River Quality Zonation using Artificial Neural Network and GIS

The first as well as the foremost phase in water quality management is zonation the quality, because it provides researchers with proper understanding of the processes and the changes of pollution, regarding the specific temporal and particular conditions. The present research has been designed to study the zonation of pollution quality of water in Simineh Rood River using Artificial Neural Networks (ANN) and Geography Information System (GIS). Qualitative data used in this study were collected by sampling in three seasons: spring and autumn of 1387 and spring of 1388. The data were then simulated by using QUAL2K model. The output was then compared with actual data and parameters Nitrate, Dissolved Oxygen, and electrical conductivity related to fall 1387 were utilized target data in neural network models. Data related to land users, geology, erosion capability, and population centers related to Simineh Rood basin were introduced to neural network models as input data. In the present study, five structures of FFBP neural networks model have been used. Results of the best structure have been compared with the output resulted from multivariable regression model, which resulted in proving superiority of the cited neural network model. The result of this research indicated that FEBP network with 4-40-40-3 structure have the best ability. This structure network considers to correlation coefficient in test data, validated data and education data and RMS error in compare to other stricter is preferred. Then used neural network model can be useful in finding the mentioned relation. Generally speaking, the result of this research shows portional ability of neural network in order to river water pollution.