[A-CG41] Satellite Earth Environment Observation

[ACG41-P20] Evaluation of the Extrapolation Capabilities of BRDF Models Using Airborne and Satellite-Based Land Surface Observations

*RUI CHEN¹, Wei Yang^1,2 (1.Department of Remote Sensing, Graduate School of Science and Engineering, Chiba University, 2.Center for Environmental Remote Sensing, Chiba University)

Keywords:BRDF, Extrapolation, principal plane, kernel-driven models, RPV model

The Bidirectional Reflectance Distribution Function (BRDF) is a crucial tool for describing the directional reflectance characteristics of surface objects and is essential for accurately inverting key surface parameters such as the Clumping Index (CI). CI is closely related to hotspot and darkspot phenomena on the principal plane, which are often not observed by satellites. This gap necessitates the use of BRDF models to simulate reflectance within the principal plane to more accurately invert these phenomena. Although many previous studies have focused on correcting hotspot phenomena, the evaluation of BRDF models' extrapolation capabilities on the principal plane is relatively insufficient. NASA’s Cloud Absorption Radiometer (CAR) provides comprehensive multi-angle observational data, including observations from the principal plane, offering valuable data resources for this study. Additionally, global observation satellite sensors such as MODIS, VIIRS, and SGLI, although potentially providing near-principal plane observations, require integration and analysis to validate the extrapolation capabilities of BRDF models. This study analyzes these multi-angle observational data, comparing the performance of kernel-driven models such as Rossthick-LiSparseR (RTLSR), Rossthick-LiTransit (RTLT), Maignan2004, Jiao2016, and the Rahman-Pinty-Verstraete (RPV) model along with its enhanced version, the Enhanced-RPV (ERPV) model. Results from CAR data indicate that the ERPV model demonstrates outstanding accuracy (R² = 0.9465, RMSE = 0.0353 for forest, and R² = 0.8535, RMSE = 0.0553 for cropland) and an exceptional ability to capture hotspot effects. Meanwhile, analysis using satellite data shows that, compared to kernel-driven models, the RPV and ERPV models better resemble the typical bowl-shaped BRDF when simulating the principal plane, demonstrating superior extrapolation capabilities. This advantage makes them particularly effective in estimating reflectance in regions where direct satellite observations are unavailable.