A METHOD FOR ESTIMATING LEAF AREA INDEX FROM LANDSAT DATA BASED ON DART MODEL AND GAUSSIAN PROCESS

Leaf area index (LAI) is a key parameter in characterizing vegetation canopy. In this paper, we proposed an efficient method to estimate LAI from Landsat surface reflectance data. The three-dimensional (3-D) discrete anisotropic radiative transfer (DART) model was used to simulate canopy reflectance by constructing real vegetation scenes. However, it took too much time to simulate reflectance because of its computational complexity. Thus, we employed Gaussian process (GP) to emulate the DART model using simulation data. LAI was retrieved by iteratively minimizing a cost function with the shuffled complex evolution (SCEUA) global optimization method. The final results demonstrated that the retrieved LAI values are in good agreement with the ground reference map.