Premium
Application of an LAI Inversion Algorithm Based on the Unified Model of Canopy Bidirectional Reflectance Distribution Function to the Heihe River Basin
Author(s) -
Ma Bo,
Li Jucai,
Fan Wenjie,
Ren Huazhong,
Xu Xiru,
Cui Yaokui,
Peng Jingjing
Publication year - 2018
Publication title -
journal of geophysical research: atmospheres
Language(s) - English
Resource type - Journals
eISSN - 2169-8996
pISSN - 2169-897X
DOI - 10.1029/2018jd028415
Subject(s) - leaf area index , remote sensing , inversion (geology) , vegetation (pathology) , environmental science , terrain , canopy , grassland , structural basin , normalized difference vegetation index , algorithm , hydrology (agriculture) , geology , mathematics , geography , cartography , geomorphology , ecology , geotechnical engineering , pathology , biology , medicine , archaeology
The leaf area index (LAI) is one of the most important parameters of vegetation canopy structure, which can represent the growth conditions of vegetation effectively. The accuracy, availability, and timeliness of LAI data can be improved greatly, which is of great importance to vegetation‐related research. There are various types of vegetation and terrain conditions in the Heihe River Basin, the second largest inland river basin in northwest China. It is not only helpful to evaluate the accuracy of LAI retrieval algorithms for the complex land surface but also useful to understand the fragile ecological status of the Heihe River Basin. In contrast to previous LAI inversion models, the bidirectional reflectance distribution function unified model can be applied for both continuous and discrete vegetation, and it is appropriate for analyzing heterogeneous vegetation distributions. In this work, we produced 30‐m LAI products once a month in the growing season of 2012. Results show that the algorithm can effectively retrieve LAIs. We verified the LAI product using field measurement data. The mean absolute errors in forest, farmland, and sparse grassland are 0.44, 0.56, and 0.38 respectively, and the R 2 is 0.8736. Further analysis shows that main errors come from three parts: errors in the parameters, mistakes in the vegetation classification, and interval of the look‐up table. Mixed pixel is also a problem for this model. Despite this, high resolution and applicability means this algorithm can be a good approach for LAI retrieval.