Proceedings of The Fourth International Iran & Russia Conference

442

Monitoring of soil salinity by Remote sensing & GIS in kashan area M. Abtahi1 , M.Pakparvar2 1-Desert research station of kashan, P.O.Box 487, Kashan,Iran ,phone: 0361-4234955,4234498,Fax:4234999, Email:[email protected] 1Fars research center for natural resources, P.O.Box 71555-617, shiraz, IRAN

Abstract A study was conducted to determine the capabilities of the successive numerical Landsat data for assessment and monitoring of soil salinization. Kashan plain, with 7220 Km2 of area, located in an arid zone of the central part of Iran, was selected as the site of investigation. It seemed to be a region prone to desertification processes. Two sorts of Landsat data: MSS (1976), TM (1998) supplementary information from soil and geo maps, surface and subsurface water data were collected.After preprocessing, the images were classified on the basis of field and subsidiary data for soil salinity. For MSS data, the Pca12, Pca34 and NDVI were merged and showed the best correlation with field samples. In TM data, merging the TM4, Pca57, Pca123 and NDVI showed the best correlation. The classification was performed by the maximum likelihood algorithm.Verification of the results showed that the differentiation between salinity classes has had a meaningful precision for both salinity maps.

A GIS network was constructed. After producing a DEM layer of the region, the other layers such as the oldest and newer maps of isodept curves of groundwater and also the salinity of groundwater maps, two classified soil salinity maps, geomap, were introduced to the GIS network as well as their documents.Merging and processing the whole data showed that 7.5% of non saline parts of region, had changed from medium to high saline, and in the same time, the size of Kashan salt lake has decreased 0.1% of total. Most of the salinized area is located in area with more salinity and reduced depth of groundwater compared with their past. Deep underlying geological material of these areas is mainly the Miocene saline evaporate deposits which is recognized as the main factor for increasing the salinity of groundwater and consequently the soil surface.

Key words: Remote sensing, Salinity, Desertification, PCA, image processing, GIS.

Proceedings of The Fourth International Iran & Russia Conference

443

Introduction Salt affected soils cover about 7 percents of the global area (szabolcs 1981). In Iran an estimated of 23.5 Mha (14.2% of total area) is affected by salinity problems (Vanaart etal 1968) and 50% of irrigated lands are salinized or prone to salinization (Pazira 1999). With this high level of salinity there is a need to monitor the problem regularly in order to take up timely reclamative and preventive measures. Remote sensing by virtue of its repetitive coverage of large areas offers great potential for monitoring dynamics, including salinization /alkalinization (Myers etal 1980, Kharin 1982, Hellden 1985, Tucker and Justice 1986, Vinogrdov 1993,). Diwivedi and Rao (1992) showed that a band combination of 1,2 and 3 bands of TM could separate degrees of salinity . A statistic index, OIF was used for band selecting. Jushi and sahai (1993), classified the salinity as severe, moderate, and slight using TM5 and MSS2 by a precision of 90 and 74 percent respectively. Damavandi (1996) showed that the best correlation between the EC of field samples and the spectral bands is obtained by TM3/4. Also a combination of CMP457 (first band from PCA on the bands 4,5,7); CMP234(first band from PCA on the bands 2,3,4) using TM4,TM3; TM6CMP1 (first band from PCA on all bands except the TM6); and TM6Cmp2 (second band from PCA on all bands except the TM6), TM4,TM2 could present the best separation of salinity classes in the regions around the Qom salt lake of Iran. Materials and methods Study site The Kashan plain is bounded by the geo coordinates 33˚,46́ to 33˚,30́ N and 51˚,03́ to 52˚,28́ E is located in the center part of Iran (200 km east of Tehran, in the Esfahan province, Figure 1). Mean annual precipitation(20 years of data) is 136.6 mm and 214 mm in the lowlands and uplands respectively. The whole basin has an area of 11000km2 .The test site is 7220 km2 in area. It is restricted by a chain of mountains to the west and south and a salt lake (Maseeleh) to north. Kashan city (350000 people) and its dependent suburban and rural parts are distributed around the plain. The plain is formed from two sorts of alluvium of Paleocene age, brought down and deposited by eleven ephemeral and perennial rivers. The alluvial fans have been gently changed to a flat evaporate salty and gypsum deposits ending in a salt lake. Because of the common wind direction from SW to NE, a chain of sand dunes, 756 km2 in area, has formed in the middle of the plain. The local lowlands that serve as accretion zones for salts and finer particles, act as evaporating pans and play a crucial role in the formation of salt affected soils.

