Optical-biophysical relationships and validation of MODIS vegetation indices with multiple fine spatial resolution data in semiarid rangelands

Abstract

The vegetation index products from the Moderate Resolution Imaging Spectroradiometer (MODIS) are designed to provide consistent, spatial and temporal comparisons of global vegetation conditions. The objective of this dissertation was to validate the robustness and global implementation of two MODIS VI algorithms, including the normalized difference vegetation index (NDVI) and "enhanced" vegetation index (EVI). Their performances have been evaluated in: (1) the normalization of canopy background (brightness) variations and the extraction of biophysical parameters across different canopy structures; (2) the characterization of seasonal vegetation profiles (phenological, intra-annual); and (3) spatial and temporal discrimination of vegetation differences (inter-annual). The validation was accomplished through multiple means, including canopy radiative transfer models which were utilized to extract pure vegetation spectra and "true" VI value free of background contamination for varying canopy structures and vegetation amount. The experimental field- and airborne-based radiometry and satellite imagery at multiple spatial resolutions were also coupled and scaled-up for comparison with coarse spatial resolution MODIS VI products to quantify characteristics of semiarid rangeland vegetation. The results showed that NDVI was advantageous in yielding biophysical relationships applicable across varying canopy types, but required knowledge of soils for biophysical estimations. The EVI provided biophysical relationships sensitive to canopy structure, thus requiring knowledge of canopy type for biophysical assessments. The MODIS VI products were successfully validated, radiometrically, by coupling field and the MODLAND Quick Airborne Looks (MQUALS) observations to high spatial resolution imagery (AVIRIS and ETM+), and appeared robust across the two parallel sites for depicting their ecological equivalents. MODIS multitemporal VI profiles were able to depict phenological activity, length of the growing season, peak and onset of greenness, and leaf turnover. Among the sensors tested, spatial resolution was found to be most important for discriminating the major land cover subtypes within the two parallel semiarid rangelands, and spectral resolution had major effects on capturing seasonal contrast due to atmosphere influences. The validation strategy utilized in this study to successively aggregate the integrity-inherent multiple fine spatial resolution data to the coarse MODIS pixel sizes appeared to perform well, thus showing potentials in the validation of other satellite products.