Precipitation provides the most crucial input for hydrological modeling. However, rain gauge networks, the most common precipitation measurement mechanisms, are sometimes sparse and inadequately distributed in practice, resulting in an imperfect representation of rainfall spatial variability. The objective of this study is to analyze the sensitivity of different model structures to the different density and distribution of rain gauges and evaluate their reliability and robustness. Based on a rain gauge network of 20 gauges in the Jinjiang River Basin, south-eastern China, this study compared the performance of two conceptual models (the hydrologic model (HYMOD) and Xinanjiang) and one process-based distributed model (the water and energy transfer between soil, plants and atmosphere model (WetSpa)) with different rain gauge distributions. The results show that the average accuracy for the three models is generally stable as the number of rain gauges decreases but is sensitive to changes in the network distribution. HYMOD has the highest calibration uncertainty, followed by Xinanjiang and WetSpa. Differing model responses are consistent with changes in network distribution, while calibration uncertainties are more related to model structures.

HIGHLIGHTS. We systematically evaluated the effect of initial parameter distribution on the estimation process of different parameters in the EnKF method.We sys

Parameter uncertainty analysis is one of the hot issues in hydrology studies, and the Generalized Likelihood Uncertainty Estimation (GLUE) is one of the most widely used methods. However, the scale of the existing research is relatively small, which results from computational complexity and limited computing resources. In this study, a parallel GLUE method based on a Message-Passing Interface (MPI) was proposed and implemented on a supercomputer system. The research focused on the computational efficiency of the parallel algorithm and the parameter uncertainty of the Xinanjiang model affected by different threshold likelihood function values and sampling sizes. The results demonstrated that the parallel GLUE method showed high computational efficiency and scalability. Through the large-scale parameter uncertainty analysis, it was found that within an interval of less than 0.1%, the proportion of behavioral parameter sets and the threshold value had an exponential relationship. A large sampling scale is more likely than a small sampling scale to obtain behavioral parameter sets at high threshold values. High threshold values may derive more concentrated posterior distributions of the sensitivity parameters than low threshold values.

Long Short-Term Memory (LSTM) proves highly effective in addressing complex nonlinear simulation challenges, making it particularly suitable for reservoir prediction applications. Based on LSTM, we conducted a comprehensive study simulating the operation mode of the Three Gorges Reservoir, accounting for its highly nonlinear inflow and outflow patterns. The model takes the inputs of inflow, outflow, and water level data of the Three Gorges Reservoir to generate the outflow predictions across different foresight periods. The research demonstrates a commendable level of prediction accuracy, underscoring the model's efficacy in simulating the reservoir discharge. Furthermore, the prediction accuracy is intimately tied to the length of the foresight period, the measured flow rate, and the number of neurons.