Generation Of All-weather Land Surface Temperature Products With High Spatio-temporal Resolution Using Multi-source Remote Sensing Data

Land surface temperature(LST),a significant parameter to characterize the surface–atmosphere interaction and the physical process of surface energy balance,is widely used in a variety of fields including climate change and environmental monitoring.Different from the traditional method of ground station measurement,which is time-consuming and labor-intensive,satellite thermal infrared(TIR)remote sensing technology can obtain continuous and large-scale LST more efficiently.With the development of the times,Fengyun series meteorological satellites have gradually acquired mature Earth observation capabilities and formed a networked observation system for polar-orbiting and geostationary satellites.Currently,limited by satellite sensors and physical mechanisms,existing LST products can only provide clear sky LST information with high spatial or high temporal resolution.Aiming at the above problems,this thesis takes the domestic polar-orbiting satellite FY-3A/B and the geostationary satellite FY-2E as the data sources,and develops a general split-window(GSW)algorithm based on dynamic land surface emissivity(LSE).The FY-3 1km LST products from 2009 to 2020 and FY-2 1h/5km LST products from 2009 to2018 were produced through the MUlti-source data SYnergized Quantitative(MuSyQ)system.To accomplish the production of all-weather surface temperature products with1h/1km high spatio-temporal resolution,this thesis carries out LST reconstruction under cloudy condition and spatial-temporal fusion for FY-3 1km and FY-2 1h/5km LST products.The specific research contents and conclusions are as follows:(1)In the production of the FY-3 1km LST product,the global static LSE base map was first produced using the ASTER Global Emissivity Dataset(GED)to improve the LSE estimation accuracy of the bare surface.Then,based on the vegetation cover method(VCM),the dynamic change information such as snow coverage are added to produce daily dynamic LSE products.Finally,the MuSyQ 1km global daily LST product was produced based on the FY-3A/B data.In the evaluation of LST products,this thesis firstly analyzes the influence of FY-3 historical recalibration coefficient on the accuracy of MuSyQ 1km LST products.The results show that the accuracy of LST products produced with historical recalibration coefficient is significantly improved compared to the original L1 data.Then combined with the data of ground stations around the world,the radiancebased(R-based)and temperature-based(T-based)methods were used for evaluating the accuracy.For the FY-3A LST product,the RMSE of the R-based method is 0.63 K in the daytime and 0.69 K in the nighttime;the RMSE of the T-based method is 3.07 K in the daytime and 2.48 K in the nighttime.For the FY-3B LST product,the RMSE of the Rbased method is 0.62 K in the daytime and 0.69 K in the nighttime;the RMSE of the Tbased method is 3.29 K in the daytime and 2.41 K in the nighttime.The accuracy obtained by the two methods is relatively high,which can be used for subsequent research on LST reconstruction under cloudy condition and spatial-temporal fusion.(2)To carry out the LST retrieval of FY-2 geostationary satellite data,this thesis further improves the GSW algorithm based on dynamic LSE,and produces MuSyQ 1h/5km LST product.The accuracy of the LST product was evaluated using two methods: 1)For the T-based method,the ground measured data of Hi WATER and Oz Flux are used for evaluation.The results show that the RMSE at Hi WATER station is 4.26 K in the daytime and 3.02 K in the nighttime.The RMSE at Oz Flux site is 4.20 K in the daytime and 2.32 K in the nighttime.2)For the intercomparison method,taking Australia as the study area,four LST products MOD11 ? MYD11 ? MOD21 and MYD21 were used for intercomparison evaluation.The RMSEs corresponding to the four products were 4.41 K,3.79 K,2.83 K and 2.72 K in the daytime and 3.00 K,3.28 K,1.95 K and 2.17 K in the nighttime.The evaluation results show that the high temporal resolution MuSyQ 1h/5km LST product has high accuracy and can be used for subsequent research.(3)Based on the high spatial resolution MuSyQ 1km and high temporal resolution MuSyQ 1h/5km LST products obtained above,the enhanced annual temperature cycle(ATCE)model is developed for LST reconstruction under cloudy conditions.Then,combined with two all-weather LSTs of polar-orbiting and geostationary satellites,the enhanced spatial and temporal adaptive reflectance fusion model(ESTARFM)is used to produce 1 hour/1 km all-weather LST products in the study area.Evaluated by the measured data of the station,the RMSE of the LST reconstructed under the cloudy condition of MuSyQ 1km and 1h/5km are 4.73 K and 4.91 K in the daytime,1.94 K and3.63 K in the nighttime.The RMSE of spatial-temporal fusion LST is 4.81 K in the daytime and 5.80 K in the nighttime.