Comparison of different reflectance indices for vegetation analysis using Landsat-TM data

Abstract

Several reflectance indices based on digital satellite data are used to describe phytosociological characteristics of vegetation. The present study was aimed at comparing the reflectance indices- simple ratio (SR), difference vegetation index (DVI), normalized difference vegetation index (NDVI), nonlinear index (NLI), aerosol free vegetation index (AFRI) and green chlorophyll index (GCI) for phytosociological characteristics of vegetation in the vicinity of river Beas in India, over a stretch of 63 km from the towns of Beas to Harike in the state of Punjab, India. Factor analysis (FA) was applied to the reflectance values of different bands and phytosociological measures. Unitless β-regression coefficients in multiple linear regression analysis give the relative importance of an independent variable eliminating the effect of other variables. Commonly used reflectance indices regressed on phytosociological variables revealed that NDVI followed by NLI explained maximum variability in density, biomass, Shannon's index and Simpson's index. GCI followed by DVI, and SR followed by DVI explained maximum variability in abundance and Margalef's index respectively. ANN models fitted well to the data. The correlations between the predicated vs observed values from ANN for density, biomass, abundance, Shannon's, Simpson's and Margalef's indices were highly significant (p < 0.001). This is the first baseline study in the vicinity of river Beas, and further studies are required to understand the changes in the pattern of vegetation and their correlation with the satellite data.