A kernel-driven model of effective directional emissivity for non-isothermal surfaces

Land surface temperature (LST) is a key geophysical parameter that reflects the combined effects of land surface energy and radiation balance. Remote sensing of LST, however, requires understanding the directional patterns of thermal emission from heterogeneous land surfaces. Recently, semi-empirical, linear, kernel-driven models have been successfully used to describe observed directional patterns of land surface reflectance. Following a similar methodology and based on a recent model of the physics of directional thermal e-mission from heterogeneous land surfaces, a kernel-driven semi-empirical model for thermal emission is developed using three kernels: (i) an isotropic kernel, corresponding to the gray-body component in a heterogeneous pixel; (ii) a two-layer canopy kernel, derived from a previous conceptual model; and (iii) a geometric optical kernel that accounts for the fact that sunlit parts are hotter than shaded parts in a pixel. The three-kernel model fits 1997 airborne directional thermal brightness data over Avignon, France, very well.