous’ ground-water quality. In this extent, 28% of the total area (both S-I andS-II) had shallow watertable. This demonstrated the connectivityof shallow table, groundwater quality and the extent of salinisa-tion. The spatial analysis in relation with hazardous water qualityand shallow watertable showed that 60–70% of the salt-affectedsoils during 1992–1995 had their occurrence in the areas of shal-low watertable having hazardous water quality. This explainedthe fact with a confidence level of, on average, 65% that the study
Fig. 10. Spatial variability of salt-affected soils.
area had been at the risk of salinisation due to waterlogged condi-tions with hazardous rating of water quality. Salinity continued topresent a threat to the sustainability of irrigated agriculture be-cause of using large quantities of poor quality water of underlyingaquifers to supplement irrigation needs. The cyclic closure of irri-gation canals forced the irrigators to use drainage water. This givesthe key message that the relationship of waterlogging, groundwa-ter quality and the ensuing incident of salinisation demands fordeveloping and implementing local land and water managementplans (LWMPs) with strong stakeholders’ engagement to minimisethe risk of further salinisation.
6. Conclusions
Poor management of canal irrigation not only results in waterwastage, but also in the degradation of soils. Soil salinisation inPakistan and elsewhere has caused an unrepairable loss of agricul-tural productivity and compromised the regional economies. Thegroundwater contains varying amounts of salts and is contributingsignificantly to salt built-up in the surface soils. It is vital to estab-lish fast monitoring systems that facilitate actions to combat salin-ity (Abbas and Khan, 2007). The present study demonstrates thepotential of using satellite data for characterizing soil salinityand environmentally degraded lands in the upper Indus Basin ofPakistan. Satellite data showed that salt-affected soils reflect moreincident energy in comparison with normal soils and other landuse classes in the visible and near-infra red spectrum. The imageswere classified using supervised maximum likelihood classificationwith an overall accuracy of about 90%. The accuracy measures builton overall accuracy and the reliability for assessing the quality ofmaps made from the supervised classification of the remotely-sensed data. The remotely-sensed data based salinity indices weredeveloped and their application was evaluated. The occurrencepattern of salt-affected soils in 1992 using the salinity indices ofS1–S4 was found to be 21%, 17%, 16%, and 19% of the total imagearea, respectively. The interactive behaviour of salt-affected soilsshows that the saline-sodic soils largely prevailed than otherknown categories in the region. The use of poor quality groundwa-ter, seepage losses from irrigation supply system, poor resourcemanagement and non-participation of farmers in the reclamationefforts are major impeding factors in the slow positive responsesof the rehabilitation efforts. There is a need to measure the bio-physical parameters of the salt-afflicted crops to avoid the con-straints that cause changes in the spectral reflectance of thesurface features. There is lack of statistical evidence for the tempo-ral and spatial trends in the salt-affected soils. More research is re-quired to establish the reclamation impacts and, in particular, thereasons why improvements have not been achieved in line withthe targets and how stakeholder engagement can make a real dif-ference to salinity management programs.
7. Abbreviations and acronyms
ANN
artificial neural network DN digital number ECe electrical conductivity of soil extract ECw electrical conductivity of water FAO Food and Agriculture Organisation FCC false-colour composite GCPs ground control points GIS geographic information system GOP Government of Pakistan IRS Indian Remote Sensing
A. Abbas et al. / Physics and Chemistry of the Earth 55-57 (2013) 43–52 51
K–T
Kauth–Thomas transformation LAI leaf area index LIDAR light detection and ranging LISS Linear Imaging Self Scanning LWMP land and water management plan mha million hectares MLC maximum likelihood classification ND normalised difference NDVI normalise differential vegetation index NSW New South Wales PLSR partial least square regression RADAR radio detection and ranging RMS root mean square SAR sodium absorption ratio SCARP Salinity Control and Reclamation Project SD standard deviation S-I schedule I S-II schedule II SPOT Satellite Pour l’Observation de la Terre TM thematic mapper USDA US Department of Agriculture UTM Universal Transverse Mercator WAPDA Water and Power Development Authority
Acknowledgements
The first author wishes to acknowledge the funding support ofthe Japan International Centre for Agricultural Sciences (JIRCAS)and the technical support from the staff of JIRCAS which made thiswork possible. We also acknowledge input from the CSIROresearchers. The valuable comments and suggestions by the anon-ymous reviewers are gratefully acknowledged, that helped im-prove the quality of work.
