Inversion and Validation of FY-4A Official Land Surface Temperature Product

Authors: DONG, L. - Collaborator; WANG, F. - Collaborator; Cosh, Michael; MIN, M. - Collaborator

Submitted to: Journal of Photogrammetry and Remote Sensing
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 4/30/2023
Publication Date: 5/5/2023
Citation: Dong, L., Wang, F., Cosh, M.H., Min, M. 2023. Inversion and Validation of FY-4A Official Land Surface Temperature Product. Journal of Photogrammetry and Remote Sensing. 15(9):2437. https://doi.org/10.3390/rs15092437.
DOI: https://doi.org/10.3390/rs15092437

Interpretive Summary: Land surface temperature is a crucial parameter for understanding the energy, water, and carbon cycles. The accurate estimation of this parameter requires accounting for signal attenuation through the atmosphere which is most easily done via a split window algorithm. For the new FY-4A mission (Chinese satellite), a collection of split window algorithms is reviewed for application to the mission product. Comparisons are made to in situ networks from the Tibetan Plateau which is a critical component of the weather and climate system across China. Good accuracy between satellite estimates and in situ measurements was achieved. This operational satellite product will have a strong impact on global climate change research, land-atmosphere interaction research, and interpretion of land surface model predictions.

Technical Abstract: The thermal infrared (TIR) channels of Advanced Geosynchronous Radiation Imager (AGRI) onboard Fengyun 4A (FY-4A) satellite provides high-frequency, high-precision, and quantitative observation data to obtain diurnal variation of land surface temperature (LST). In this paper, seven candidates split window LST algorithms are compared to evaluate their applicability for the AGRI sensor using simulated data by the MODTRAN radiative transfer model. The Ulivieri & Cannizzaro (1985) algorithm is determined to be optimal for the enterprise algorithm of FY-4A LST operational products. The refined algorithm coef'cients were strati'ed for dry and moist atmospheres as well as for daytime and nighttime conditions by the least-square regression fitting. The FY-4A LST products are produced in clear sky condition and the results show that the diurnal variation characteristics of LST can be efficiently obtained. The FY-4A LST products is validated by using one year data of in situ measurements and Moderate Resolution Imaging Spectroradiometer (MODIS) LST product. The validation results indicate that the preferred LST algorithm meets the required accuracy (2.5 K) of the FY-4A mission: 1) Compared with in situ data of the HeBi crop measurement network, the root mean square errors (RMSE) were 2.139K and 2.447K. According to statistics, the proportion of errors between -2.5k and 2.5k accounts about for 78.0% of the error, and that of between -3.0k and 3.0k accounts for 85.0%. Compared with in situ data in Naqu site of Tibet plateau, the RMSE was 2.86K and the proportion of sample data with errors between -2.5k and 2.5k accounts for 71.9% of the error, and that of between -3.0k and 3.0k accounts for 83.63%. 2) When compared with the MODIS LST product, the RMSE was 1.64k, 2.17K, 2.6k and1.73k in March, July, October and December respectively. From the bias time change between FY4A LST and MODIS LST, RMSE of the XLHT (city) and GZH (desert) sites were 2.735 and 2.97K respectively. Further studies are necessary to improve the algorithm accuracy under the large water vapor condition. Sensitivity analysis shows that obtaining high quality land surface emissivity (LSE) is crucial for a highly accurate of FY-4A LST product. And water vapor content (WVC) is also an important concern for the algorithm to improve, particularly when the view zenith angle is large or the atmosphere is very moist. Overall, our algorithm exhibits good accuracy to produce FY-4A LST operational products. The long time series FY-4A LST data will help to the global climate change research, land-atmosphere interaction research, and weather prediction and land surface models assimilation.