Data base The remote sensing data base comprised Landsat2 MSS digital data of 2339 by 3264 that was acquired on 25 may 1976 and landsat5 TM of 5000 by 8944 pixels for 18 may 1998. The survey of Iran topographical sheets of 1:50000 and 250000 and published soil survey reports and geomaps as well as the data of level and quality of ground water were also used. A pentium II PC, the IDRISI 2.008, photo shop 5 were the main tools for investigation. Approach Geometric correction Eight ground control points which had been accurately obtained in the field observation by the GPS instrument, were used for geocorrection of the images. TM data obtained from Eosat, had been originally georeferenced by non-parametric systemic correction. The MSS was corrected by a linear polynomial model with the RMSE of 0.92 pixel. Its pixel size was reformatted to 25 X 25 (equal that of the originally resampled TM). Field sampling

Proceedings of The Fourth International Iran & Russia Conference

444

On the basis of the soil maps and a preliminary unsupervised classification, a network of welldistributed sampling areas were developed. Some 84 soil samples was collected from 21 training area. The samples were analyzed for texture components, caco3, gypsum, EC and SAR. Classification Original bands, a variation of spectral ratios, band combinations and Principal Component Analysis (PCA) on different bands previously proposed for separating salinity classes as well as the differentiating between gypsum and saline area were constructed (table 1). The DN values related to pixels of training samples were extracted for each of the bands or bands combinations. According to the correlation between DN values and EC, percent of gypsum and caco3 of soil samples (Table2) the best combination was selected. The Field samples were divided to two categories in a viewpoint of well spatial distribution through the area for each category. Supervised classification was done on the appropriate band combination applying the maximum likelihood algorithm on the basis of the first category.

A variation of EC classification was considered and the best result was obtained using the classification as give in Table 3. Because of the lack of observation for MSS data, a set of training pixels was selected from classified TM image as a basis for MSS supervised classification Precision analysis The second category of field samples was used for checking the precision of the classified TM data. Results showed the total precision of 92%. For the MSS, a set of 23 checking points was axtracted from the field samples (applying the local experiences and different from those had been used as the training samples).Total precision of 72% was obtained. GIS Soil surface salinity classified maps of MSS and TM were overlaid and a new map of salinity changes was constructed. This was merged to the geomap, and DEM layer and the document table of the final algorithm was constructed. Results and discussion Results of image classification are shown in figure 2 and the data relating to change of salinity through the period are presented in table 3. A cross matrix of changes between the classes is shown in table 4. According to table 4, 7.5 % of the non-saline area has been changed to saline (6.1% to class 2 and 1.4% to class 3). In the same time, 1.5 % of the area previously covered by salt lake, has changed to lower classes (0.5% to class 1, 0.1% to class 2 and 0.9% to class 3). All of the changes from non-saline to high salinity (class 1 to 3) has occurred in marginal lands.Most of the change from non saline to low and medium (class 1 to 2) and also medium to high (class 2 to 3) Is shown in central parts especially around the Kashan City and its other sub urban and villages

Proceedings of The Fourth International Iran & Russia Conference

445

Considering the whole overlaid data, it may be synthesized that salinization has accurred in relation to two main mechanisms namely exposure underlying saline layers in some marginal area because of the wind erosion which is sequentially due to over grazing and deterioration of green cover as shown in 3, and salinization of the groundwater, which is the only source of irrigation water. The latter is due to over exploitation water aquifer (more than 6 meters lowering of sub surface water and a increase in the number of wells over the 22 years period) and in the same time, a decrease in recharge from upland catchments.The increase in runoff coefficient is evidence of less water absorption by soils of uplands due to deterioration of plant cover, which in turn results in increasing sediment concentration. Two problems are caused by that phenomenon; a) In marginal area, the saline ground water front that previously had restricted and controlled by hydraulic pressure the sweet water aquifer is now allowed to flow beneath the arable lands. b) In central parts, salinized groundwater has developed on the alluvial aquifer with a very deep underlying evaporate of Miocene age.The decrease in water table (more than 20 meters in some cases) has caused more contact with saline layers and dessolution of minerals. Figure 4 is an example of degraded lands affected by newly salinized irrigation water.