References
Abbas, A., 1999. Analysis of Environmental Land Degradation in the Semi-aridRegion of Faisalabad, Pakistan using Remote Sensing and GIS. JIRCAS ResearchReport, Japan (unpublished).
Abbas, A., Khan, S., 2007. Using remote sensing technology for appraisal of irrigatedsoil salinity. In: Oxley, L., Kulasiri, D. (Eds.), MODSIM 2007 InternationalCongress on Modelling and Simulation. Modelling and Simulation Society ofAustralia and New Zealand, pp. 2632–2638.
Abbas, A., Uchida, S., Yamamoto, Y., 2005. GIS applications for hydrologic risks ofgroundwater levels in irrigated lands, Pakistan. Journal of Water Resources 9(2), 1–5.
Ambast, S.K., Singh, O.P., Tyagi, N.K., Menenti, M., Roerink, G.J., Bastiaanssen,W.G.M., 1999. Appraisal of irrigation system performance in saline irrigatedcommand using SRS and GIS. In: Nieuwenhuis, G.J.A., Vaughan, R.A., Molenaar,M. (Eds.), Operational Remote Sensing for Sustainable Development. Balkema,Rotterdam, pp. 457–462.
Aslam, M., Prathapar, S.A., 2006. Strategies to Mitigate Secondary Salinisation in theIndus Basin of Pakistan: A Selective Review. Research Report No. 97,International Water Management Institute (IWMI), Colombo, Sri Lanka.
Barrett-Lennard, E.G., Hollington, P.A., 2006. Development of a National Program onSaline Agriculture for Pakistan. <http://www.cazs.bangor.ac.uk/salinity/reports/Nationalprogram.htm>.
Bastiaanssen, W.G.M., Molden, D.J., Makin, I.W., 2000. Remote sensing for irrigatedagriculture: examples from research and possible applications. AgriculturalWater Management 46, 137–155.
Chowdary, V.M., Chandran, R.V., Neeti, N., Bothale, R.V., Srivastava, Y.K., Ingle, P.,Ramakrishnan, D., Dutta, D., Jeyaram, A., Sharma, J.R., Singh, R., 2008.Assessment of surface and sub-surface waterlogged areas in irrigationcommand areas of Bihar state using remote sensing and GIS. AgriculturalWater Management 95 (7), 754–766.
de Clercq, W.P., van Meirvenne, M., Fey, M.V., 2009. Prediction of the soil-depthsalinity-trend in a vineyard after sustained irrigation with saline water.Agricultural Water Management 96 (3), 395–404.
Douaoui, A.E.K., Nicolas, H., Walter, C., 2006. Detecting salinity hazards within asemiarid context by means of combining soil and remote-sensing data.Geoderma 134 (1–2), 217–230.
Dwivedi, R.S., Rao, B.R.M., 1992. The selection of the best possible Landsat TM bandcombination for delineating salt-affected soils. International Journal of RemoteSensing 13 (11), 2051–2058.
Dwivedi, R.S., Sreenivas, K., 1998. Delineation of salt-affected soils and waterloggedareas in the Indo-Gangetic plains using IRS-1C LISS-III data. InternationalJournal of Remote Sensing 19 (14), 2739–2751.
FAO 2002. Pakistan – Development of a Research Programme in Irrigation andDrainage, International Programme for Technology and Research in Irrigationand Drainage. Programme Formulation Report No. 9. <http://www.fao.org/docrep/005/y3690e/y3690e0a.htm>.
FAO, 2007. Extent and Causes of Salt-affected Soils in Participating Countries. AGL:Global Network on Integrated Soil Management for Sustainable use of Salt-affected Soils. <http://www.fao.org/ag/agl/agll/spush/topic2.htm>.
Farifteh, J., Farshad, A., George, R.J., 2006. Assessing salt-affected soils using remotesensing, solute modelling, and geophysics. Geoderma 130 (3–4), 191–206.
