Daily precipitation data covering the period of 1961-2009 from 72 stations in Northwest Arid Region, China, are analysed to investigate coherent subregions of variations in extreme precipitation events. First, four extreme precipitation indices, i.e. number of heavy precipitation days (DPh), precipitation intensity above heavy precipitation (IPh), number of torrential precipitation days (DPt), and precipitation intensity above torrential precipitation (IPt), were defined and analysed seasonally. Then rotated empirical orthogonal function (REOF) technique is used to identify the coherent subregions of these seasonal extreme precipitation indices and related trends. The results indicate that the number of coherent subregions of DPh, IPh, DPt, and IPt is 6, 5, 14, and 10 for spring; 4, 18, 6, and 6 for summer; 15, 8, 10, and 8 for autumn; and 7, 7, 2, and 5 for winter, respectively. Attribute shifts of underlying surfaces should be an important driving factor behind changes of extreme precipitation events. Intensity variations of heavy and torrential precipitation are the main reason behind significant changes of extreme precipitation, especially intensity variations of the heavy precipitation. The highest increasing magnitude of IPh is 584% (10a) −1 which is found in summer in the west parts of the Southern Xinjiang, the middle Tianshan Mountains, and the west parts of the Junggar Basin in the Northern Xinjiang. The findings of this study are potentially relevant for the development of human understanding of hydrological cycle responses in arid regions to regional and global climate changes, particularly in terms of extreme weather extremes.