Water conservation function is the core of ecosystem services, which has received great attention from governments and scholars around the world. The spatiotemporal variability of water conservation capacity under the joint activity of climate change and human activities resulted in great unbalance in the development of social economy in different regions. However, the previous methods have difficulty in accurately characterizing the interactions among driving factors of water conservation capability and can only be applied to areas with a single underlying surface, which dramatically reduces the reliability of global application. Therefore, this paper couples the water conservation index (WCI) and a complex mapping relationship exploration method (geographic detector) to evaluate the long-term dynamic changes of water conservation capacity from 1985 to 2022, and analyzes its driving factors in different periods, taking the Yellow River water conservation area of China with complex underlying surface and severe climate change, as the study area. This overcomes the shortcomings of previous methods and improves the reliability of large-scale water conservation capacity research. The results show that the trend of water conservation capacity in the area changes abruptly in 2000, first decreasing before and then rising after, which is consistent with the trend of climate change. The capacity is higher in the south and lower in the north, with the best capacity in the south and the worst in the north. The temporal and spatial variation of the capacity is similar to that of SWC. SWC determined the spatial distribution pattern of water conservation capacity, which had the greatest impact, followed by evapotranspiration (ET), land surface temperature (LST) and leaf area index (LAI). SWC was most affected by precipitation, lower in the northwest of the 400mm isoprecipitation line in the basin, and higher in the southeast of the isoprecipitation line. Moreover, two-factor interactions have a greater effect on water conservation than single-factor effects. The synergistic effect of SWC and ET had the greatest impact, while the synergistic effect of SWC and Slope had the lowest impact. We found that land use change, urban expansion, and GDP growth in the study area accelerated the growth rate of overall fractional vegetation cover (FVC), which directly increased the capacity after 2000. The above research methods and results can lay a solid theoretical and methodological foundation for the protection of water conservation function under complex underlying surface conditions, providing a scientific basis for the optimal allocation of water resources, the optimization of ecosystem structure, and the making of policies related to climate change adaption, for the purpose of sustainable development of China and other regions across the world.