Method for predicting nitrogen concentration in water on lower Mississippi River in USA

Yan Baowen;Mark D.Tomer;Wen Deping;College of Water Resources and Architectural Engineering,Northwest A&F University;National Laboratory for Agriculture and the Environment;Water and Hydrologic Survey Bureau of Qinghai Province;

nitrate nitrogen forecast;;Mississippi River;;concentration of nitrate nitrogen;;artificial neural network;;Loadruner procedure

In order to investigate pollution and eutrophication caused from agricultural nonpoint source pollution in river water,it is essential to predict the relation between the concentration of nitrate nitrogen in a river and the runoff for such an ion-dominated pollutant. Hence,the lower Mississippi river with intensive farming land in USA was chosen as a model and the nitrate nitrogen concentration,runoff water quality data collected from Vicksburg Hydrological Station on the river were analyzed. Then the whole runoff data set was separated into daily and monthly individual data sets by using Baseflow program,furthermore,the individual runoff data sets were extended and complemented by using Loadrunner Program of Yale University to form a continuously daily nitrate nitrogen concentration series; eventually,the monthly concentration sequences were established. The monthly nitrate nitrogen concentration in the water body was predicted by means of neural network method,as a result,the corresponding procedure and predication formulas were proposed. The results showed that the average error between predicted and observed concentrations is 7. 5%,implying the procedure and formulas proposed are accurate and feasible. It suggests that this method can be applied to predict monthly nitrate nitrogen concentration in a real river.