The lower part of Figure 2 illustrates the observation of distributional characteristics of several kinds of indices derived from LANDSAT TM data. Soil Moisture Index (SMI) and Normalized Vegetation Index (NVI) are both calculated from from band 2 and 4 data. For the calculation of NVI, the minimum value appearing in the histogram would be subtracted from the data of each band to correct simply rediometric bias associated with component of principal component analysis for the two dimensional data (band 3 band 4 ) provided that the NVI value ranges between o and o.2 because SMI is not effective in areas densely covered with vegetation . Coverage Index (PDRG) is a modification from the index PD54 developed by Pickup et al. (1993) in which the original concept was applied to band 4 and band 5 of LANDSAT MSS data Another information is a subtracted value of normalized SPOT data . Although the observation time varies with the years, seasonal allocation which is presented on the opposite side of the peak of vegetation activity in the Rabi season may enable to distinguish some interesting features of the regional engironment.
5. Results and Discussion
1) Analysis of vegetation changes using GVI data
By overlaying six thematic maps tow land units dominated the area of the Pothwar plateau. The conditions between the two units differed only in the pattern of precipitation, with more than 250 millimeters in both Rabi and Kharif seasons for one and more than 250 millimeters in Kharif but less than 250 millimeters in Rabi season for the other.
Figure 3. Change of GVI in the Pothwar plateau (low rainfall)
Figure 3 shows the changes of GVI in the land unit with relatively less precipition and Figure 4 shown the changes for the unit with higher precipitation . It was indicated that the condition could be fulfilled only if the GVI data in a temporal and spatial range were representative (Goward et al. (1993)). Procedure for spatial averaging in this study could mitigate the sause of error in the analysis , so that these figures indicate two distination maxima corresponding to Rabi and Kahrif crop cultivation and extremely low values between maxima where the climate is hot and dry. GVI values at maxima are evidently higher in the unit with higher precipitation and the presence of a peak is unstable in the Kharif season in both units . Especially no distinctive Kharif peak was noticed in 1987 owing to the shortage of precipitation .
Figure 4. Change of GVI in the Pothwar plateau (high rainfall)
Similar changes in the pattern of GVI, that is the presence of two maxiam, were observed in the irrigation agricultural areas. On the other hand no depression of change in the pattern in summer season was observed the wood and forest areas and the same in rangland These findings support that it is possible to distinguiest agricultural lands from other land uses by analyzing the temporal changes of GVI data. Figure 5 shows an attempt to differentiate agricultural lands from non-agricultural lands, where the horizontal axis denotes themaximum value in February to April and August to October and the vertical exis denotes the minimum value in May to July. A small number of exceptions can be seen for degraded land where vegetation after rainfall grow under susally and conditons. The figure shows that the largest number dosts representing non-agricultural lands would be allocated along a line according to their averaged vegetation activities and those of agricultural lands would be distant from the lone.
2). Extraction of severely eroded area using SPOT data
Soil erosion in the Pothwar lateau is characterized by the development of dissected valleys since the widely distributed locess soil is highly erodibel by water in the vertical direction . These valleys appear ;on the satellite imagery with high frequency features when the spatial resolution of the imagery corresponds to the scale of topographic variations . Figure 6 compares two satellite imagerics of the area including a severely eroded area . SPOT panchromatic mode with 10 meter spatial resolution shows clearly the dissected patterns by soil erossion located in the upper left and upper right sides of the imagery. The portions occurring with a high frequency but relatively regular pattern represented on the lower side probably correspond to arable land. Comparison of the SPOT imagery with LANDSAT TM (band2) shows that the former is evidently superior to the latter for delineating eroded areas.Land features occurring with a high frequency can be expressed by applying statistical data analysis. Auto-correlation function is one of the methods which represent spectral characteristics of objects . Pickup and Chewings (1988) applied this technique to the temporal change of land degradation induced by soil erosion. In this study auto-correlation functions have been differences the high path filtering with 3X3 window images Figure 7 shows the results of calculation of the whole target area and Figure 8
Figure 5. Scattergram of seasonal GVI by land unit
