A semi-empirical approach for modeling the vegetation thermal infrared directional anisotropy of canopies based on using vegetation indices

摘要

Measurements of the surface thermal infrared (TIR) radiance provides an estimate of the land surface temperature (LST). However, any TIR measurement must be acquired under a certain geometry observation, which may refer to strong directional anisotropies. Although physical radiative transfer models can provide high precision directional brightness temperature simulation, they are too complex for processing large volumes of satellite data. With the objective to compare TIR measures acquired under different viewing angles, the topic of angular normalization issue for retrieved LSTs could be treated based on semi-empirical modelling. In this paper, we consider such category of models to simulate the directional anisotropy of surface brightness temperatures in combination with visible and near-infrared (VNIR) data. In these models, the vegetation fraction and the hot spot effect are depicted by a vegetation index and a brightness factor, respectively. An evaluation of the method is performed with both synthetic and measured datasets. The directional anisotropies that are fitted by this semi-empirical model demonstrate good agreement with an extensive synthetic dataset that is generated with the Soil Canopy Observation, Photochemistry and Energy Fluxes (SCOPE) soil-vegetation-atmosphere transfer model. An evaluation using airborne multi-angle TIR data also reveals that this model performs well when predicting BT directional anisotropies, with root mean square errors (RMSEs) of less than 0.31 degrees C over a maize-planted area. Relative to Roujean-Lagouarde (RL) and Vinnikov models using only TIR data, the proposed model offers better performances. In addition, for future use with satellite data, the proposed model using observations at different times and the combination with VNIR BRDF model are also evaluated, and good results are obtained. It yields a promising approach for the angular normalization of LST and mosaics of fine-scale images.