Revealing the mechanism of hydrological and agricultural drought has been challenging and vital in the environment under extreme weather and water resource shortages. To explore the evolution process from meteorological to hydrological and agricultural drought further, multi-source remote sensing data, including the Gravity Recovery and Climate Experiment (GRACE) product, were collected in the Huaihe River basin of China during 2002–2020. Three machine learning methods, including long short-term memory neural network (LSTM), convolutional neural network (CNN), and categorical boosting (CatBoost), were constructed for hydrological and agricultural drought forecasting. The propagation time from meteorological drought to surface water storage and terrestrial water storage drought, evaluated by the standardized precipitation evapotranspiration index, was 8 and 11 months with Pearson correlation coefficients (R) of 0.68 and 0.48, respectively. Groundwater storage drought was correlated with evapotranspiration and vegetation growth with a 12-month lag time, respectively. In addition, vegetation growth was affected by the drought of soil moisture at depths ranging from 100 to 200 cm with an 8-month lag time with an R of −0.39. Although the forecasting performances of terrestrial water storage drought were better than those of groundwater storage drought and agricultural drought, CNN always performed better than LSTM and CatBoost models, with Nash–Sutclife efficiency values during testing ranging from 0.28 to 0.70, 0.26 to 0.33, and −0.10 to −0.40 for terrestrial water storage drought, groundwater storage drought, and agricultural drought at lead times of 0–3 months, respectively. Furthermore, splitting training and testing data at random significantly improved the performances of CNN and CatBoost methods for drought forecasting rather than in chronological order splitting for non-stationary data.

Pan evaporation (E_pa) has been decreasing in many regions of the world in the past decades. Analysis of the changes in E_pa helps to understand the variation of vapour in the hydrological cycle and the impact of climate change on E_pa. This study aims to investigate the spatial and temporal trends in E_pa and analyse the causes of E_pa variation in the Qiantang River Basin, East China. A self-organizing map neural network is first applied to classify the 40 hydrological stations. Based on the clustering results, a trend-free pre-whitening Mann–Kendall test is then used to investigate the historical trends in E_pa based on more than 50 years of monthly observations. Finally, sensitivity analysis is conducted for evaluating the relationship between E_pa and five climate variables and for investigating the causes of changes in E_pa using a global, variance-based method. It is found that the hydrological stations located in this basin can be classified into six sub-areas and the E_pa trends vary substantially at six sub-areas. Generally, there is a decreasing trend of E_pa in a majority of months in all sub-areas except one and the most significant decreases occur in summer. The results of sensitivity analysis reveal that E_pa is more sensitive to solar radiation, followed by minimum air temperature, wind speed, relative humidity and maximum air temperature. The contribution of interactions between different climate variables to the variation of E_pa cannot be ignored in some specific stations. Overall, the change of E_pa is mainly due to the decrease of solar radiation. This study provides an in-depth understanding of the causes of E_pa variation and valuable information for sustainable water and crop management in this region.

Sustainable water resources management is facing a rigorous challenge due to global climate change. Nowadays, improving streamflow predictions based on uneven precipitation is an important task. The main purpose of this study is to integrate the ensemble technique concept into artificial neural networks for reducing model uncertainty in hourly streamflow predictions. The ensemble streamflow predictions are built following two steps: (1) Generating the ensemble members through disturbance of initial weights, data resampling, and alteration of model structure; (2) consolidating the model outputs through the arithmetic average, stacking, and Bayesian model average. This study investigates various ensemble strategies on two study sites, where the watershed size and hydrological conditions are different. The results help to realize whether the ensemble methods are sensitive to hydrological or physiographical conditions. Additionally, the applicability and availability of the ensemble strategies can be easily evaluated in this study. Among various ensemble strategies, the best ESP is produced by the combination of boosting (data resampling) and Bayesian model average. The results demonstrate that the ensemble neural networks greatly improved the accuracy of streamflow predictions as compared to a single neural network, and the improvement made by the ensemble neural network is about 19–37% and 20–30% in Longquan Creek and Jinhua River watersheds, respectively, for 1–3 h ahead streamflow prediction. Moreover, the results obtained from different ensemble strategies are quite consistent in both watersheds, indicating that the ensemble strategies are insensitive to hydrological and physiographical factors. Finally, the output intervals of ensemble streamflow prediction may also reflect the possible peak flow, which is valuable information for flood prevention.

A novel multiple precipitation assimilation framework based on Bayesian merging with Gamma distribution (BMG) is proposed in this study. Thirty-four rainfall events are collected from 2010 to 2015 over Jinhua River watershed, China. Seven assimilating scenarios consisted of three satellite rainfall products, namely Tropical Rainfall Measuring Mission Multi-Satellite Precipitation Analysis 3B42 (TMPA 3B42), Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks (PERSIANN), and Climate Prediction Centre Morphing technique (CMORPH) and the interpolation of gauge products are compared with the ground measurements at spatial resolutions of 0.250° × 0.250°, 0.100° × 0.100°, 0.050° × 0.050°, 0.025° × 0.025°, and 0.010° × 0.010°. The results show that (1) the performance of all merging scenarios gradually decreases with the increase of spatial resolutions, however all scenarios perform consistently at different spatial scales, indicating that the BMG is robust and insensitive to spatial resolutions; (2) the merging product synthesized from PERSIANN, CMORPH, and gauge measurements outperforms other merging scenarios, suggesting their compatibility is higher than that of TMPA 3B42 and others; and (3) the contribution of an individual satellite product on different merging scenarios becomes larger as the spatial resolution increases. Overall, the proposed BMG has great potential to assimilate various rainfall products for continuous rainfall events even at large scales.