Spatiotemporal distribution of short-duration heavy precipitation and its correlation with geographical-meteorological factors in Xinjiang, China

This study aims to investigate the spatiotemporal distribution characteristics of short-duration heavy precipitation in Xinjiang and explore its spatially heterogeneous relationships with key geographical-meteorological factors (altitude, slope, NDVI, PWEI, and mean temperature), thereby providing a scientific basis for improved meteorological forecasting and disaster prevention in this arid to semiarid region. Using hourly meteorological data from 926 ground meteorological stations across Xinjiang from May to September between 2016 and 2024, we performed a comprehensive analysis of the temporal variations and spatial patterns of short-duration heavy precipitation. A multiscale geographically weighted regression (MGWR) model was employed to quantify the varying influences of geographical-meteorological factors on the amount of short-duration heavy precipitation. The results revealed considerable interannual variability in short-duration heavy precipitation events. The frequency exhibited a declining trend from 2016 to 2023, followed by a notable recovery in 2024. Seasonal analysis showed a clear unimodal pattern from May to September, with precipitation amount and frequency initially increasing, reaching their peak in July, and then gradually decreasing. Diurnally, precipitation events gradually increased from 06:00 to 12:00 local time, with a distinct peak during the 10:00-12:00 period, followed by a gradual decline in the afternoon and evening hours. Spatially, the short-duration heavy precipitation distribution is highly heterogeneous. Areas with high values are primarily concentrated in the Tianshan Mountains and their surrounding regions, whereas plain areas and arid basins experience lower amounts. Model performance comparisons demonstrate that the MGWR model provides a substantially better fit than both the ordinary least squares and standard geographically weighted regression models, effectively capturing the spatial heterogeneity of factor influences. Analysis of standardized regression coefficients from the MGWR output indicates that altitude has the most significant impact on precipitation amount, particularly in mid-elevation regions between 1000 and 2000 m. Precipitation shows high sensitivity to slope within the 0°-10° range. It also exhibits increased sensitivity to NDVI in areas with higher values. Meanwhile, the influences of PWEI and mean temperature on precipitation amounts are relatively limited. Spatial analysis of dominant factors across meteorological stations reveals that altitude dominates 73.43% of the stations, primarily located in the central, northern, and southern mountainous regions and their piedmont transition zones. Slope is the dominant factor in 14.26% of the stations, concentrated in western Bortala, the southern foothills of the central Tianshan Mountains, the Bogda Mountain region in Changji, and the eastern Tianshan Mountains in Hami. NDVI dominates 10.26% of the stations, mainly distributed in central Kashgar, where vegetation coverage is relatively high. PWEI and mean temperature dominate only 1.62% and 0.43% of the stations, respectively, indicating their secondary roles in influencing precipitation patterns across most of Xinjiang.