Modelling Thermal-infrared Directional Brightness Temperatures and Inversion of Component Temperatures Over Vegetation Canopies

Author:Bian Zunjian

Supervisor:Willow admire fire, xiao Qing, du Yongming

Database:Doctor

Degree Year:2018

Download:71

Pages:150

Size:13901K

Keyword:

Land surface temperature(LST)is a vital variable for processes of energy budget and water cycle of the surface-atmosphere interface.Currently,inversion for LSTs can be performed at different scales using remote sensing technique.However,retrieved LSTs are not only dependent on surface temperature state,but also affected by viewing angles.Thermal infrared(TIR)directional models can be used to explain the relationship between temperatures and emissivity of individual components and top-of-canopy(TOC)directional brightness temperatures(BT).However,limitations of these proposed directional BT models still exist:1)the vegetation clumping effect are not considered;2)many directional BT models can only simulate instantaneous observational results,in which component temperatures are required;3)existing algorithms are hardly applied for satellite operational data preprocessing.Because of above mentioned problems,in this paper,we aim to explore the TIR directionality over vegetation canopies in modelling and inversion perspectives.Major conclusions can be found as follows:(1)An inversion algorithm for leaves,sunlit soil and shaded soil temperatures was proposed using airborne TIR multi-angle observations.Based on an analytical model,FR97,components’effective emissivity were calculated by introducing sunlit and shaded fractions of visible soil.Temperatures of leaves,sunlit and shaded soil were inverted by a Bayes optimal strategy.The temperature difference between sunlit and shaded soil can reach up to 5.0°C in both inverted and measured results.This proposed algorithm performed better than a two-component algorithm,and displays a low sensitivity to inversion noise compared to a four-component algorithm.(2)An analytical four-component directional BT model was proposed for crop and forest canopies.Based on(1),a four-component directional BT model was proposed.Then,a vegetation clumping index was introduced for geometric information of crop and forest canopies.Based on simulated datasets generated by the TRGM(Thermal Radiosity-Graphic-Model)and measured datasets,the proposed model with vegetation clumping index performed well for both crop and forest canopies,and the proposed model that considers temperature differences between sunlit and shaded components performed well for hot spot effect.(3)In the paper,a combined model of a radiative transfer model and energy budget method was proposed.It is because of the difficulty to obtain component temperatures in a large-scale complex surface.Therefore,we adopted an iteration strategy between radiative transfer and energy budget processes to simulated component temperatures and TOC directional BTs synthetically.A validation was performed based on airborne TIR multi-angle observations.Simulated results agreed well with measured results with coefficients of determinate(R~2)larger than 0.6 and Root Mean Squared Errors(RMSEs)less than 0.32°C.(4)A semi-empirical algorithm was proposed for angular normalization of BTs using visible and near-infrared data.Because of limitations of satellite TIR multi-angle observations during a short period and lack of a priori information of surface structure,many existing directional BT models can not be applied for satellite applications directly.Therefore,based on a local regression relationship between spectral variables,i.e.vegetation index and brightness factor,and BTs,a semi-empirical algorithm was proposed.Based on airborne measured TIR data,this proposed algorithm performed better for directional anisotropies of BTs than Vinnikov and RL models.A series of modifications for a directional BT model can help understand and invert component temperatures from directional BTs.The angular normalization algorithm can provide potential solutions for operational preprocessing of satellite data in the future.