Analyzing variation of inflow from the Syr Darya to the Aral Sea: A Bayesian-neural-network-based factorial analysis method

In this study, a Bayesian-neural-network-based factorial analysis (abbreviated as BNN-FA) method is developed for quantifying the effects of multiple factors on inflow from the Syr Darya to the Aral Sea (abbreviated as ISA). BNN-FA cannot only reflect the complex relationship between inputs and outputs, but also reveal the individual and interactive effects of multiple factors. BNN-FA is applied to the downstream of the Syr Darya river basin, where effects of human-activity, hydrological, and ecological factors are analyzed. Results reveal that (i) during 1960–1991, the main factors affecting ISA are: agricultural water consumption of Uzbekistan (31.2%) > agricultural water consumption of Kazakhstan (23.8%) > agricultural water consumption of Tajikistan (14.3%) > reservoir water storage (10.6%) > evapotranspiration (7.5%); interactions among agricultural water consumptions of the three countries (i.e. Kazakhstan, Uzbekistan, and Tajikistan) and interactions between evapotranspiration and agricultural water consumptions have noticeable effects on ISA; (ii) during 1992–2015, the main factors are: agricultural water consumption of Uzbekistan (21.4%) > upstream inflow (19.0%) > agricultural water consumption of Kazakhstan (16.0%) > agricultural water consumption of Tajikistan (14.2%) > industrial water consumption of Uzbekistan (9.4%); interactions among agricultural water consumptions of the three countries and interactions between upstream inflow and agricultural water consumptions have important effects on ISA; (iii) the contribution of agricultural activity decreases from 69.3% in 1960–1991 to 51.6% in 1992–2015. The findings are helpful for decision makers to formulate effective strategies to increase the runoff of the Syr Darya and restore the eco-environment of the Aral Sea.