High-resolution reconstruction and correction of FY-4A precipitable water vapor in China using back propagation neural network

Abstract

Precipitable water vapor (PWV) with high spatial and temporal resolution is essential for a deeper understanding of the study area's hydrological cycle and climate change. This study uses the European Center for Medium-Range Weather Forecasts Reanalysis 5 (ERA5) PWV as auxiliary verification data and proposes a reconstruction method based on back propagation neural network (BPNN) downscaling. Combining with high temporal resolution meteorological elements, digital elevation model (DEM), location, and time, reconstructed the Fengyun-4A (FY-4A) PWV data set with high spatial and temporal resolution in China was obtained. Subsequently, using the characteristics of FY-4A PWV with DEM, location, and time, a BPNN-based correction method is proposed to correct the reconstructed FY-4A PWV using the high-precision global navigation satellite system (GNSS) PWV. The results showed that the root mean square error (RMSE), Bias, and Pearson correlation coefficient (R) of the reconstructed model with respect to the FY-4A PWV are 1.32 mm, 0 mm, and 0.99, respectively. Compared with the auxiliary verification data ERA5 PWV, the RMSE, Bias, and R of the reconstructed FY-4A PWV are 0.85 mm, 0.18 mm, and 0.99, respectively. Compared with the GNSS PWV, the RMSE of the corrected FY-4A PWV is 2.44 mm, which is 28.86% lower than that of the reconstructed FY-4A PWV, the Bias is improved from −1.46 mm to 0 mm, and the R is enhanced from 0.98 to 0.99. The reconstruction and correction method of FY-4A PWV with high spatial and temporal resolution proposed in this study dramatically improves the completeness and accuracy of FY-4A PWV in China. It lays the foundation for studying rapid changes in China's hydrological cycle and climate.