Runoff simulation is essential for effective water resource management and plays a pivotal role in hydrological forecasting. Improving the quality of runoff simulation and forecasting continues to be a highly relevant research area. The complexity of the terrain and the scarcity of long-term runoff observation data have significantly limited the application of Physically Based Models (PBMs) in the Qinghai–Tibet Plateau (QTP). Recently, the Long Short-Term Memory (LSTM) network has been found to be effective in learning the dynamic hydrological characteristics of watersheds and outperforming some traditional PBMs in runoff simulation. However, the extent to which the LSTM works in data-scarce alpine regions remains unclear. This study aims to evaluate the applicability of LSTM in alpine basins in QTP, as well as the simulation performance of transfer-based LSTM (T-LSTM) in data-scarce alpine regions. The Lhasa River Basin (LRB) and Nyang River Basin (NRB) were the study areas, and the performance of the LSTM model was compared to that of PBMs by relying solely on the meteorological inputs. The results show that the average values of Nash–Sutcliffe efficiency (NSE), Kling–Gupta efficiency (KGE), and Relative Bias (RBias) for B-LSTM were 0.80, 0.85, and 4.21%, respectively, while the corresponding values for G-LSTM were 0.81, 0.84, and 3.19%. In comparison to a PBM- the Block-Wise use of TOPMEDEL (BTOP), LSTM has an average enhancement of 0.23, 0.36, and −18.36%, respectively. In both basins, LSTM significantly outperforms the BTOP model. Furthermore, the transfer learning-based LSTM model (T-LSTM) at the multi-watershed scale demonstrates that, when the input data are somewhat representative, even if the amount of data are limited, T-LSTM can obtain more accurate results than hydrological models specifically calibrated for individual watersheds. This result indicates that LSTM can effectively improve the runoff simulation performance in alpine regions and can be applied to runoff simulation in data-scarce regions.
Runoff simulation and forecasting is of great significance for flood control, disaster mitigation, and water resource management. Alpine regions are characterized by complex terrain, diverse precipitation patterns, and strong snow-and-ice melt influences, making accurate runoff simulation particularly challenging yet crucial. To enhance predictive capability and model applicability, this study takes the Upper Jinsha River as a case study and comparatively evaluates the performance of a physics-based hydrological model BTOP and the data-driven deep learning models LSTM and BiLSTM in runoff simulation and short-term forecasting. The results indicate that for daily-scale runoff simulation, the LSTM and BiLSTM models demonstrated superior simulation capabilities, achieving Nash–Sutcliffe efficiency coefficients (NSE) of 0.82/0.81 (Zhimenda Station) and 0.87/0.86 (Gangtuo Station) during the test period. These values are significantly better than those of the BTOP model, which achieved a validation NSE of 0.57 at Zhimenda and 0.62 at Gangtuo. However, the hydrology-based structure of the BTOP model endowed it with greater stability in water balance and long-term simulation. In short-term forecasting (1–7 d), LSTM and BiLSTM performed comparably, with the bidirectional architecture of BiLSTM offering no significant advantage. When it came to flood events, the data-driven models excelled at capturing peak timing and hydrograph shape, whereas the physical BTOP model demonstrated superior stability in flood peak magnitude. However, forecasts from the data-driven models also lacked hydrological consistency between upstream and downstream stations. In conclusion, the present study confirms that deep learning models achieve superior accuracy in runoff simulation compared to the physics-based BTOP model and effectively capture key flood characteristics, establishing their value as a powerful tool for hydrological applications in alpine regions.
With accelerating urbanisation, extreme rainfall events have become increasingly frequent, leading to rising urban flooding risks that threaten city operation and infrastructure safety. The rapid expansion of impervious surfaces reduces infiltration capacity and accelerates runoff responses, making cities more vulnerable to short-duration, high-intensity storms. Although the SWMM is widely used for urban stormwater simulation, its application is often constrained by the lack of detailed drainage network data, such as pipe diameters, slopes, and node connectivity. To address this limitation, this study focuses on the main built-up area within the Second Ring Expressway of Chengdu, Sichuan Province, in southwestern China. As a regional core city, Chengdu frequently experiences intense short-duration rainfall during the rainy season, and the coexistence of rapid urbanisation with ageing drainage infrastructure further elevates flood risk. Accordingly, a technical framework of “open road data substitution–automated modelling–SWMM-based assessment” is proposed. Leveraging the spatial correspondence between road layouts and drainage pathways, open road data are used to construct a virtual drainage system. Combined with DEM and land-use data, Python-based automation enables sub-catchment delineation, parameter extraction, and network topology generation, achieving efficient large-scale modelling. Design storms of multiple return periods are generated based on Chengdu’s revised rainfall intensity formula, while socioeconomic indicators such as population density and infrastructure exposure are normalised and weighted using the entropy method to develop a comprehensive flood-risk assessment. Results indicate that the virtual drainage network effectively compensates for missing pipe data at the macro scale, and high-risk zones are mainly concentrated in densely populated and highly urbanised older districts. Overall, the proposed method successfully captures urban flood-risk patterns under data-scarce conditions and provides a practical approach for large-city flood-risk management.
