Inland river runoff variations in arid regions play a decisive role in maintaining regional ecological stability. Observation data of inland river runoff in arid regions have short time series and imperfect attributes due to limitations in the terrain environment and other factors. These shortages not only restrict the accurate simulation of inland river runoff in arid regions significantly, but also influence scientific evaluation and management of the water resources of a basin in arid regions. In recent years, research and applications of machine learning and in-depth learning technologies in the hydrological field have been developing gradually around the world. However, the simulation accuracy is low, and it often has over-fitting phenomenon in previous studies due to influences of complicated characteristics such as “unsteady runoff”. Fortunately, the circulation layer of Long-Short Term Memory (LSTM) can explore time series information of runoffs deeply to avoid long-term dependence problems. In this study, the LSTM algorithm was introduced and improved based on the in-depth learning theory of artificial intelligence and relevant meteorological factors that were monitored by coupling runoffs. The runoff data of the Yarkant River was chosen for training and test of the LSTM model. The results demonstrated that Mean Absolute Error (MAE) and Root Mean Square error (RMSE) of the LSTM model were 3.633 and 7.337, respectively. This indicates that the prediction effect and accuracy of the LSTM model were significantly better than those of the convolution neural network (CNN), Decision Tree Regressor (DTR) and Random Forest (RF). Comparison of accuracy of different models made the research reliable. Hence, time series data was converted into a problem of supervised learning through LSTM in the present study. The improved LSTM model solved prediction difficulties in runoff data to some extent and it applied to hydrological simulation in arid regions under several climate scenarios. It not only decreased runoff prediction uncertainty brought by heterogeneity of climate models and increased inland river runoff prediction accuracy in arid regions, but also provided references to basin water resource management in arid regions. In particular, the LSTM model provides an effective solution to runoff simulation in regions with limited data.

Climate change and anthropogenic activities drive the shrinkage of terminal lakes in arid areas to varying degrees. Ecological water conveyance (EWC) projects have emerged globally to restore the ecology of terminal lakes. However, there remains a lack of qualitative evaluation of the benefits of EWC on terminal lakes. This study compared the Taitema Lake Basin with the Aral Sea Basin in Central Asia, representative of terminal lake basins with and without EWC, respectively. The results show that the water area of Taitema Lake increased by 7.23 km²/year due to EWC (2000–2019), whereas that of the Aral Sea Basin decreased by 98.21% over the entire process of natural evolution (1972–2019). Land use changes before and after the EWC (1990–2019) included an increase and decrease in desert land and water bodies in the Aral Sea Basin, and a decrease and increase in desert land and arable land in the Tarim River Basin, respectively. The normalized difference vegetation index (NDVI) and actual evaporation (ETa) are the main factors influencing the change in the water area of the Aral Sea Basin with the changing environment, while EWC is the main factor influencing the change in the water area of Taitema Lake. The results confirm that EWC is a feasible measure for achieving ecological restoration of a terminal lake watershed in an arid area.

The uncertainties in soil erosion (SE) are further intensified by various factors, such as global warming, regional warming and humidification, and vegetation cover changes. Moreover, quantitative evaluations of SE in major basins of Central Asia (CA) under changing environments have rarely been conducted. This study conducted quantitative evaluation of SE in four major basins (Syr Darya Basin (SDB), Amu Darya Basin (ADB), Ili River Basin (IRB) and Tarim River Basin (TRB) using the Revised Universal Soil Loss Equation (RUSLE) and analyzed the main driving factors. SE quantities in the basins presented relatively consistent upward fluctuating trends from 1982 to 2017. Vegetation cover variation fluctuated significantly from 1982 to 2017. Specifically, vegetation cover decreased continuously in SDB, ADB, and IRB, but increased gradually in TRB. Pixels with positive spatial variation of vegetation mainly occurred around lakes and oases near rivers. The Normalized Difference Vegetation Index (NDVI) showed higher correlation with precipitation (80.5%) than with temperature (48.3%). During the study period, the area of arable land (AL) exhibited the largest change among all land use types in CA. Under long-term human activities, the proportion of NDVI of other land types converting to AL was the highest. In the structural equation model (SEM), precipitation, temperature, Shannon Diversity Index (SHDI), and NDVI strongly influenced SE. Overall, the major basins in CA were jointly affected by climate, human activities, and vegetation. Specifically, climatic factors exerted the strongest influence, followed by SHDI (human activities). SE was found to be relatively serious in ADB, SDB, and IRB, with SE in SDB even approaching that in the Loess Plateau. Under the background of global changes, appropriate water and land resource management and optimization configurations should be implemented in CA with reference to TRB in order to relieve local SE problems.