| 摘 要： |
Despite the benefits of global coverage with high spatiotemporal resolutions, satellite precipitation products (SPPs) still suffer from inadequate accuracy in natural hazard forecasts, hydrology, and water resources management. Rain/no-rain (R/NR) detection error significantly affects the accuracy of daily SPPs, which has attracted increasing attention in recent years. This paper proposed a precipitation bias correction framework (PBCF) to improve the accuracy of daily SPPs, focusing on improving the ability of SPPs to detect the occurrence of the precipitation based on a R/NR discriminative model. Multiple land and climate variables derived from ERA5Land reanalysis dataset were used to construct the R/NR discriminative model using the artificial neural network (ANN) method. A case study on the bias correction of daily precipitation of Integrated Multi-satellitE Retrievals for Global Precipitation Measurement (GPM) mission (IMERG) over Hanjiang River Basin (HRB) was conducted for the period 2004-2018. Daily precipitation of 64 meteorological stations in HRB were spatially and randomly divided into two groups: 44 stations were used for training, validating and testing the constructed R/NR discriminative model, and the other 20 stations were used to evaluate the performance of the R/NR discriminative model in different topographic areas. The results indicate that the proposed PBCF could reduce the bias of IMERG, with the correlation coefficient (R) increased by 19.4%, the root mean square error (RMSE) and the mean absolute error (MAE) decreased by 19.0% and 29.8% on the daily scale, respectively. The constructed R/NR discriminative model could improve the ability of IMERG for detecting the precipitation occurrence, with a classification accuracy of about 86.5% and the equitable threat score (ETS) increased from 0.15 to 0.58. Further analyses showed that the proposed PBCF was more efficient than the cumulative distribution function mapping method in correcting IMERG. This study provides a novel insight for the correction of daily SPPs. (c) 2021 Elsevier B.V. All rights reserved. |
在线留言
联系我们
所长信箱