Modeling directional anisotropies of forest emissions using a thermal FRT model

Land surface temperature (LST) is an essential indicator of forest growth conditions, drought and fire. However, retrieved LSTs from satellite-based thermal infrared (TIR) observations prone to directional anisotropies, and a pixel-averaging temperature, that is usually aggregated by different components such as overstory canopy, tree trunk, understory vegetation and bottom soil, cannot directly reflect the temperature distribution inside a forest. Currently, existing analytical thermal radiative transfer models (RTM) can hardly fill this gap due to complicated structural and thermal variations in a forest stand. In this paper, a famous visible and near-infrared (VNIR) RTM, FRT, was introduced to the TIR domain, in which a themral emission term of components was added. Moreover, the effect of vertical heterogeneity of leaves’ temperatures on top-of-canopy (TOC) brightness temperatures (BTs) was also considered. Based on a simulated dataset, the effect of the tree trunk and understory vegetation on TOC BTs was found significant, and the temperature profile of leaves can increase the directional anisotropy of TOC BTs. An inter-comparison with a three-dimensional radiosity model indicated a stable performance of the modified thermal FRT model. The thermal FRT model was also validated using a measured dataset from an airborne sensor, and the evaluation result revealed a satisfactory performance with root mean squared error and coefficient of determination of 0.55°C and 0.74 in the solar principal plane (SPP), respectively, and the corresponding values were 0.21°C and 0.90 in cross-SPP. The thermal FRT model has capable of acting as not only a tool to link satellite TIR observations and component temperatures for a further sub-pixel inversion and an angular normalization study of LSTs, but also a supplement for combined applications using both VNIR and TIR remotely sensed observations.