A Consistency Analysis of Land Surface Temperatures Retrieved From Several Polar-Orbiting Satellite Observations

• Published in: IEEE transactions on geoscience and remote sensing Vol. 62; pp. 1 - 16
• Main Authors: Zhong, Shouyi, Li, Hua, Bian, Zunjian, Liu, Qiang, Du, Yongming, Cao, Biao, Xiao, Qing
• Format: Journal Article
• Language: English
• Published: New York IEEE 2024; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Algorithms; Consistency; Consistency analysis; cross-satellite comparison; Data integration; Data sources; Decision trees; Drought; Earth sciences; Imaging radiometers; In situ measurement; Infrared imaging; Infrared radiometers; Intercomparison; Land cover; Land surface; Land surface temperature; land surface temperature (LST); Meteorology; MODIS; Ocean temperature; Polar orbiting satellites; Radiometers; Radiometry; Regression analysis; Satellite broadcasting; Satellite observation; Satellites; Sea surface; Spectroradiometers; Surface structure; Surface temperature; Temperature; validation; Vegetation
• ISSN: 0196-2892; 1558-0644
• DOI: 10.1109/TGRS.2024.3398067

Land surface temperature (LST) is an important variable in Earth science research and can be measured using various thermal infrared (TIR) observations. Due to variations in data sources and inversion algorithms, LST products yield inconsistent results, further affecting subsequent applications (e.g., drought and vegetation monitoring). Although many evaluation studies have been conducted, most of them have focused on product validation or differences between inversion algorithms. There is a lack of analysis of the data sources, which is important for the data fusion. Therefore, a consistency analysis was conducted herein. Mainstream polar-orbiting satellite data were selected, including data from the Moderate Resolution Imaging Spectroradiometer (MODIS), Sea and LST Radiometer (SLSTR), and Visible Infrared Imaging Radiometer Suite (VIIRS). The same inversion algorithm (split window, SW) for LST was employed across all datasets, thereby ensuring that differences in satellite data were the primary factor. Following validation based on in situ measurements, the polar-orbiting LST results were intercompared and analyzed with the LST results derived from the Himawari-8 Advanced Himawari Imager (AHI) observations. The results indicated the following: 1) similar conclusions were obtained from the intercomparison results and the ground-based validation results, with root mean square errors (RMSEs) for intercomparison results ranging from 2.9 to 3.3 K; 2) based on the intercomparison results, regression analysis revealed that surface temperature status, land cover and vegetation information, and angular factors had a significant impact on the evaluation results, with t tests yielding p values less than 0.05 for all of these factors; and 3) based on a decision tree analysis, the contributions of angular factor, surface temperature status, and land surface structure were 50.9%, 31.3%, and 17.8%, respectively. These findings enhance knowledge regarding the impact of different satellite data on LST inversion results and emphasize the necessity of preprocessing before the joint application of satellite data, such as angle normalization and radiometric calibration.