Land Surface Temperature Reconstruction Based On FY-2F Geostationary Meteorological Satellite Data

With the intensification of climate change and the frequent occurrence of various natural disasters,it has a serious impact on urban ecosystems.There is an increasing demand for environmental monitoring and research on the regional surface thermal environment.Land surface temperature(LST)is a key parameter for measuring the water-heat balance of the Earth's surface,which affects the energy exchange process and the interaction between ground and air.Therefore,it is of great significance to obtain high-quality surface temperature data.Because meteorological station observations can no longer satisfy people's needs,the remote sensing technology which can acquire large-area,synchronously-observed surface temperature images become particularly important Different from conventional polar-orbiting satelli.te surface temperatUre products,FY-2F LST products have high time resolution and can acquire one image per hour.However,its low spatial resolution and the lack of data due to factors such as cloud,aerosols,and observation angles has limited the application of its products.Based on these,this paper takes the Yangtze River Delta as research area and FY-2F LST product as data source to reconstruct and downscaling the land surface temperature.Finally,a high spatial-temporal resolution surface temperature dataset is obtained,and using the datasets to analyze regional thermal environment changes.First,the paper uses random forest to construct a 5km resolution FY-2F LST and latitude and longitude,DEM,surface reflectance bl-b7 waveband of MODIS and solar incident angle factor regression model.Then,reconstructing and downscaling land surface temperature based on the model.After reconstruction,both the time missing value rate and the spatial vacancy rate of each image are greatly reduced,and the spatial resolution is increased from 5km to 1km.The accuracy of the random forest model reconstruction is high.The determination coefficient R2 of more than 90%of the data is greater than 0.6,the average absolute error MAE is basically less than 2K,and the partial correlation error RMSE of most image is between 0.5K-2.0K.Subsequently,the paper reconstructs land surface data of 2013 using Savitzky-Golay(S-G)filter based on the characteristic of long time-series LST.After simulation verification,the simulated value is highly correlated with the true value,the MAE is 0.37,and the reconstruction accuracy is high.At the same time,the FY-2F LST daily product is reconstructed by S-G filtering,and the reconstruction result is consistent with the accuracy of the original data.Therefore,using S-G filtering to reconstruct land surface temperature is reliability.Finally,this paper analyzes the spatial-temporal variation of surface temperature in the Yangtze River Delta region based on the high spatial-temporal resolution surface temperature products which was obtained by reconstruction.The surface temperature in the southern part of the Yangtze River Delta region is slightly higher than that in the north.The surface temperature in Shanghai is the highest,southern Jiangsu Province is higher than that in northern Jiangsu,and the urban area in Zhejiang Province is higher than the mountainous area.The surface temperature changes of farmland and forest land is basically the same throughout the year,and the temperature of the town is slightly higher than the two.