This study applied a GR4J model in the Xiangjiang and Qujiang River basins for rainfall-runoff simulation. Four recurrent neural networks (RNNs)—the Elman recurrent neural network (ERNN), echo state network (ESN), nonlinear autoregressive exogenous inputs neural network (NARX), and long short-term memory (LSTM) network—were applied in predicting discharges. The performances of models were compared and assessed, and the best two RNNs were selected and integrated with the lumped hydrological model GR4J to forecast the discharges; meanwhile, uncertainties of the simulated discharges were estimated. The generalized likelihood uncertainty estimation method was applied to quantify the uncertainties. The results show that the LSTM and NARX better captured the time-series dynamics than the other RNNs. The hybrid models improved the prediction of high, median, and low flows, particularly in reducing the bias of underestimation of high flows in the Xiangjiang River basin. The hybrid models reduced the uncertainty intervals by more than 50% for median and low flows, and increased the cover ratios for observations. The integration of a hydrological model with a recurrent neural network considering long-term dependencies is recommended in discharge forecasting.

Agricultural drought can have long-lasting and harmful impacts on both the ecosystem and economy. Therefore, it is important to monitor and predict agricultural drought accurately. Soil moisture is the key variable to define the agricultural drought index. However, in situ soil moisture observations are inaccessible in many areas of the world. Remote sensing techniques enrich the surface soil moisture observations at different tempo-spatial resolutions. In this study, the Level 2 L-band radiometer soil moisture dataset was used to estimate the Soil Water Deficit Index (SWDI). The Soil Moisture Active Passive (SMAP) dataset was evaluated with the soil moisture dataset obtained from the China Land Soil Moisture Data Assimilation System (CLSMDAS). The SMAP-derived SWDI (SMAP_SWDI) was compared with the atmospheric water deficit (AWD) calculated with precipitation and evapotranspiration from meteorological stations. Drought monitoring and comparison were accomplished at a weekly scale for the growing season (April to November) from 2015 to 2017. The results were as follows: (1) in terms of Pearson correlation coefficients (R-value) between SMAP and CLSMDAS, around 70% performed well and only 10% performed poorly at the grid scale, and the R-value was 0.62 for the whole basin; (2) severe droughts mainly occurred from mid-June to the end of September from 2015 to 2017; (3) severe droughts were detected in the southern and northeastern Xiang River Basin in mid-May of 2015, and in the northern basin in early August of 2016 and end of November 2017; (4) the values of percentage of drought weeks gradually decreased from 2015 to 2017, and increased from the northeast to the southwest of the basin in 2015 and 2016; and (5) the average value of R and probability of detection between SMAP_SWDI and AWD were 0.6 and 0.79, respectively. These results show SMAP has acceptable accuracy and good performance for drought monitoring in the Xiang River Basin.

The extended range temperature prediction is of great importance for public health, energy and agriculture. The two machine learning methods, namely, the neural networks and natural gradient boosting (NGBoost), are applied to improve the prediction skills of the 2-m maximum air temperature with lead times of 1–35 days over East Asia based on the Environmental Modeling Center, Global Ensemble Forecast System (EMC-GEFS), under the Subseasonal Experiment (SubX) of the National Centers for Environmental Prediction (NCEP). The ensemble model output statistics (EMOS) method is conducted as the benchmark for comparison. The results show that all the post-processing methods can efficiently reduce the prediction biases and uncertainties, especially in the lead week 1–2. The two machine learning methods outperform EMOS by approximately 0.2 in terms of the continuous ranked probability score (CRPS) overall. The neural networks and NGBoost behave as the best models in more than 90% of the study area over the validation period. In our study, CRPS, which is not a common loss function in machine learning, is introduced to make probabilistic forecasting possible for traditional neural networks. Moreover, we extend the NGBoost model to atmospheric sciences of probabilistic temperature forecasting which obtains satisfying performances.

Potential evapotranspiration (PET) is used in many hydrological models to estimate actual evapotranspiration. The calculation of PET by the Food and Agriculture Organization of the United Nations (FAO) Penman–Monteith method requires data for several meteorological variables that are often unavailable in remote areas. The China Meteorological Assimilation Driving Datasets for the SWAT model (CMADS) reanalysis datasets provide an alternative to the use of observed data. This study evaluates the use of CMADS reanalysis datasets in estimating PET across China by the Penman–Monteith equation. PET estimates from CMADS data (PET_cma) during the period 2008–2016 were compared with those from observed data (PET_obs) from 836 weather stations in China. Results show that despite PET_cma overestimating average annual PET and average seasonal in some areas (in comparison to PET_obs), PET_cma well matches PET_obs overall. Overestimation of average annual PET occurs mainly for western inland China. There are more meteorological stations in southeastern China for which PET_cma is a large overestimate, with percentage bias ranging from 15% to 25% for spring but a larger overestimate in the south and underestimate in the north for the winter. Wind speed and solar radiation are the climate variables that contribute most to the error in PET_cma. Wind speed causes PET to be underestimated with percentage bias in the range −15% to −5% for central and western China whereas solar radiation causes PET to be overestimated with percentage bias in the range 15% to 30%. The underestimation of PET due to wind speed is offset by the overestimation due to solar radiation, resulting in a lower overestimation overall.

The soil water supply and atmospheric humidity conditions are crucial in controlling plants’ stomatal behavior and water use efficiency. When there is water stress caused by an increase in saturated water vapor pressure (VPD) and a decrease in soil water content (SWC), plants tend to close stomata to reduce water loss. This affects the gross primary productivity (GPP) and evapotranspiration (ET), subsequently leading to changes in water use efficiency (WUE) and carbon use efficiency (CUE) in plants. However, land–atmosphere interactions mean that water vapor in the atmosphere and soil moisture content causing water stress for plants are closely related. This study aims to compare and estimate the effects of VPD and SWC on the carbon cycle and water cycle for different plant functional types. Based on the fluxnet2015 dataset from around the world, the WUE and CUE of five plant functional types (PFTs) were estimated under varying levels of VPD and SWC. The results showed that high VPD and low SWC limit the stomatal conductance (Gs) and gross primary productivity (GPP) of plants. However, certain types of vegetation (crops, broad-leaved forests) could partially offset the negative effects of high VPD with higher SWC. Notably, higher SWC could even alleviate limitations and partially promote the increase in GPP and net primary production (NPP) with increasing VPD. WUE and CUE were directly affected by Gs and productivity. In general, the increase in VPD in the five PFTs was the dominant factor in changing WUE and CUE. The impact of SWC limitations on CUE was minimal, with an overall impact of only −0.05μmol/μmol on the four PFTs. However, the CUE of savanna plants changed differently from the other four PFTs. The rise in VPD dominated the changes in CUE, and there was an upward trend as SWC declined, indicating that the increase in VPD and decrease in SWC promote the increase in the CUE of savanna plants to some extent.