Click on “Download PDF” for the PDF version or on the title for the HTML version.

If you are not an ASABE member or if your employer has not arranged for access to the full-text, Click here for options.


Published by the American Society of Agricultural and Biological Engineers, St. Joseph, Michigan

Citation:  Watershed Management to Meet Water Quality Standards and Emerging TMDL (Total Maximum Daily Load) Proceedings of the Third Conference 5-9 March 2005 (Atlanta, Georgia USA) Publication Date 5 March 2005  701P0105.(doi:10.13031/2013.18099)
Authors:   M. Kim, M. Kim, J. Chung, C. Kim
Keywords:   Artificial Neural Network, Generalized Regression Neural Network, Total Nitrogen, Total Phosphorus, Drainage water, Irrigation water, Rainfall, Time-series prediction

Agricultural surface drainage can deposit nitrogen and phosphorus into surrounding rivers and streams accelerating eutrophication and threatening the ecosystem. Surface drainage from paddy fields and other agricultural lands are influenced by rainfall and other events. The relationship between surface drainage and rainfall is complex and non-linear. Numerous factors impact agricultural surface drainage, including the temporal and spatial distribution of rainfall, land topographic, and soil characteristics. This study used a Generalized Regression Neural Network (GRNN) model, which is one of variants of Radial Basis Function (RBF), to define the influence of rainfall and surface drainage on nutrient load into the neighboring water systems by predicting the surface water quality and quantity. Various network structures and input variables were tested and compared to seek the best performance of a network based on a series of measured water quantity and quality data. The data was obtained from a 15-ha paddy field with drainage and irrigation channels. Simulations showed reasonably good prediction of surface drainage based on historical data of rainfall (R2=0.84). However, the resulting prediction for nutrient concentration corresponding to surface drainage was somewhat varied (R2=0.72 and 0.40 in Total nitrogen and phosphorus, respectively). It is believed that the models poor prediction on nutrient concentration is due to natural and artificial variations of nutrient content in the irrigation streams. Therefore, it is recommended to provide a comprehensive and sufficient representation of both environmental inputs and hydrological process to improve the model performance on nutrient concentration prediction for a further investigation.

(Download PDF)    (Export to EndNotes)