Water conservation is an essential parameter for maintaining the ecological balance. The Three-River Source Region (TRSR) cannot be an exception, since it is one of the most influential water conservation reserves in the Qinghai–Tibet Plateau in China. Therefore, the realization of its scientific significance can determine its future regional sustainable development and the optimal allocation of water resources. The study of the past is critical to predict the future temporal and spatial changes in the water conservation of the TRSR. The first task of this study was to obtain the optimal runoff simulations in the TRSR from 1981 to 2014 by calibrating the adjustable the parameters of the Soil and Water Assessment Tool (SWAT) model. Then, the water conservation of the TRSR from 1981 to 2014 was quantified. Finally, the future water conservation of the TRSR was also predicted using the optimal SWAT model. The predication took into consideration the three terms including the near-term (2015–2044), mid-term (2045–2074), and long-term (2075–2099) in three different climate scenarios of SSP1-1.9 (SSP119), SSP2-4.5 (SSP245), and SSP5-8.5 (SSP585). The main findings are as follows: (1) both the coefficient of determination (R²) and Nash–Sutcliffe efficiency coefficient (NSE) for runoff simulation on the three sub-basins reached above 0.78 during the calibration and validation periods, which indicates the reasonableness of the SWAT model in the TRSR. (2) From 1981 to 2014, the water conservation capacity of the TRSR showed an increasing trend (0.5135 mm/year), and its changes had significant positive correlations with precipitation and temperature. The Yellow River Source (YR) and the Yangtze River Source (YZ) had the strongest and weakest water conservation capacities, respectively. (3) From 2015 to 2099, the water conservation in the TRSR in the SSP119, SSP245, and SSP585 scenarios decreased first and then increased, increased first and then decreased, and increased steadily, respectively.

Dissolved oxygen (DO) dynamics in river systems exhibit complex nonlinear and non-stationary characteristics driven by interactions among meteorological and water-quality factors, posing challenges for accurate prediction and identification of dominant drivers. To address these challenges, this study establishes a VMD-Bayes-LSTM model by integrating Variational Mode Decomposition (VMD), Bayesian optimization, and Long Short-Term Memory (LSTM) networks. The datasets used in this study were collected from monitoring data at eight sites in the Yangtze River Basin, including water-quality and meteorological factors. Compared with several benchmark models, the VMD-Bayes-LSTM model achieves the best performance in daily DO concentration prediction, with mean R², KGE, MSE, and MAE values of 0.876, 0.900, 0.208, and 0.247, respectively. Meanwhile, this model supports multi-step prediction and achieves five-day-ahead forecasting of DO concentrations. Based on this, a SHAP value-interpretable VMD-Bayes-LSTM model is constructed. Analysis indicates that water temperature and average air temperature are the primary predictive factors at the CC, PT, DQ, LS, NJG, and ZT sites with a cumulative contribution rate exceeding 60%, reaching a maximum of 90% at the PT site; notably, at the LD and LJG sites, turbidity emerges as the dominant water-quality parameter, accounting for 59.6% and 36.2% of the predictive contribution, respectively.

Recently, Integrated Multi-satellite Retrievals for the Global Precipitation Measurement (IMERG) mission and European Centre for Medium-Range Weather Forecasts Reanalysis v5 (ERA5) precipitation datasets have been widely used in remote sensing and atmospheric studies, respectively, because of their high accuracy. A dataset of 268 site-gauge precipitation measurements over the Yellow River Basin in China was used in this study to comprehensively evaluate the performance of three high-resolution precipitation products, each with a spatial resolution of 0.1°, consisting of two satellite-derived datasets, IMERG and multisource weighted-ensemble precipitation (MSWEP), and one ERA5-derived dataset, ERA5-Land. The results revealed that the spatial distribution of IMERG annual precipitation closely resembled that of the observed rainfall and generally exhibited a downward trend from southeast to northwest. Among the three products, IMERG had the best performance at the annual scale, whereas ERA5-Land had the worst performance due to significant overestimation. Specifically, IMERG demonstrated the highest correlation coefficient (CC) above 0.8 and the lowest BIAS and root mean square error (RMSE), with values in most regions of 24.79 mm/a and less than 100 mm/a, respectively, whereas ERA5-Land presented the highest RMSE exceeding 500 mm/a, BIAS of 1265.7 mm/a, and the lowest CC below 0.2 in most regions. At the season scale, IMERG also exhibited the best performance across all four seasons, with a maximum of 17.99 mm/a in summer and a minimum of 0.55 mm/a in winter. Following IMERG, the MSWEP data closely aligned with the observations over the entire area in summer, southern China in spring and winter, and middle China in autumn. In addition, IMERG presented the highest Kling–Gupta efficiency coefficient (KGE) of 0.823 at the annual scale and the highest KGE (>0.77) across all four seasons among the three products compared with ERA5-Land and MSWEP, which had KEG values of −2.718 and −0.403, respectively. Notably, ERA5-Land exhibited a significant positive deviation from the observations at both the annual and seasonal scales, whereas the other products presented relatively smaller biases.