Farifteh, J., Van der Meer, F., Atzberger, C., Carranza, E.J.M., 2007. Quantitativeanalysis of salt-affected soil reflectance spectra: a comparison of two adaptivemethods (PLSR and ANN). Remote Sensing of Environment 110, 59–78.
Farooq, M., Din, N.U., 1980. Application of multi-temporal Landsat data in theidentification of salinity in the Khairpur pilot project, Pakistan. In: Proceedingsof the National Seminar on Application of Remote Sensing Techniques in WaterResources Development and Management, Lahore, Pakistan, pp. 74–31.
Fernández-Buces, N., Siebe, C., Cram, S., Palacio, J.L., 2006. Mapping soil salinityusing a combined spectral response index for bare soil and vegetation: a casestudy in the former lake Texcoco, Mexico. Journal of Arid Environments 65 (4),644–667.
Ghassemi, F., Jakeman, A.J., Nix, H.A., 1995. Salinisation of Land and WaterResources: Human Causes, Management and Case Studies. Centre forResources and Environmental Studies, ANU, Canberra.
Government of Pakistan [GOP], 1999. Soil Survey of Pakistan, Islamabad, Pakistan.Hanjra, M.A., Qureshi, M.E., 2010. Global water crisis and future food security in an
era of climate change. Food Policy. doi:10.1016/j.foodpol.2010.1005.1006.Kelleners, T.J., Chaudhry, M.R., 1998. Drainage water salinity of tubewells and pipe
drains: a case study from Pakistan. Agricultural Water Management 37 (1), 41–53.
Khan, S., Hanjra, M.A., 2008. Sustainable land and water management policies andpractices: a pathway to environmental sustainability in large irrigation systems.Land Degradation and Development 19 (3), 469–487.
Khan, S., Hanjra, M.A., 2009. Footprints of water and energy inputs in foodproduction – global perspectives. Food Policy 34 (2), 130–140.
Khan, G.M., Siddiqui, Z.A., 1987. Study of Waterlogging and Salinity in Larkana,Khairpur and Gujranwala using Remotely Sensed Data. Remote SensingApplications Division, SUPARCO, Karachi, Pakistan.
Khan, N.M., Rastoskuev, V.V., Sato, Y., Shiozawa, S., 2005. Assessment of hydrosalineland degradation by using a simple approach of remote sensing indicators.Agricultural Water Management 77 (1–3), 96–109.
Kijne, J.W., 2006. Salinisation in irrigated agriculture in Pakistan: mistakenpredictions. Water Policy 8 (4), 325–338.
Mahmood, K., Marcar, N.E., Naqvi, M.H., Arnold, R.J., Crawford, D.F., Iqbal, S., Aken,K.M., 2003. Genetic variation in Eucalyptus camaldulensis Dehnh, for growth andstem straightness in a provenance – family trial on saltland in Pakistan. ForestEcology and Management 176 (1–3), 405–416.
Malins, D., Metternicht, G., 2006. Assessing the spatial extent of dryland salinitythrough fuzzy modeling. Ecological Modelling 193 (3–4), 387–411.
Metternichet, G.I., Zinck, J.A., 1997. Spatial discrimination of salt- and sodium-affected soils surfaces. International Journal of Remote Sensing 18 (12), 2571–2586.
Metternichet, G.I., Zinck, J.A., 2003. Remote sensing of soil salinity: potential andconstraints. Remote Sensing of Environment 85, 1–20.
Metternicht, G., 2001. Assessing temporal and spatial changes of salinity usingfuzzy logic, remote sensing and GIS. Foundations of an expert system. EcologicalModelling 144 (2–3), 163–179.
Odeh, I.O.A., Onus, A., 2008. Spatial analysis of soil salinity and soil structuralstability in a semiarid region of New South Wales Australia. EnvironmentalManagement 42, 265–278. doi:10.1007/s00267-008-9100-.
Peng, W., 1998. Synthetic analysis for extracting information on soil salinity usingremote sensing and GIS: a case study of Yanggao basin in China. EnvironmentalManagement 22 (1), 153–159.
