|
|
|
Land Degradation
|
Mapping of Salt-Affected Soils using Remote Sensing and Geograp
Information Systems: A Case Study of Nakhon Ratchaslma, Thailandl
3.2 Analysis of Satellite Data
Various steps involved in the analysis of satellite data are presented in Fig. I. In brief, after geometric correction and linear contrast stretch, the image was visually interpreted in individual bands and various band combinations, and by applying filtration, Principal Component Analysis (PCA) and band ratioing. Supervised classification was employed with maximum likelihood, mahalanobis and minimum distance methods including the bands 1,3,4,5 and 7. Band six had poorer separation of features and band two was highly correlated with one. Unsupervised classification was also tried with sequential clustering including seven separable classes. .
Fig.1 Flow Chart Showing Various Steps Involved in Image Classification
3.3 Development of GIS Thematic Layers and Their Integration
3.3.1 Selection of Parameters
The ideology behind the selection and further classification of parameters (Table 2) were as follows
- Landform:
Among the five types of landform present in study area, virtually no salt crust is seen in high and middle race while it can be frequently seen in flood plain and basin. Lower terrace and peneplain remain somewhere in the middle.
- Ground Water Quality and Well Yield:
Along with the quality ofeds to
be considered for salinity assessment. Hence for overlaying purpose, the twelve classes of ground water quality and quantIty so found m study area were grouped to four by considering the expected salt quantity (in term of kg/hr) that can come out under average condition along with same operation hour. The criteria set was <1.25, 1.25 to 6.00,6.01 to 12.00 and > 12.00 kg/hr for Good (WI), Medium (W2), Worse (W3), Worst (W4) , respectively (Table I).
- Soil Drainage:
Generally, poorer the soil drainage, greater the possibility of rise in the level of perched ground water especially in low- Iying areas.
- Irrigated Area:
Regular water supply to an area generally enhances the salinization by leaving the salts on soil surface and/or increasing the ground water table. This becomes more concerning factor if it contains more salt and/or ground water quality is worse.
- Slope:
Generally, slope percent two or less has been reported as associated with the salinization and again one or less
percent as the most serious one (Ratanavong, 1991 ).
3.3.2 Creation and Integration of GIS Layers
The digitized contour lines were further processed to create Digital Elevation Model (DEM) fIrst which was used to derive slope GIS layer. Similarly, to create landform and drainage GIS layer, the digitized soil coverage was assigned with respective value. Rest of GIS files were obtained just by digitization (Fig. 2).!Integration of thematic layers was done using logical approach as employed by (Mongkolsawat and Thiragnoon, 1990).This involved the identification of saline areas where a specified combination of conditions occurred (Table 3). For the first two classes, main priority was given to landform type while for the rest classes, ground water quality and well yield was given highest consideration followed by landform and rock type. Area under , being considered as not affected by salinization themselves, were classified as Gland G2 except the irrigated one. These were further separated by assigning Rl with Wl as Gl and rest as G2. Similarly, to separate G4 and G5, further base was taken from the slope and drainage conditions. Alluvial complex with all area as R2 and IRl and majority as W3 and SL 1, was classified as G4.
Fig. 2 Flow Chart for Creating and Integrating GIS Coverages and with Image
3.3.3 Integration of Classified GIS Coverage with Image
In brief, G 1 and G2 classes having L 1 type of landform were assigned the same class name. Rest of extremely as well as moderately saline areas so classified from image were put as such while that of slight saline area was further categorized into potential (S4) and slightly saline area (S3) (Table 5). The interpretation of salinity result with respect to crop f~ production was based on the salt tolerance guidelines given by Mass (1984).
4.0 Results and Discussions
4.1 Image Analysis ~
-
Visual Interpretation:
In contrast to others, band six showed darker tone for the area having >50% salt crust coverage which may be due to the discontinuity of pores/channels in soil that resulted into more moisture. Among the Various band combinations, 5,3,1 (R, G, B) had clearest distinction of salt crust area followed by standard FCC (4,3,2) and 7,5,3. The other band combinations had similar color appearance to one of above combinations. For instance, the combinations 5,3,2; 7,3,2; 7,3,1 and 5,2,1 were similar to 5,3,1. Filtering, band rationing and PCA were found no more helpful in this respect than as in above band combinations.
- Unsupervised Classification:
Majority of the area (84.86% ) so classified as nearly barren and vegetative, from this classification, were not interpretative from salinity point of view (Table 4).
- Supervised Classification:
Four classes viz; Extremely saline (11), Moderately saline (12), Slightly saline (S3) and Deep water bodies could be separated from this classification. Various methods had similar coverage of each class, especially maximum likelihood and mahalanobis. However, the spatial distribution of the classes was different in these two methods, too. Also, the accuracy assessment showed that maximum likelihood with standard deviation three had the highest accuracy (Fig. 3). The result was more accurate at standard deviation three than two. It can be attributed to well separablity of different features even at three. The result of maximum likelihood at standard deviation three was used for further analysis.
Fig.3 Accuracy of Various Methods of Supervised Classification
|
|
|
|
|
|
|
