A study on position error correction for remote sensing precipitation products from Fengyun-4 geostationary observations in extreme precipitation

Spatial and temporal resolution of satellite precipitation products is high but they have errors in geographic distribution and data accuracy, while rain gauge data have low resolution but high accuracy. Therefore, the two precipitation observation data can be fused to obtain high accuracy and high resolution precipitation products. In this paper, we propose a data fusion method based on image registration and warping in image processing for data correction of satellite precipitation products. The essential element is to construct a cost function containing a term constrained on the precipitation field differences and a term constrained on the mapping domain. The warping vector field is obtained by minimizing the cost function and applied to satellite precipitation products for the purpose of position correction. The quantitative precipitation estimate (QPE) of FY-4A is corrected by rain gauge station data in position for extreme precipitation in Henan Province in July 2021. The results show that the QPE distribution has obvious shape and position error, but after position correction, the mean absolute error, root mean square error, and position error of the QPE are reduced, while the correlation coefficient and the fit to the station improved. However, since the registration algorithm is actually a process of finding large-scale feature matching in image processing, the iterative process will be difficult to converge due to the widely varying precipitation fields.