The Improved Angular & Spectral Kernel BRDF Model Based On The Spectral Library

Land surface albedo is a critical parameter controlling the earth's radiative energy budget,which is also considered as one of the Essential Climate Variables(ECVs)monitoring the global climate change.The land surface albedo is acquired by the integration of bidirectional reflectance distribution function(BRDF)over the observation geometry,since the surface reflectance is directional.Operational algorithms for land surface BRDF/Albedo products are currently mainly developed from the kernel-driven model,which combines atmospherically corrected,multi-date and multi-band surface reflectance to extract BRDF parameters.The Angular and Spectral Kernel Driven model(ASK model),which incorporates the component spectra as a priori knowledge in kernel function,provides a potential way to make full use of the multi-sensor data with multispectral information and accumulated observations.However,currently the ASK model is still not feasible for global BRDF/Albedo inversions due to the lack of sufficient field measurements of component spectra at the large scale.In this study,we outlines a parameterization scheme on the soil and leaf component spectra in the frame of ASK model for better application on BRDF/Albedo inversions.By reviewing the history in the development of the spectral libraries,the soil and leaf spectral at the global scale are found for parameterization.After analyzing and comparing each spectral library from different wavelength range,types,ancillary data and resources sharing views,and making a summarization on the application of spectral libraries for varied disciplines,the globally distributed soil spectral library,GOSPEL,ANGERS and LOPEX leaf optical property database were used for calculating the ratio of leaf and soil reflectance.To consider the intrinsic variability in both the land cover and spectral dimension,the mean and standard deviation of ?(?)for 28 soil units and 4 leaf types were calculated,to find out the trend of the parameters in and among the different land types in seven MODIS bands.Compared to the retrievals from BRF datasets simulated by the PROSAIL model,the ASK model shows an acceptable accuracy on the parameterization strategy,with the RMSE 0.007 higher at most than inversion by true component spectra.The results indicate that the classification on ratio contributed to capture the spectral characteristics in BRDF/Albedo retrieval,whereas the ratio range should be controlled within 8% in each band.Ground-based measurements in Heihe River Basin were used to validate the accuracy of the improved ASK model,and the generated broadband albedo product shows good agreement with in situ data,with the RMSE mostly lower than 0.02.This suggests that the parameterization strategy on the soil and leaf component spectra has improved the ability of better application of the ASK model,which has the great potential for global scale BRDF/Albedo inversions.