Modeling nitrate leaching using neural networks

Accurate evaluation of nitrate leaching potential in agricultural fields is a major challenge. Field data are expensive to gather and use of existing prediction models is limited by inadequate understanding of the physical and chemical processes underlying nitrate leaching. A neural network model was developed to acquire the inherent characteristics of an experimental data set, and successfully used to simulate nitrate leaching in agricultural drainage effluent under various management systems. Simulation results indicated that: (i) sub-irrigation with a 0.5 m water table depth could reduce nitrate leaching to negligible levels, (ii) intercropping corn with ryegrass could reduce nitrate leaching by 50%, and (iii) the application of more than 180 kg N ha⁻¹ of fertilizer may cause excessive nitrate leaching.