Stochastic optimal control of agrochemical pollutant loads in reservoirs for irrigation

Abstract

Agriculture intensification through the use of chemical nitrogen fertilizers is contributing substantially to the tremendous increases in global food production. However, eutrophication of downstream waters associated with excess nitrate pollutant runoff has become a major environmental concern in agricultural areas. Reservoirs used for irrigation are well-positioned to intercept substantial fertilizer runoff from upslope agricultural fields. The authors hypothesized that such reservoirs can be used to influence the quality of downstream surface water. A Markov decision process was formulated to find optimal operating policies for controlling stochastic agrochemical pollutant loads in reservoirs for irrigation. The model was applied to an irrigation reservoir in Japan that is prone to nitrate pollution from upslope green tea plantations. The computational results reveal that, in all nitrate-nitrogen states in which the reservoir storage state is below 10%, the optimal policy is to introduce cleaner water supplements from the upstream reservoir. When the reservoir storage volume exceeds 35% of active capacity and when the nitrate-nitrogen concentration is less than 3 mg/L, the optimal policy would be only to release the irrigation water. When the reservoir storage volume exceeds 35% of active capacity and when the nitrate-nitrogen concentration exceeds 5 mg/L, the optimal operation decision is to first release irrigation water from the reservoir before introducing cleaner water supplements. Therefore, the operation of irrigation reservoirs can be optimized to control intercepted agro-fertilizer pollutants from intensive agricultural areas. The developed method for reservoir irrigation operation has the potential to support decision-making processes to provide greater protection of the water environment from agrochemical pollution while sustaining the productivity in the intensive agricultural areas.