LAUSR

Given the threats from rainstorm events that have been intensified due to climate change, the investigation of heavy/extreme precipitation is attracting increasing attention. The investigation of extreme precipitation, with its restricted tools and data, is quite different from that of nonextreme precipitation. Satellite precipitation estimates (SPEs) are promising precipitation measurements due to fine spatial resolution. However, the spatiotemporal patterns of SPE extremes remain unclear. Therefore, this study attempted to systematically analyse the spatiotemporal patterns of extreme SPEs provided by the Tropical Rainfall Measurement Mission (TRMM) research product from statistical characterization to stochastic behaviour modelling. The statistical characterization was depicted using comprehensive extreme precipitation indices (EPIs), while the stochastic behaviour was modelled using the generalized Pareto distribution (GPD) to fit the ‘peak over threshold’ samples and employing the max‐stable process to fit the ‘block maximum’ samples (Max: annual 1‐day maximum, RX1day: monthly 1‐day maximum, and RX5day: monthly 5‐day maximum) of the TRMM data. The TRMM extremes were analysed across the Xijiang River Basin, China. The results of the statistical characterization revealed that compared with gauge precipitation, most TRMM EPIs were discovered to have higher temporal dynamic levels. Moreover, the spatial median of the return level for each return period derived from temporal TRMM extreme modelling with the GPD was lower than that from gauge extremes. The spatial dependence of Max in TRMM was weaker than that of either RX1day or RX5day, primarily due to different precipitation‐generating processes existing between two extreme events at annual scale compared with monthly scale. Each of the Max, RX1day, and RX5day in TRMM exhibited stronger spatial dependence than that in gauge extreme precipitation, maybe related to the area‐average‐based representation of the SPEs. These findings can enhance our insight into the spatiotemporal patterns of extreme satellite precipitation, which is beneficial for operational flood risk management.