Sub-pixel Model for Vegetation Fraction Estimation based on Land Cover Classification

Vegetation fraction, the radio of vegetation occupying unit area, is a very important parameter in developing climate and ecology model. However, to measure the vegetation fraction by fieldwork a job of wasting manpower and financial resources with low_precision work, which requires estimation of vegetation fraction from remote sensing data. This study explores the potential of deriving vegetation fraction from normalized difference vegetation index (\%NDVI)\% using the TM data. Under the assumption that the pixel of TM image is a mosaic structure, sub_pixel models for vegetation fraction estimation are introduced firstly in this paper. Then the idea of using different sub_pixel model for vegetation fraction estimation based on land cover classification is proposed. The "dense vegetation model" is used to calculate the vegetation fraction in woodland, orchard and city zone, and the "nondense vegetation model" is used to calculate the vegetation fraction in cropland and meadow area.\;As a result of case study in Haidian district, Beijing, the accuracy rate of vegetation fraction estimation by using "dense vegetation model" and "nondense vegetation model" synchronously based on land cove classification is obtained about 75.4%, which is 5.8% higher than that of using "dense vegetation model" only. The accuracy rate of vegetation fraction estimation by using this model is high.\;Despite the difference between observed and estimated values for some conditions, the Sub_pixel model seems to be a good approach for estimating vegetation fraction at a regional scale. This approach may be an important tool for solving the problems in the monitoring of regional vegetation fraction over large area.