Simulation Of Land Surface Temperature At Pixel Scale Based On Ground Scene Model

Land surface temperature(LST)retrieved by remote sensing is one of the most important and sensitive input parameters of many land surface models and widely used in weather forecasting,forest fire monitoring,freezing injury monitoring,water resources utilization,urban climate monitoring and so on.LST is not only beneficial for the improvement of algorithms for remotely sensed LST retrieval,but also meaningful for further application of remote sensing products.Due to the mismatch of observation scale and the difference of the observation methods between in-site measurements and remotely sensed products in heterogeneous areas,it may be lead to large errors to the validation of remotely sensed LST when using the in-situ measurements directly.Considering geometry relationship of Sun-ground-sensor and spatial heterogeneity,a ground scene model is constructed to upscale the in-situ measurements temperature to the target pixel scales.In this paper,an observation experiment aimed to obtain the parameters of ground three-dimensional(3-D)structure and surface heat radiation was carried out in Hunhelin(HHL)station,Huyang(HY)station and Chaoji(CJ)station which located on the downstream of the Heihe River basin in July 2014,August 2014 and July 2016.The scene models are constructed by ground data and airborne data respectively for the purpose of simulating the LST in the FOV of the four component longwave radiation sensor.In the 3-D scene model,the crown is ellipsoid and the directions of illumination and observation are modeled to obtain the proportion of the components.The radiant temperatures of land surface measured by the thermal infrared cameras and the infrared radiometers are transformed into the component temperature after instrument calibrating,atmospheric correcting and other processing.The airborne data is also used to construct the ground scene model.The airborne LiDAR data is used to construct the 3-D scene,and airborne thermal infrared camera is used to acquire the component temperature.Based on the ground data,the results show that the simulated LST has a high consistency with the LST retrieved by four components longwave radiation with component temperatures obtained from the thermal infrared cameras.The Bias is 1.9 K(1.9 K)at HY(HHL)station.The RMSE(root-mean-square error)is 1.7 K(1.2 K)at HY(HHL)station.The simulated LST is 5~10 K larger than the LST retrieved from four components longwave radiation when the infrared radiometer is input as component temperatures.According to comparing the simulated LST based on airborne data and the LST retrieved by four components longwave radiation,the results show that the bias(RMSE)is 0.7 K(2.6 K).The scene model proposed in this paper shows a satisfied performance in simulating the truth values of LST in heterogeneous areas,which will be valuable for validating and quantifying the remotely sensed LST.