Translator Disclaimer
21 November 2012 Exploring optimal design of look-up table for PROSAIL model inversion with multi-angle MODIS data
Author Affiliations +
Physical remote sensing model inversion based on look-up table (LUT) technique is promising for its good precision, high efficiency and easily-realization. However, scheme of the LUT is difficult to be well designed, as lacking a thorough investigation of its mechanism for different designs, for instance, the way the parameter space is sampled. To studying this problem, experiments on several LUT design schemes are performed and their effects on inversion results are analyzed in this paper. 1,000 groups of randomly generated parameters of PROSAIL model are taken to simulate multi-angle observations with the observation angles of MODIS sensor to be inversion data. The correlation coefficient (R2) and root mean square error (RMSE) of input LAIs for simulation and estimated LAIs were calculated. The results show that, LUT size is a key factor, and the RMSE is lower than 0.25 when the size reaches 100,000; Selecting no more than 0.1% cases of the LUT as the solution with a size of 100,000 is usually valid and the RMSE is usually increased with the increasing of the percentage of selected cases; Taking the median of the selected solutions as the final solution is better than the mean or the “best” whose cost function value is the least; Different parameter distributions have a certain impact on the inversion results, and the results get better when using a normal distribution. Finally, winter wheat LAI of one research area in Xinxiang City, Henan Province of China is estimated with MODIS daily reflectance data, the validate result shows it works well.
© (2012) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Wei He, Hua Yang, Jingjing Pan, and Peipei Xu "Exploring optimal design of look-up table for PROSAIL model inversion with multi-angle MODIS data", Proc. SPIE 8524, Land Surface Remote Sensing, 852420 (21 November 2012);

Back to Top