Remote sensing is generally accepted as a powerful technique for land surface studies in arid and semi-arid regions. This technique is also very useful for other environmental studies such as shallow and undeveloped soil properties. In this research, band relations and principal component analysis were utilized to develop an appropriate model for vegetation cover monitoring and also finding a relation between vegetation cover and soil properties according to Digital Number of pixels in the LandSat image. The relations between spectral values of the pixels with the parameters were analyzed in order to develop a statistical model. The model was also assessed by a regression analysis and comparing of R-Square coefficients for each of the variables. Two different series of corresponding digital numbers for vegetation cover were distinguished and categorized into 40 and 70 samples. Different indices and band combinations were used for a regression stepwise analysis. The results showed that bands 1, 2 and 3 with R-Square 0.26, 0.33 and 0.38, respectively showed higher correlation than other spectrum bands. Other indices such as brightness analysis, greenness component, tasseled cap and PCA showed higher correlations. The highest R-square statistically significant at 95% confidence level was derived when vegetation cover data and DN in different band compositions were considered in the regression analysis. Radiation reflection varies with soil properties and other soil factors depending on vegetation cover density; therefore, estimation of vegetation cover based on remote sensing data is very complicated. However, this study showed that ETM+ images are very useful for monitoring vegetation spatial variability especially in remote arid and semi-arid areas.

Keywords