Classification of the images in order to differentiation between saline and gypsum area did not give any reasonable results on the basis of the precision analysis. All of the soil samples having different quantities of gypsum over the zero had EC values above 10 ds/m. so, our problem was separation of gypsum saline area from the saline area and there was a lake of DN differentiation. However there was not a great error for detecting the salinity changes, because there was no change in gypsum area from past to present.

Conclusion The main reason for progressive salinization in the Kashan plain as a typical example of central deserts area of Iran is disturbance of hydrological cycle and mismanagement of water which reflected on soil and water salinity. In addition, plant cover degradation may be the cause of salinization of the marginal lands. PCA of blue, green and red could increase the seperability of salinity classes especially when combine with the PCA of the near IR (5 and 7)bands. Incorporation of NDVI in PCA combination, could increase the correlation of DN values and soil EC; it seems to be a way to minimize the effect of plant cover absorbency when soil salinity is the main point of notice.

Proceedings of The Fourth International Iran & Russia Conference

446

There is a need for better recognition of the interaction effects of saline gypsum and carbonate minerals in spectral reflectance as well as in selecting the best band or PCA combination for their differentiation. References 1- Alavi Panah, S. K, (1998). Study of soil salinity in desert based upon field observation, remote sensing and a GIS (case study: Ardakan area, Iran). Unpublished paper presented at the INT Symposium of New Technologies to Combat Desertification, held in Tehran, Iran, 12-15 October 1998. 2- Dapper, Goossens, (1996). Modeling and Monitoring of soil salinity and water logging hazards in the desert - delta fringes of Egypt based on geomorphology, remote sensing and GIS. Proceedings of the 16th Earsel symposium Malta, 20-23 MAY. 3- Dwivedi, R.S, Rao, BRM, (1992). The selection of the best possible landsat TM and combination for delineating salt-affected soils. INT. J. Remote sensing. vol. 13: No. 11, 2051-2058 4- Hellden, V, (1985). Remote sensing for drought impact assessment-a study of land transformation in Kordofan, Sudan. Advances-in-Space-Research, 4: 11, 165-168 5- Joshi, M.D, Sahai, B, (1993). Mapping of salt affected land in Soura Shtra coast using landsatsattelite data. INT.J. Remote sensing, vol. 14: No. 10, 1919-1929. 6- Jurio, Elsie-M, Zuidam,(1998). Remote sensing, synergism and information system for desertification analysis: an example from northwest geographical patagonia. Argentina. ITC Journal 3/4. 7- Kharin, N. G, (1982). Remote sensing and monitoring of desertification in arid lands. Alternative strategies for desert development and management, vol. 4 (UN Institute for Training and Research) 1295-1309, New York, USA, Pergamon press. 8- Kaushalya, Ramachandran, (1992). Monitoring the impact of desertification in Western Rajasthan using remote sensing. Journal of Arid Environments 22: 293-304. 9- Lyon, Johng, Yuan, Ding, ET al, (1998). A change detection experiment using vegetation 10- indices. PE and RS. February. 11- Mishra, J.K, Joshi, M.D. (1994). Study of desertification Process in Aravialli using remote sensing techniques. INT.J. Remote sensing, vol. 1: No. 1, 87-94. 12- Metternicht, Graciela, Zinck, Alfred, (1996). Modeling salinity alkalinity classes for mapping salt-affected top soils in the semi arid valleys of Cochabamba (Bolivia). ITC Journal, vol. 2, 125-134 13- Myers, V. I, Mann, H.S, Moore, D, Derries, M, Abdel-hady, M, (1980). Remote sensing for monitoring resources for development and conservation of desert and semi-desert areas. Mann, H.S (Ed): Arid zone research and development, 505-513 14- Tucker, C. J, Justice, C. O, (1986). Satellite remote sensing of desert spatial extent. Desertification Control Bulletin, no. 13: 2-5 15- Rao, BRM, Dwivedi, RS, (1998). An inventory of salt-affected soils and water logged areas in thenagar Jun Sagar canal command area of southern India, using multispectral space-Borne data. Land degrad. develop. 9: 357-367. 16- Vinogradov, B. V, (1993). Remote indicators of soil desertification and degradation. Eurasian-Soil-Science, 25: 8, 66-75

Proceedings of The Fourth International Iran & Russia Conference

447

Fig.1. Location of the Kashan plain

Table 1 – List of combinations which have compared for salinity classification Type of band or combination

