Combining ATC and 3D-CNN for reconstructing spatially and temporally continuous land surface temperature
2022
Fu, Huyan | Shao, Zhenfeng | Fu, P. (Peng) | Huang, Xiao | Cheng, Tao | Fan, Yewen
More than half of the satellite-derived Land surface temperatures (LSTs) data are missing due to poor weather conditions (e.g., clouds, shadows, and other atmospheric conditions) and/or sensor failure, which has greatly limited the acquisition of spatially consistent and temporally regular LST data, leading to reduced utilization and accuracy of the data and hindering the understanding of the spatial-temporal patterns of surface thermal environments. Although several reconstruction methods have been developed, they are not effective in reconstructing daily LSTs for high spatial-temporal dynamics in urban regions. In this study, we developed a hybrid reconstructing method based on the annual temperature cycle (ATC) and three-dimensional convolutional neural network (3D-CNN) models. The proposed method has been validated on both Landsat 8 Thermal Infrared Sensor (TIRS) and Moderate Resolution Imaging Spectroradiometer (MODIS) Terra satellite LST datasets. The reconstructed results show that the proposed model incorporated with auxiliary data and 3D-CNN model significantly outperforms standard annual temperature cycle (ATCs) and enhanced annual temperature cycle (ATCe) models (Root Mean Square Error (RMSE): Landsat ATCs = 4.21 K, Landsat ATCe = 3.25 K, Landsat ATCe&SATs&CNN = 0.96 K, MODIS ATCs = 3.83 K, MODIS ATCe = 3.15 K, MODIS ATCs&SATs&CNN = 0.61 K). The proposed method is simpler, more efficient with higher robustness in terms of reconstructing high spatiotemporal LSTs in urban settings compared with existing reconstruction methods.
Afficher plus [+] Moins [-]Mots clés AGROVOC
Informations bibliographiques
Cette notice bibliographique a été fournie par National Agricultural Library
Découvrez la collection de ce fournisseur de données dans AGRIS
