River Flow Forecasting using Remote Sensing data and NARX Artificial Neural Network (Case study: Qarasu River, Ardabil Province)
River Flow Forecasting using Remote Sensing data and NARX Artificial Neural Network (Case study: Qarasu River, Ardabil Province)
Amin Akbari Majd1 Javanshir Azizi Mobasr2 Ali Rasulzadeh3 Mahsa Hasanpour Kashani4
1) Master s student of water resources, department of water science and engineering, University of Mohaghegh Ardabili Email:
2) Associate. Prof., Dept., Range & watershed Management, Faculty of Agriculture and Natural Resources, University of Mohaghegh Ardabili Email:
3) Professor., Dept., Range & watershed Management, Faculty of Agriculture and Natural Resources, University of Mohaghegh Ardabili Email:
4) Assistant. Prof., Dept., Range & watershed Management, Faculty of Agriculture and Natural Resources, University of Mohaghegh Ardabili Email: m.hkashaniuma.ac.ir
IOI Code
:
XGKG-KCCCC
Publication :
Organic vs Conventional Agriculture(umaorganic.ir)
Abstract :
River flow forecasting is very important for planning and managing water resources, especially in dry areas where water resources are limited. In recent years, ANN methods have been widely used in a wide range of engineering applications in the field of hydrology. One of these models is the NARX model. This model is a self-correlated time series model that can model nonlinear patterns well in a feedforward neural network system. In this study, the second data with a period of 10 years was used; The first series is related to the observational data at the selected hydrometric station and the second series is the data extracted through remote sensing. The main variables that were extracted from remote sensing databases and by the GEE platform with a daily time step are: vegetation cover index (low and high), volume of water in the soil, runoff and precipitation. Then the final model was built using the NARX model with a 3-day delay and 15 intermediate layers. The results obtained by evaluating by r, RMSR and NSE statistics showed that the accuracy of the model in the training phase (first 5 years) was equal to 0.94, 0.035 and 0.887 respectively and the accuracy of the model in the Test (second 5 years) was equal to 0.89, 0.057 and 0.787 respectively. According to these results and comparison with similar researches, it was shown that the NARX model has a good and acceptable ability in predicting the river flow when used with remote sensing data.
Keywords : forecasting river flow remote sensing artificial neural network gee narx