Development of high-resolution summer precipitation data for Xinjiang Region by fusing satellite retrieval products and Gauge observations

Abstract

High-resolution precipitation data aid climate research and forecasting, reveal precipitation mechanisms, assess extreme events, provide empirical support for models, enhance prediction accuracy, and have application value for weather forecasting and beyond. The Xinjiang region of China, characterized by its vast expanse and complex terrain, exhibits a pronounced spatial and temporal disparity in precipitation distribution. Traditional ground meteorological observation stations are sparse and unevenly distributed, leading to considerable limitations and uncertainties in precipitation observation data. The Integrated Multi-satellite Retrievals for Global Precipitation Measurement products (i.e., IMERG) provide new-generation satellite precipitation measurements, but they are inaccurate in regions with complex terrain. Leveraging the advantages of multiple data sources to achieve complementary fusion of precipitation data can effectively increase the accuracy and spatiotemporal resolution of data. In this study, we proposed a merged (automatic weather station and IMERG measurements) high-spatiotemporal resolution (0.1° × 0.1°) hourly precipitation product (M-AWSI), and then evaluated its applications. For the 2027 AWS in Xinjiang, the RBFN (radial basis function neural network) method was used to obtain the gridded data, and RBFN can overcome the insufficient of traditional interpolation in local approximation ability. Furtherly, the gridded data is fused with the IMERG data by using an optimized probability matching total correction scheme, where multiple constraints are incorporated, such as effective correction radius and distance weight correction to avoid temporal and spatial discontinuity of the data in neighboring areas. Compared with observational data, the IMERG product effectively captures the spatial distribution characteristics of precipitation in the Xinjiang region. However, it exhibits significant underestimation of heavy precipitation and overestimations of weak precipitation, while failing to accurately depict the peak time in the diurnal precipitation variation. The M-AWSI data have markedly elevated the representation indices for daily precipitation across various intensities, with particularly prominent performance in augmenting the hit rate for identifying heavy rain and rainstorm events. Furthermore, in relation to the hourly probability density distribution and the attributes of daily precipitation variability, the alignment between M-AWSI and observational data has been significantly strengthened. Additionally, the M-AWSI data demonstrates a substantial improvement in its ability to represent extreme precipitation zones and their evolutionary characteristics compared to IMERG data. The M-AWSI data effectively overcomes the limitations of IMERG, which tend to underestimate heavy precipitation and overestimate weak precipitation. The establishment of this dataset will contribute to a deeper understanding of precipitation characteristics, particularly extreme precipitation events, in the Xinjiang region.