The arid and semi-arid cities in northern China, such as Ordos City in the Inner Mongolia autonomous region, have faced severe water scarcity and vegetation degradation. The situation is driven by the dual pressures of climate change and intensive anthropogenic activities. In order to mitigate urban water resource shortage and associated ecological challenges, it is critical to analyze the changing trends of precipitation, and to quantify the potential of rainwater resources for utilization. Taking the main urban area of Ordos City (with an area of 361 km²) as a case study, this paper examined the spatiotemporal evolution of precipitation and evaluated the potential rainwater that could be harvested, based on the daily precipitation data (1958-2022) from 11 national meteorological stations (obtained from the National Tibetan Plateau Data Center). The results could offer a scientific basis to enhance rainwater resource management in similar arid urban areas.　　　　The methodological framework integrated multiple analytical techniques, including linear regression analysis, Mann-Kendall test. The annual precipitation showed a slight declining trend at a rate of −3.30 mm per decade, although the trend is statistically insignificant. The trend exhibited seasonal variation: the precipitation decreased in spring (−0.20 mm/decade) and summer (−3.60 mm/decade), contrasted by marginal increases in autumn (+0.60 mm/decade) and winter (+0.30 mm/decade). The Mann-Kendall test identified a major abrupt change point in 1961, which marked a regime shift from a wetter to a drier phase. Spatially, a consistent southeast-to-northwest declining gradient in precipitation was demonstrated based on interpolation technique analysis. The main urban and eastern regions, receiving significantly higher rainfall, were identified as strategically suitable for rainwater collection. To assess the potential rainwater to be used, a modified Soil Conservation Service-Curve Number (SCS-CN) model was developed, which integrated three land use types and USDA soil classes (B: sandy loam, C: loam, D: clay). The initial abstraction factor (λ=0.2) was used under Antecedent Moisture Condition II (AMC II). The model simulated the runoff depths for five hydrological year scenarios defined by exceedance probability based on the Pearson Type III distribution: 10% (wet year, 375.74 mm), 25% (moderately wet, 337.41 mm), 50% (median, 306.49 mm), 75% (moderately dry, 267.92 mm), and 90% (dry year, 237.80 mm). The results showed the potential rainwater resources ranged from 45.34 million m³ (in dry years) to 87.61 million m³ (in wet years), with an annual average of 73.50 million m³. The results were validated using a runoff coefficient method (coefficient of 0.67), which yielded an average annual potential rainwater resources of 66.33 million m³. The comparison showed an average relative error of 9.75%, which indicated the model performance was reasonable. Furthermore, the urban areas constituted a significant portion of the total runoff, although urban construction land covered a relatively small area. The main reason is the urban extensive impermeable surfaces generating more runoff, which highlights the critical role of this land type in urban rainwater harvesting.　　This paper proposed several strategic recommendations to address the challenges of high evaporation and low precipitation in Ordos. Firstly, the tailored rainwater harvesting infrastructure should be established by integrating retention measures into urban planning, such as reservoirs and cisterns. Secondly, based on specific demand and potential supply, the water resource allocation schemes should be optimized in the urban water management system. Thirdly, it is essential to promote broader societal participation in rainwater resource utilization, such as public education and awareness campaigns, disseminated through media and community programs.