Development of Scale-dependent Water Quality Model to Predict Nutrient Loading.

Modeling non-point pollution across multiple scales has been an issue in the environment. For more representative and practical approach in quantifying and qualifying surface water, Modular Neural Network (MNN) was implemented in this study. Two different site-scales, 1.5 × 10⁵ and 1.62 × 10⁶ square meters (m²), with the same plants, soils, and paddy field management practices were selected. Hydrologic and water quality data including amounts of rainfall, irrigation and surface discharge, and nutrient loadings were continuously monitored throughout the investigated period, and used for the verification of MNN. Correlation coefficients (R) for the resulting predictions from the networks versus measured values were in the range of 0.41 to 0.95. This study found that the extrapolation from small to large fields can be well achievable for rainfall-surface drainage process. Nutrient prediction produced less favorable results due to complex phenomena of nutrients in drainage water, however, this study revealed the feasibility of using MNN for better prediction if more hydrologic and environmental data are provided. The finding of this study further proved that the up-scaling from small-segment plot to large-scale of paddy field can provide accurate and reliable estimation, which contribute to establish the water quality management for sustainable agriculture.