The spatiotemporal regimes of glacier runoff (GR) under a warming climate are of great concern, especially in dryland areas in northwestern China (DAC). Due to the difficulty of observing GR, little attention has been given to the spatiotemporal change in GR at regional scales. This study uses the regional individual glacier mass balance (GMB) dataset developed by digital elevation models (DEMs) to simulate the spatiotemporal regime of GR using atmospheric parameters considering both ablation and accumulation processes on glaciers. In this study, GR, including glacier meltwater runoff (MR) and delayed water runoff (DR) of the DAC, was quantitatively assessed at a catchment scale from 1961 to 2015. The total annual GR in the DAC was (100.81 ± 68.71) × 10⁸ m³ in 1961–2015, where MR accounted for 68%. Most basins had continuously increasing tendencies of different magnitudes from 1961 to 2015. The least absolute shrinkage and selection operator (LASSO) and random forest techniques were used to explore the contributions of climate factors and glacier physical properties to GR, and the results indicated that climate factors could explain 56.64% of the variation. In comparison, the remaining 43.36% could be explained by the physical properties of glaciers themselves (i.e., degree-day factor on ice, degree-day factor on snow, glacier median height, aspect, and slope). This study not only improves our understanding of the spatiotemporal change in GR in the drylands of northwestern China at spatial and temporal resolutions but also highlights the role of physical properties in explaining the heterogeneous dynamics among GRs unlike previous studies that only emphasize rising temperatures.