Monday, November 13, 2006

The Coupling of Site-scale on the Prediction of Agricultural Water Quality.

Minyoung Kim1, Min-Kyeong Kim2, Nam-Jong Lee2, Kee-An Roh2, and Mun-Hwan Koh2. (1) USDA, Agricultural Research Service, 120 Keim Hall, East Campus, Univ of Nebraska, Lincoln, NE 68583-0934, (2) National Institute of Agricultural Science, RDA, 249 Seosundong Kwonsungu, Suwon, South Korea

The state of the art knowledge of the behavior of excessive nutrients enables us to establish reliable assessment of risks and the design of efficient management practices to mitigate this problem. Therefore, it is critical to develop a modeling technique to predict the accurate and reliable water quality. The implementation of Artificial Neural Network (ANN) has been increased in Korea due to its unique agricultural characteristics, small-scale and complex system. Therefore, this study attempted to investigate Modular Neural Network (MNN), one of ANN algorithms, can be further implemented as a forecasting tool for nutrient loadings. Two different site-scales, but same plants, soils, and even management practices of agricultural paddy fields, 15 ha and 162 ha, were applied and compared. Hydrologic and water quality data including rainfall, irrigation and surface discharge amounts, and nutrient loadings (total nitrogen and total phosphorus) were continuously monitored throughout the investigated period, and used for the verification of MNN. A well-trained MNN using data from a small-scale paddy field (15 ha) was tested for the larger-scale paddy field (162 ha). Correlation coefficients (R) for the resulting predictions from the networks versus measured values were generally in the range of 0.41 to 0.95, more closely to present that 0.948 (surface discharge) generated by rainfall and 0.524 (total nitrogen) and 0.408 (total phosphorus) corresponding to runoff events. The practical implication in this study showed that MNN technique can achieve superior performance in predicting rainfall-runoff process, and also relatively acceptable predictions in small and even large-scale agricultural paddy fields in the case of water quality forecasting.