Global climate change threatens cultural heritage via water erosion, presenting considerable challenges for heritage conservation. Assessing and predicting the risk of water erosion to heritage helps formulate scientifically informed protection strategies, thereby enhancing the adaptability of cultural heritage sites to future climate variability. This paper investigates the current status of water erosion on the Great Wall in the Yellow River Basin using unmanned aerial vehicles (UAVs), assesses the probability of water erosion with the H2O-AutoML algorithm, analyses the driving factors through SHapley Additive exPlanations (SHAP), and presents future climate scenarios based on different Shared Socioeconomic Pathways (SSPs) from the Coupled Model Intercomparison Programme6 (CMIP6) to predict the risk of water erosion up to the year 2100. Projections indicate that the Gradient Boosting Machine (GBM) model (accuracy: 0.740, AUC: 0.809), demonstrates the best predictive performance and reliably reflects the occurrence of water erosion at heritage sites. The Great Wall in the study area encounters significant water erosion risks, with high-risk areas constituting 56.50 % of the total length analysed; these areas are predominantly concentrated, while very low-risk areas are more geographically dispersed. The main driving factors contributing to water erosion at heritage sites include the terrain ruggedness index (TRI), topographic wetness index (TWI), slope, stream power index (SPI), and precipitation seasonality (PS). Water erosion risk to heritage sites is projected to escalate under future climate scenarios, particularly under the SSP585 scenario. This research establishes a scientific basis for formulating more accurate and comprehensive climate change adaptation and conservation plans for cultural heritage.