Modeling Vegetation Reflectance from Satellite Remote Sensing Data

摘要

The development of better techniques for land vegetation cover and forest ecosystems monitoring is a major requirement for local, regional and global policy and global change science. The influence of climatic variability and anthropogenic activities on the condition of the vegetation (agricultural fields, forests, sparse) is growing up continuously. In order to characterize current and future state of vegetation and localize zones of changes must be defined the proper criteria. Vegetation land cover monitoring by satellite remote sensing data is one of the most important application of satellite imagery. Vegetation reflectance has variations with sun zenith angle, view zenith angle, and terrain slope angle. To better providing of this these effects corrections in the visible and near-infrared region of electromagnetic spectrum, was used a three parameters model and was developed a simple physical model of vegetation reflectance, by assuming a homogeneous and closed vegetation canopy with randomly oriented leaves. Multiple scattering theory was used to extend the model to function for both near-infrared and visible light. This paper aims to improve the model to be used to correct satellite imagery for bidirectional and topographic effects. Thresholding based on biophysical variables derived from time trajectories of satellite data was applied for classifying using Landsat TM and ETM, SAR ERS-1 imagery for Cernica forested area in the Eastern part of Bucharest town, Romania. Classification accuracies are function of the class, comparison method and season of the year.