In any hydrological processes the flow and the pollutants, specially coming from non-point sources (NPS), are always important for accurate planning and smooth operation of the water resource system. The NPS pollutant such as nitrogen etc. poses a threat to human and environment. A reliable prediction of pollutants and its correlation with past values of river flow and observed non-point pollutants in river system is vital. Present work demonstrates reliable prediction of streamflow and non-point pollutants in river system. Non-point pollution scenario in this work is simulated using fuzzy rule based and ANFIS (Adaptive Neuro fuzzy inference system) models. The FIS, ANFIS are examined using the long-term observations of monthly river flow discharges, total nitrogen loads, total phosphorous loads and sediment loads. The four quantitative standard statistical performance evaluation measures, the coefficient of correlation (R), Nash-Sutcliffe efficiency coefficient (E), root mean squared error (RMSE), mean absolute error (MAE), are employed to evaluate the performances of various models developed. The performance of Fuzzy Ruled Based models and ANFIS models show potential applicability of developed models for flow and nitrogen load predictions. The results of ANFIS models are comparatively better than fuzzy rule based models. Performances of these AI techniques are also compared with statistical methods. Further, data mining techniques are employed to pre-process and to reduce the time series inputs. Results obtained with reduced inputs when compared to the earlier results gives improved efficiency.

Keywords: prediction, streamflow, fuzzy inference system, ANFIS, non-point pollution.