机载WIDAS地表观测的BRDF原型反演算法验证

Surface albedo qualifies the proportion of incoming light reflected by the land surface and plays an important role in the earth's energy budget. In the WATER experiment of Heihe in 2008, we developed an algorithm for estimating the albedo of Wide-angle Infrared Dual-mode line/area Array Scanner (WIDAS) based on the MODIS Bidirectional Reflection Distribution Function (BRDF) archetype-based algorithm for the narrow-angle observations of the airborne WIDAS. However, in the HiWATER experiment in 2012, the WIDAS observation angle range was upgraded from the early 30° to the maximum observation zenith angle of 52°, which impacted the data preprocessing (radiation scaling, atmospheric correction, and multi-angle registration). This condition caused significant noise and created new challenges for the surface albedo inversion of WIDAS. In the current study, we addressed new problems and adopted new surface observation data to verify the ability of the BRDF archetype-based algorithm to retrieve the albedo of WIDAS. To obtain quasi-real-time BRDF prior knowledge, we first extracted five BRDF archetypes as a priori information from 500 m MODIS BRDF parameter product (MCD43A1) within the Heihe experimental region. Then, we applied these BRDF archetypes to airborne WIDAS multi-angular observations for albedo estimates based on the hotspot-corrected linear kernel-driven BRDF model, that is, RossThickChen-LiSparseReciprocal model. Finally, field albedo measurements were conducted to validate the broadband albedo estimates. We compared three commonly used albedo estimate methods, namely, the BRDF archetype-based algorithm that was developed in our previous paper; full-inversion BRDF/albedo algorithm that has been adopted as the operational MODIS BRDF/albedo algorithm; and Lambertian assumption method that is commonly used to estimate surface albedo, especially when only nadir observations are available. The performance of the BRDF archetype-based algorithm was verified by comparison and analysis of the algorithms. Unsurprisingly, the accuracy of the albedo retrievals by using the BRDF archetype-based algorithm was obtained at 0.034, which was 18% and 71% higher than those of the full inversion algorithm and the Lambertian assumption method, respectively. The inversion and verification of the WIDAS demonstrated that the BRDF archetype-based method was noise resistant and obtained stable albedo estimates. Therefore, our previous conclusions were confirmed by using new WIDAS observations. Given the merit of the proposed archetype-based algorithm, we strongly recommend it to the user community, especially for narrow-angle observations that need a priori information for stable retrievals of surface albedo.