To reconstruct hotspot effect for MODIS BRDF archetypes using a hotspot-corrected kernel-driven BRDF model

Abstract

Previously, a few bidirectional reflectance distribution function (BRDF) archetypes were distilled from the routine MODIS BRDF product for capturing the major variability of anisotropic reflectance of a large number of land surfaces from MODIS, based on the RossThick-LiSparseReciprocal (RTLSR) model. Since the routine RTLSR BRDF model tends to underestimate the hotspot effect, the resulting hotspot signatures of these MODIS BRDF archetypes are underestimated to some degree inevitable. In this study, we use the entire available POLDER hotspot data as a priori to optimalize two hotspot parameters regarding hotspot height and width in a new hotspot-corrected RTLSR model. Then, the corrected model with the optimal hotspot parameter values is used to reconstruct the hotspot effect of the MODIS BRDF archetypes. This study assumes that hotspot signatures are not largely related to the overall pattern of the anisotropic reflectance provided with the routine MODIS BRDF product, particularly for MODIS that rarely acquires hotspot observations; therefore, the BRDF parameters retrieved by using the routine RTLSR model can be used as a baseline to generate hotspot effect in conjunction with the hotspot-corrected RTLSR model that fully inherits the property of the RTLSR model except for hotspot effect.