On May 26, the Leading Sci-tech Achievements Award of the China International Big Data Industry Expo (Big Data Expo) was officially released. Led by the Institute of Agricultural Resources and Regional Planning (IARRP) of the Chinese Academy of Agricultural Sciences (CAAS), and jointly completed by Ningxia University, the National Satellite Meteorological Center, the National Space Science Center of the Chinese Academy of Sciences, the National Meteorological Center and Shandong Jianzhu University The " Reconstruction technology of surface temperature products based on thermal infrared remote sensing retrieval and meteorological data fusion" project won the "Excellent Project" award of the 2022 Big Data Expo and obtained the Leading Scientific and Technological Achievement Award.

The Leading Scientific and Technological Achievement Award is co-sponsored by the National Development and Reform Commission of the PRC, the Ministry of Industry and Information Technology, the Cyberspace Administration of China and the Guizhou Provincial People's Government, and the Big Data Expo is the world's first big data exposition. Since its launch in 2015, guided by the philosophy of "global vision, national height, industrial perspective and business standpoint", it has explored a new mechanism for international cooperation in the digital era, contributed Chinese solutions to global big data development, and promoted the technical application and industrial development of big data around the world.

Surface temperature is a key parameter for meteorological forecasting and agricultural disaster monitoring. The rapid and accurate acquisition of large-area and long-term surface temperature information with remote sensing technology is in line with the current trend of scientific and technological development, so as to respond to various global natural disasters and monitor crop growth around the world to further ensure global and national food security. The technology uses deep learning to couple physical and statistical methods to establish a deep learning surface temperature inversion framework driven by "physical model + statistical methods + expert knowledge", which would seem to solve the separation problem of surface temperature and emissivity. By using radiation transfer model simulation and high-precision statistical data, the training and test data required for deep learning are available, and the problem of surface temperature inversion is technically solved with high precision. When the observation angle is less than 10 degrees and 65 degrees, the theoretical accuracies are below 0.1 K and 0.5 K respectively.

This technology provides a general algorithm model for thermal infrared sensor surface temperature inversion, and also provides a scheme for thermal infrared sensor band design. On this basis, the team further fuse the inversion results with the surface meteorological station data to solve the problem of missing surface temperature data under cloudy conditions, and produce a spatiotemporally continuous surface temperature dataset. This technical framework will become the main general paradigm for remote sensing inversion and reconstruction of global surface temperature and other parameters, especially the main standard algorithm for satellite surface temperature inversion and data product reconstruction. The combination of deep dynamic learning neural network with radiative transfer model and high-precision statistical data, that is, deep learning coupled with physical and statistical methods to retrieve surface temperature and emissivity is a milestone in the history of remote sensing surface temperature and emissivity inversion.