عنوان مقاله [English]
نویسنده [English]چکیده [English]
Background and objectives: Quantitative prediction of river discharge one of the most important elements in the management of surface water resources, especially take suitable decisions in occurrence of floods and drought events. Various approaches introduced in hydrology to predict river discharge which conceptual models as well as data-driven models are the most important ones.In this study, long term recorded data sets in Araz-Kouseh watershed with 1678 km2 area located in northern Iran (Golestan province) were used to investigate the precision of different river discharge prediction models. The IHACRES model as a conceptual hydrological model and KNN and M5 as data mining models selected for modeling of monthly river discharge and the results were compared to examine the accuracy of studied models. In some studies, the expressed models used for daily river discharge prediction but the main objectives of this study are application of these models to predict monthly discharge for a watershed.
Material and Methods: The 29 years (1985-2013) daily rainfall and discharge data belonging to Araz-Kouseh hydrometry and meteorological stations used to extract monthly time series for modeling. The required quantity and quality conditions of datasets for modeling confirmed using different statistical tests. Recorded datasets divided to two subseries, first one used for calibration period and second one used for validation of investigated models. The results of models in calibration and validation period analyzed considering model efficiency goodness of fit criteria.
Results:The results of IHACRES conceptual hydrological model for both calibration and validation periods (correlation coefficients equal to 0.81 and 0.79 for calibration and validation periods respectively) show suitable ability of this model to predict monthly river discharge. Moreover investigation of results of both M5 and KNN data mining models (correlation coefficient equal to 0.94 and 0.89 for calibration and validation periods respectively for KNN model and equal to 0.92 and 0.88 for calibration and validation periods respectively for M5 model) reveals that application of these models led to significant increase in prediction precision in comparison with IHACRES model.
Conclusions:The results of this study indicate the data mining models, i.e. M5 and KNN, outperform conceptual hydrological model, i.e. IHACRES, for prediction of monthly river discharge considering different goodness of fit criteria. It is clear that the accuracy of the prediction of data mining models are very close to each other but the M5 model is selected as best model in this study because of its explicit equations for prediction. Furthermore, investigation of time series of predicted river discharge show data maining models had better prediction for low discharges in comparison high discharges.