Qureshi, R.H., Aslam, M., Rafiq, R., 1993. Expansion in the use of forage halophytes inPakistan. In: Davidson, N., Galloway, R. (Eds.), Proceedings WorkshopProductive Use of Saline Land, Perth, Australia, 10–14 May, 1991. ACIARReport 42, pp. 12–16.
Silvestri, S., Marani, M., Marani, A., 2003. Hyperspectral remote sensing of saltmarsh vegetation, morphology and soil topography. Physics and Chemistry ofthe Earth, Parts A/B/C Applications of Quantitative Remote Sensing toHydrology 28 (1–3), 15–25.
Stockle, C.O., 2007. Environmental Impact of Irrigation: A Review. <http://www.swwrc.wsu.edu/newsletter/fall2001/IrrImpact2.pdf> (accessed 01.08.07).
Tabet, D., 1995. Le suivi de la salinité dans les périmètres irrigués par télédétectionet systèmes d’information géographique: application au périmèter de Fordwah,Pakistan. Montpellier, France, DEA National d’Hydrologie. 174pp.
Tabet, D., 1999. Intérêt d’une approche spatiale pour suivi de la salinité des solsdans les systèmes irrigués: cas de la subdivision de Chistian dans le Punjab,
52 A. Abbas et al. / Physics and Chemistry of the Earth 55-57 (2013) 43–52
Pakistan. Ph.D. Thesis. Montpellier, France, Ecole Nationale du Génie rural, desEaux et Forêts.
Thenkabail, P.S., Hanjra, M.A., Dheeravath, V., Gumma, M.A., 2010. A holistic view ofglobal croplands and their water use for ensuring global food security in the21st century through advanced remote sensing and non-remote sensingapproaches. Remote Sensing 2 (1), 211–261.
Tripathi, N.K., Rai, B.K., Dwivedi, P., 1997. Spatial modelling of soil alkalinity inGIS environment using IRS data. In: Proceedings 18th Asian Conference onRemote Sensing. 20–24 October, 1997, Kuala Lumpur, Malaysia, pp. A-8-1–A-8-6.
USDA, 1954. The US Salinity Laboratory Handbook No. 60. The United StatesDepartment of Agriculture.
van Lynden, G.W.J., Mantel, S., 2001. The role of GIS and remote sensing in landdegradation assessment and conservation mapping: some user experiences andexpectations. International Journal of Applied Earth Observation andGeoinformation 3 (1), 61–68.
Verma, K.S., Saxena, R.K., Barthwal, A.K., Deshmukh, S.N., 1994. Remote-sensingtechnique for mapping salt affected soils. International Journal of RemoteSensing 15 (9), 1901–1914.
Vidal, V., Tabet, D., Ahmad, M.D., Asif, S., Zimmer, D., Strosser, P., 1998. SalinityAssessment in Irrigation Systems using Remote Sensing and GeographicalInformation Systems – Application to Chistian Subdivision, Pakistan. IIMIReport No. R-47.
Wang, F., Xu, Y.J., 2008. Development and application of a remote sensing-basedsalinity prediction model for a large estuarine lake in the US Gulf of Mexicocoast. Journal of Hydrology 360 (1–4), 184–194.
WAPDA, 1981. Atlas Soil Salinity Survey of Irrigated Areas in the Indus Basin, 41million acres. Publication. No. 274, 5 Volumes. Directorate of Reclamation andInvestigation Organization, Lahore, Pakistan.
WAPDA, 1984. Applicability of Landsat Imagery for Monitoring Soil Salinity Trends.Publication No. 529. Directorate of Reclamation and Investigation OrganizationLahore, Pakistan.
WAPDA, 1997. Monitoring and Evaluation, Fourth Drainage Project Publication No.SM-267. Annual Report vols. I and II. SCARP Monitoring Organisation (SMO),Lahore, Pakistan.
Yu, R., Liu, T., Xu, Y., Zhu, C., Zhang, Q., Qu, Z., Liu, X., Li, C., 2010. Analysis ofsalinization dynamics by remote sensing in Hetao Irrigation District of NorthChina. Agricultural Water Management 97 (12), 1952–1960.