Description

Objective

Source

TM 1 to 7 except 3

Combination

Salinity alkalinity

Metternicht and Zinck (1996)

TM 1,3,5

Combination

Salinity classes

Dwivedi and Rao (1992)

TM 3,4,5,6

Combination

Salinity and gypsum

Alvi Panah (1997)

TM 2,3,4

Combination

Salinity alkalinity

Rao and Dwivedi (1998)

TM 3 / 4

Spectral ratio

Correlation against EC

Damavandi (1996)

TM 4 / 3

Spectral ratio

Minimizing the salinity interfere Lyon and Yuan (1998)

TM 1,2,4,5

Combination

Salinity classes

Joshi and Sahai (1993)

TM 5

Original band

Salinity mapping

Joshi and Sahai (1993)

CMP457; CMP234,TM4,TM3; and TM6CMP1;TM6CMP2,TM4,TM2

Combination of PCA and Salinity classes original bands

TM4, PCA57, NDVI

PCA123

and Combination of PCA and Salinity classes original bands

Damavandi (1996) Proposed in this work

MSS 2

Original band

Salinity mapping

Joshi and Sahai (1993)

PCA of MSS

Higher order of PCA

Salinity classes

Dwivedi (1996)

Proceedings of The Fourth International Iran & Russia Conference

PCA12, PCA34 and NDVI

Combination

448

Salinity classes

Proposed in this work

Table 2 – correlation between DN values and soil properties

Soil properties

Type of band or combination

EC ds/m

Gypsum%

Correlation

Significant level

TM 1 to 7 except 3

0.57

TM 1,3,5

0.21

TM 3,4,5,6 TM 2,3,4 TM 3 / 4

%Carbanate

Correlation (a)

Significant level

Correlation (b)

Significant level

0.001

0.36

0.05

0.11

0.05

0.25

NS

NS

0.22

NS

0.31

0.01

0.42

0.01

0.26

NS

0.48

0.001

0.22

NS

0.24

NS

0.51

0.001

0.28

NS

0.35

NS

TM 4 / 3

0.28

0.01

0.25

NS

0.28

0.05

TM 1,2,4,5

0.51

0.001

0.15

NS

0.24

NS

TM 5

0.41

0.001

0.20

NS

0.21

NS

0.65

0.001

0.21

NS

0.35

0.05

TM4, PCA57, PCA123 and NDVI

0.79

0.001

0.22

NS

0.33

0.05

MSS 2

0.39

0.001

0.12

NS

0.26

NS

MSS 3

0.44

0.001

0.13

NS

0.29

NS

PCA12, PCA34 and NDVI

0.66

0.001

0.21

NS

0.○

0.05

CMP457; CMP234,TM4,TM3; TM6CMP1; and TM6CMP2,TM4,TM2

a and b – some samples having zero values for the property were omitted Table 3 – Salinity classes based upon the field samples and the image characteristics

No. of class

Description

Range of EC ds/m

1

Non saline

0–2

2

Low to medium salinity

2 – 10

3

High salinity

> 10

4

Salt flat

-

Table 4. Change in salinity classes through the 22 years period Class

Description

EC ds/m

Size of area (1976)

Size of area (1998)

Ha

%

Ha

%

Deference

Change at the class %

Total change %

1

Non saline

0–2

209133

29

154781

21.4

-54352

-26

-7.5

2

Low to med.

2 – 10

301650

41.8

310437

43

8787

2.9

1.2

3

High

> 10

112917

15.6

169577.8

23.5

56660.8

50.2

7.8

4

Lake

-

98510

13.6

87414.2

12.1

-11095.8

-11.3

-1.5

Proceedings of The Fourth International Iran & Russia Conference

Total

722210

100

722210

449 100

0

-

Figure 2 – Kashan plain classified images of MSS (1976) right and TM (1998) left. Table 5 . Cross matrix of changes between the classes 1976

Non saline

Low to medium

High

Salt lake

Percent of change

Non saline

-

-6.1

-1.9

0.5

-7.5

Low to medium

6.1

-

-5

0.1

1.2

High

1.9

5

-

0.9

7.8

Salt lake

-0.5

-0.1

-0.9

-

-1.5

1998

0

Proceedings of The Fourth International Iran & Russia Conference

Fig.3. Marginal lands with apearance of saline layers

450

Fig.4. An example of degraded lands Underlying which were arable farmlands in the past