Comparison of the four gridded precipitation products for estimating regional rainfall erosivity in mainland China

Soil erosion has been one of the major environmental factors to restrict the economy and social development. Rainfall erosivity can be used to represent the potential capacity of rainfall for soil erosion. It is a high demand to accurately assess the rainfall erosion force for the decision-making on soil and water conservation. Among them, the gridded precipitation products can be expected to calculate rainfall erosivity. But there is great regional variability in the different precipitation products. Therefore, this study aims to assess the adaptability of the various types of precipitation products in different regions, in order to realize the accurate estimation of rainfall erosivity. The daily precipitation data was collected from 738 meteorological stations and four sets of high-resolution gridded precipitation products (China gauge-based daily precipitation analysis, CGDPA; China meteorological forcing dataset, CMFD; Dataset of gridded daily precipitation in China (Version 2.0), CN0.5; tropical rainfall measurement mission-multi satellite precipitation analysis, TRMM-TMPA), from 2008 to 2013 in mainland China. The daily precipitation erosion force model was used to calculate the rainfall erosivity, and then to compare it with the standard data from the meteorological stations. The annual rainfall erosion distribution of precipitation products and measured data were compared to determine the multi-year trends of rainfall erosion of precipitation products and measured data by one-dimensional linear regression. Finally, the Bayesian model averaging (BMA) was used to calculate the weights of different precipitation products in the nine agricultural regions and different time scales. The results show that: 1) In terms of spatial distribution, the rainfall erosivity calculated by the four precipitation products and change rates were consistent with the ground station data. All rainfall erosivity showed an increasing pattern from the northwest to the southeast. Specifically, the CN0.5 was underestimated in the Loess Plateau, Huang-huai-hai Plain, Sichuan Basin and surrounding areas, the middle and lower reaches of the Yangtze River, Yunnan-Guizhou Plateau, and Southern China. CGDPA was overestimated in the Tibetan Plateau region near the Himalayas. 2) At different time scales, there was the consistency in the temporal trends of the four precipitation products and the measured precipitation. On the monthly scale, there was a low region-wide average rainfall erosion force between December and January, whereas, the peak values all occurred from June to August. On the seasonal scale, the low values of the global average rainfall erosion occurred in winter, whereas, the peak values occurred in summer. On the annual scale, the highest values of the global average rainfall erosion occurred in 2010, while the lowest was in 2011. Four precipitation products underestimated the rainfall erosivity. 3) The statistical indicators of CGDPA and CMFD were better than those of TRMM and CN0.5 at different time scales. CGDPA presented the highest Bayesian weight, rather than the Qinghai-Tibet Plateau region as the highest weight of the rainfall erosion force in CMFD. In general, the CGDPA and CMFD performed overall better applicability, compared with CN0.5 and TRMM. The gridded precipitation products can be expected to estimate the rainfall erosion calculation in mainland China. The finding can provide a strong reference to selecting the precipitation products for the rainfall erosion calculation in various agricultural sub-regions, particularly for regional soil and water conservation and agricultural management.