Biophysical interpretation of spectral indices for semi-arid soil and vegetation types in Niger.

Abstract

In situ radiometric field data and data simulated with a radiative transfer model were used to evaluate the performance and biophysical interpretation of spectral indices Concurrently with remotely sensed measurements, temporal biophysical measurements for different vegetation types for two semi-arid regions in Niger were made, including leaf area index (LAI), fraction of absorbed photosynthetically active radiation (fAPAR), percent vegetation cover, and biomass. The spectral dynamics of vegetation and soil were characterized at the leaf and canopy scale by optical measurements under many adverse conditions, including variability in vegetation optical and structural properties, soil reflectance properties, sun and view geometry and atmospheric perturbations. The spectral indices evaluated in this research comprised spectral vegetation indices and spectral mixture model indices, computed from spectral reflectances. The performance of different vegetation indices and their sensitivity to green and non-green vegetation and soils were compared and quantified by utilizing estimates of percent relative error in spectral vegetation indices, and estimates of vegetation equivalent noise expressed in terms of biophysical parameters (LAI, fAPAR). The soil adjusted vegetation index (SAVI) and modified normalized vegetation index (MIND VI) were improvements over the normalized difference vegetation index (NDVI), but were still sensitive to many perturbing variables such as soil and vegetation distribution, soil optical properties, litter and green vegetation optical properties and leaf angle distribution. The spectral mixture model indices were designed to be sensitive to vegetation, soil and non-green vegetation components and were shown to provide useful surface information that can aid in minimizing the noise in spectral vegetation indices, and also in improving their biophysical interpretation. Vegetation and soil brightness imagery were created from remotely sensed reflectance data, by calibrating the spectral mixture model with the data generated with a radiative transfer model. The effect of standing litter on spectral indices was shown to possibly cause both an increase and a decrease in the vegetation index, depending on the coupled spectral and structural properties of litter, green vegetation and soil. In situ measurements confirmed the results obtained from the analysis of data sets generated with a radiative transfer model. The implications of the effect of perturbing variables on spectral indices were also discussed.