This chapter reviews recent advances in the application of artificial intelligence for satellite precipitation data fusion, downscaling, and flood forecasting. Against the backdrop of global climate change and frequent extreme hydrometeorological events, particular emphasis is placed on the persistent challenges encountered by satellite precipitation products in complex terrain. AI and deep learning techniques have overcome many limitations of traditional forecasting methods by effectively addressing non-stationary spatiotemporal issues and delivering superior performance. In flood forecasting, rapid high-resolution simulations driven by AI not only significantly enhance the accuracy of numerical weather prediction (NWP) but also provide novel insights into the complex process through which precipitation uncertainty translates into hydrological risk. Emerging paradigms such as physics-informed neural networks exemplify the potential for an organic integration of process-driven hydrometeorology and data-driven AI, offering promising prospects for improved forecast accuracy and the development of adaptive warning systems, particularly in flash flood-prone regions such as the Tibetan Plateau. This chapter synthesizes the latest progress in understanding the error characteristics of satellite precipitation products, multi-source data fusion, downscaling, and flood forecasting. It advocates for the deep embedding of domain-specific physical mechanisms into AI frameworks, thereby providing a scientific foundation and decision-making support for flood control and disaster mitigation in the Upper Yangtze urban agglomeration, while simultaneously advancing hydrometeorological forecasting technologies on a global scale.
The terrain and climate of the Qinghai–Tibetan Plateau make it hard to assess satellite precipitation. GSMaP (Global Satellite Mapping of Precipitation) is a widely used rainfall dataset, but direct comparisons of its versions and products over the Plateau are still limited. In this study, we evaluate four GSMaP products—Gauge, GNRT, MVK and NRT—across four versions (v05–v08) using daily station precipitation data from 2001 to 2022 as the reference. We assess both precipitation amount and precipitation event detection. The analysis is carried out at the station scale and then examined by month, season, year, rainfall intensity and space. We also compare regional patterns across the Plateau. The results show that GSMaP performance generally improves in later versions. Among them, v08 is usually more stable and more consistent, especially for gauge-corrected products. This improvement appears not only in better agreement with station data but also in smaller differences among stations for some products. Still, the size of the improvement is not the same for all products, seasons, rainfall classes and regions. The improvement is more clear in wetter areas and in warm seasons. By contrast, uncertainty is still relatively large in cold seasons, under strong rainfall and in the high-elevation interior of the Plateau. Non-gauge products also show wider variation than the Gauge product, which suggests that gauge correction still plays an important role in improving consistency. In general, version updates help improve GSMaP performance under some conditions, but the gains are not the same across different climate settings, rainfall intensities, or elevation zones. This study provides a systematic evaluation of GSMaP over the Qinghai–Tibetan Plateau for 2001–2022 and offers practical support for choosing and using GSMaP products in complex terrain.
Accurate characterization of precipitation in complex terrain is essential for hydrological modeling and climate studies. This study uses daily observations from 156 rain gauges in Sichuan Province (2015–2020) to evaluate two high-resolution satellite products (GSMaP-GNRT and IMERG-Early) and to develop a Transformer-based fusion framework at the gauge scale. All three datasets reproduce the regional seasonal cycle with more rainfall in summer and less in winter. At the daily scale, the fused product attains correlation comparable to GSMaP, while GSMaP and the fusion slightly overestimate precipitation (Bias = 6.24% and 5.21%), and IMERG shows stronger underestimation (Bias = −11.46%). At the monthly scale, the fused dataset achieves the best overall performance in terms of correlation, bias and RMSE. Spatially, the fusion reduces bias and RMSE and yields more homogeneous patterns over Sichuan’s complex terrain. Detection metrics indicate that the fused product increases the probability of detection and slightly improves the critical success index, while the false alarm ratio remains relatively high and comparable to the original products. This implies a gain in event sensitivity and spatial consistency rather than substantially reduced false alarms. Overall, the Transformer-based fusion provides a useful compromise between GSMaP and IMERG, adding value particularly for bias reduction, monthly statistics and event detection. The fused dataset offers a promising input for precipitation monitoring, hydrological simulation and disaster-risk analysis in Sichuan and similar mountainous regions.