A systematical water allocation scheme for drought mitigation

The severe drought events worldwide have increased the awareness of serious impacts to various social and economic sectors. It is a challenge to make efficient water resources management that optimizes economic and social well-beings under great uncertainty of hydro-meteorology. Artificial intelligence techniques possess an outstanding ability to handle non-linear complex systems. This study proposes a systematical water allocation scheme, which integrates system analysis with artificial intelligence techniques, for decision makers to mitigate drought threats. We first derive evaluation diagrams through a large number of interactive evaluations based on long-term hydrological data to provide a clear perspective of all possible drought conditions and their corresponding water shortages, and then configure neural-fuzzy networks to learning the associations between events and outcomes for estimating water deficiency levels under various hydrological conditions. The adaptive neuro-fuzzy inference system (ANFIS) is adopted to construct the mechanism between designed inputs (water discount rate and the exceedence probabilities of hydrological conditions) and simulated outputs (water deficiency levels). The water allocation in the Shihmen Reservoir watershed of northern Taiwan is used as a case study. The results suggest that the drought thresholds of reservoir storage in the beginning of the first paddy crops can be recommended as: Q50, Q60, Q70 and Q90 for precautionary, preliminary, moderate and severe drought conditions, respectively. The inference system further indicates reservoir storage is identified as the most influential variable that significantly affects water shortage. We demonstrate the proposed water allocation scheme significantly avails water managers of reliably recommending drought thresholds and determining a suitable discount rate on irrigation water supply. This study has direct bearing on more intelligent and effectual water allocation management, which is expected to substantially benefit water managers.