A Radiative Transfer Model for Patchy Landscapes Based on Stochastic Radiative Transfer Theory
Yelu Zeng,Jing Li,Qinhuo Liu,Alfredo R. Huete,Baodong Xu,Gaofei Yin,Weiliang Fan,Yixuan Ouyang,Kai Yan,Dalei Hao,Min Chen
DOI: https://doi.org/10.1109/tgrs.2019.2952377
IF: 8.2
2019-01-01
IEEE Transactions on Geoscience and Remote Sensing
Abstract:The availability of global high-resolution land cover maps provides promising a priori knowledge for characterizing subpixel heterogeneity and improving predictions of directional reflectance of coarse-resolution pixels. Due to mutual shadowing and sheltering effects between the adjacent forest and cropland patches, the spectral nonlinear mixing of patchy ecotones is significant, especially when the sun illuminates the ecotone from the forest side with high solar zenith angle. The spectral linear mixture (SLM) approach leads to overestimation of the bidirectional reflectance factor (BRF) in the red band in the principal plane (PP), with a maximum absolute error (MAE) of 0.0063 and a maximum relative error (MRE) of 52.5%, and to underestimation in the near-infrared band in PP with an MAE of 0.0940 and an MRE of 14.5%. In a scenario with randomly distributed boundary orientations, the overestimation of SLM increases with the degree of fragmentation and the view zenith angle. We propose a Radiative Transfer model for patchy ECotones (RTEC). which improves $R^{2}$ from 0.61 to 0.94 in the red band of Landsat-8 directional reflectance at the validation site. The RTEC model provides an efficient and analytical approach for directional reflectance predictions over heterogeneous patchy landscapes at coarse resolution and will be used for biophysical parameter retrievals [e.g., the leaf area index (LAI)] in future applications.