Objective In the context of global climate change, frequent extreme precipitation events have become a major threat to the sustainable development of mountainous rural areas. Rural areas, as composite ecosystems dominated by natural surface conditions, maintain a high proportion of vertical water exchange with strong natural hydrological regulation advantages. However, under the combined effects of disaster-bearing vulnerability and natural hydrological cycle characteristics, eco-hydrological regulation services in rural areas often exhibit a relatively tense but basically balanced “tight equilibrium” state. Increased storm risks induced by climate change and surface condition changes caused by land use alterations pose severe challenges to this “tight equilibrium” state. Current research on eco-hydrological regulation services in mountainous rural areas is relatively weak, particularly lacking systematic evaluation frameworks for supply – demand matching relationships under future extreme precipitation scenarios. Therefore, it is urgent to predict the dynamic evolution of rural ecosystem hydrological regulation services and evaluate the matching relationship between supply and demand, so as to provide a scientific basis for proactive responses to future risk challenges in rural areas.

Methods This research takes Mentougou District, Beijing as a typical case study area, which belongs to the Taihang Mountain foothills, with the Yongding River and its tributary Qingshui River as main waterways, characterized by steep mountains, deep valleys and narrow banks, where summer monsoons during flood seasons easily form topographic convergence and shear, leading to frequent heavy precipitation events. The research selects 57 gully catchment s covering rural areas in Mentougou District as basic analysis units, establishing a comprehensive analytical framework of “climate change – land use change – eco-hydrological service supply and demand”. Three Shared Socioeconomic Pathways (SSPs) scenarios are employed, including SSP1-2.6 low emission scenario (representing strong emission reduction pathways), SSP2-4.5 medium emission scenario (reflecting moderate emission reduction efforts), and SSP5-8.5 high emission scenario (corresponding to the development pathways featuring high emissions from fossil fuels), which are combined with Generalized Extreme Value (GEV) distribution models to calculate extreme precipitation events with a 100-year return period. The patch-generating land use simulation (PLUS) model is used in combination with Markov chain model to predict land use changes in 2050 and 2100, with model validation showing overall accuracy of 0.98 and Kappa coefficient of 0.98. Based on the SCS-CN model, eco-hydrological regulation service supply capacity is quantified, using runoff regulation rate to characterize the surface runoff regulation ability of each gully catchment. Fourteen flood susceptibility characteristic indicators are selected, including meteorological characteristics (20-year, 50-year, 100-year return period precipitation), topographic surface characteristics (slope, topographic position index, terrain ruggedness index, topographic wetness index, stream power index, plan curvature, profile curvature, normalized difference vegetation index), and construction characteristics (road density, river network density, land use). Eight machine learning models are compared, including Artificial Neural Network, Support Vector Machine, Extreme Gradient Boosting, Gradient Boosting, Random Forest, Extra Trees, Light Gradient Boosting Machine (LGBM), and Categorical Boosting. The LGBM model with optimal comprehensive performance (comprehensive score: 0.754) is ultimately selected for flood susceptibility mapping to represent eco-hydrological regulation service demand levels. The Z-score method is used to standardize supply capacity and demand levels, with quadrant division determining four supply – demand matching types: high supply – high demand, high supply – low demand, low supply – high demand, and low supply – low demand.

Results Results indicate that future extreme precipitation intensity will significantly increase with emission scenarios. Historical baseline extreme precipitation intensity averages 147.2 mm/day, with a relatively small increase in mid-21st century, but a notably larger increase in the late-21st century, particularly reaching 24.8% under high emission scenarios to 184.5 mm/day. The eco-hydrological regulation service supply capacity gradually weakens over time, with a median regulation rate of 49.9% in the baseline period. By the late-21st century, low emission scenarios decline to 46.28% (8.31% decrease) and high emission scenarios significantly decline to 43.61% (13.60% decrease). Conversely, eco-hydrological regulation service demand continues rising, with a median demand level of 0.181 in the baseline period, significantly increasing to 0.220 under high emission scenarios in the late-21st century (21.55% increase). Supply – demand matching analysis shows low supply – high demand gully catchments are mainly distributed along the mainstreams of the Yongding River and the Qingshui River. In the mid-21st century, the number of low supply – high demand gully catchments is 4, 6, and 10 under the low, medium and high emission scenarios respectively. While the late-21st century sees a significant increase in the number of low supply – high demand gully catchments to 11, 12, and 20 respectively under the aforesaid three emission scenarios, with over 100% growth under all emission scenarios, showing a trend of expansion from downstream to upstream areas.

Conclusion Based on supply – demand matching type analysis, the research proposes multi-level stormwater resilience planning strategies. For low supply – high demand gully catchments, it recommends enhancing stormwater retention capacity according to eco-hydrological supply – demand gaps, adopting comprehensive measures featuring “gray – green integration” to construct flood interception ditches and collection pools, thereby forming a spatial pattern of hierarchical retention. It recommends adopting the spatio-temporal planning strategy of “zoning-based progression, dynamic adjustment” that prioritizes low supply-high demand gully catchments and then gradually expands to other gully catchment types, and adjusting comprehensive intervention measures according to emission scenario evolution, thus forming systematic response mechanisms adapted to climate change processes. The research results provide a scientific basis and methodological support for mountainous rural areas to address future extreme precipitation challenges and optimize ecological flood control, offering important reference value for enhancing regional climate adaptation capacity and helping strengthen systematic stormwater resilience in mountainous rural areas against future extreme precipitation.