Study On Downscaling Of Land Surface Temperature Considering Urban Spatial Morphological Parameter

Land Surface Temperature(LST)is a key physical parameter for studying the hydrothermal balance of the Earth system,and plays an important role in estimating surface thermal inertia,soil moisture,surface evapotranspiration,urban heat island and other studies.Compared with the natural surface,the spatial variability of LST in urban areas is higher,so how to obtain reliable high spatial and temporal resolution LST has been a hot spot for urban thermal environment research.However,looking around the current global thermal infrared data generally has the problem of conflicting spatial and temporal resolutions.To solve this problem,this study downscales the existing surface temperature data to obtain high spatial and temporal resolution surface temperature products.Firstly,to address the problem that surface temperature downscaling studies with spatial resolution of 30 m and above often lack validation data,this study uses an unmanned aerial vehicle mounted thermal camera to obtain local high-resolution thermal infrared data of the study area and invert to obtain 10 m surface temperature to validate the local accuracy of the 10 m downscaled LST.Secondly,to address the problems that the current urban LST downscaling span is large,which often leads to the reduction of accuracy and inaccurate LST spatial distribution prediction,the study constructs a step-by-step downscaling model based on random forest to downscale 900 m Sentinel-3LST to 450 m,150m,30 m and 10 m step-by-step to improve the urban surface temperature downscaling effect when the scale span is large.Finally,in view of the fact that the current urban downscaling drivers rarely consider the influence of urban three-dimensional features on surface temperature,the study introduces a variety of spatial morphological parameters as drivers to analyze the influence of spatial morphological parameters on surface temperature downscaling..The results show that(1)the surface temperature inversion of UAV thermal infrared data is highly accurate with absolute error between-2℃ and 3℃ and root mean square error of 2.35℃,using the actual ground sample temperature as the validation data,and can be used for downscaling validation.(2)At 30 m spatial resolution,direct downscaling LST cannot accurately characterize the spatial distribution of urban LST,while step-by-step downscaling LST performs well in both high and low temperature regions.The correlation between stepwise downscaling LST and Landsat LST is also improved.Compared with the two downscaling results,the stepwise downscaling accuracy improves significantly,and the RMSE decreases from 2.01°C to 1.08°C.(3)After considering the urban spatial morphology parameter in the driving factor,the temperature underestimation in the streets of the densely built area is reduced and the LST distribution is more continuous at 30 m.At10 m resolution,the RMSE of the campus synchronous experiment area is reduced by 0.31 ℃-0.37℃ after adding the urban spatial morphology parameter.In contrast,the accuracy of the same area is only improved by 0.12 ℃ after adding the urban spatial morphology parameter at 30 m resolution.This indicated that the urban spatial morphology parameters effectively improved the phenomenon that the temperatures of buildings and surrounding roads were underestimated while the temperatures of water bodies and vegetation were overestimated when only the remote sensing spectral index was used to downscale,and the effect was more obvious at a higher spatial resolution of 10 meters.