To address the challenges of obtaining driving data and soil-water parameters, as well as the computational load in real-time forecasting of flood and landslide disasters on a large scale, based on the coupled hydrological-landslide model(CRESLIDE), seamless refined intelligentgrid precipitation data were introduced as the driving factor. A soil thickness model and soil type data were employed to obtain distributed soilwater parameters, land cover information and accumulated rainfall were combined to dynamically identify landslidesensitive areas, and parallel computing technology was utilized to accelerate the computational process. Subsequently, a nationalscale CRESLIDE model was constructed to conduct realtime forecasting of rainfallinduced flood and landslide disasters from 2022 to 2023, and it was verified in longsequence and individual disaster cases during the flood period from June to August. The results show that the maximum parallel computing speedup ratio of the CRESLIDE model can reach up to 5.53. Flood disasters are primarily distributed across North and Southwest China, and landslide disasters are predominantly concentrated in South China, with a notable increase in frequency observed in North China during July and August. In longsequence validation, the area under ROC curve for evaluation of the forecast accuracy of the model is above or close to 0.7, and the forecast accuracy for flood disasters is higher than that for landslide disasters. In individual case validation, the location and time of disasters forecast by the model are consistent with the observed results.