| Abstrakt: |
The planning and control of water resource systems are meaningfully influenced by streamflow forecasting. Recently, artificial neural network (ANN) techniques have been widely used to forecast the different tasks related to forecasting. This work offers the advancement of an ANN model for long-term stream flow forecasting using machine learning with multilayer perceptron (MLP) and Levenberg algorithms. As a case study of the Goulburn River, 48 years (1974-2022) of monthly lagged streamflow and climate indices such as El Nino Southern Oscillation (ENSO), Interdecadal Pacific Oscillation (IPO), Pacific Decadal Oscillation (PDO), and Indian Ocean Dipole (IOD) are used as input predictors for long-term streamflow forecasting. The performance of the developed ANN model was analyzed with Pearson regression R, Mean Square Error (MSE), Root Mean Square (RMSE), Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE). The developed models using climate indices (ENSO) as an input predictor have shown a close correlation Pearson regression R value ranging from 0.72 to 0.81, with fewer errors in terms of MSE, RMSE, MAE, and MAPE. The results of the developed model showed that the ANN models can predict six-month-ahead streamflow using ENSO climate indices with better accuracy, and one-month-ahead streamflow can be predicted by using streamflow as an input predictor. Moreover, the other indices, i.e., IPO, PDO, and IOD, have shown the random correlation for the selected station. It is observed that the ANN model using climate indices Nin~o3.4 at one lag month is considered a best fit model with fewer errors and a regression R value of 0.81, and the statistical accuracy measures MSE, RMSE, MAE, and MAPE of the Goulburn River model are 62.34 to 73.17, 7.89 to 8.55, 4.47 to 5.20, and 79.25 to 108.00, respectively, in comparison to other developed models, thus proved to be a promising model and can be used to forecast the six months ahead streamflow for the other regions using past ENSO indices data as an input predictor. It is therefore concluded that ENSO indices have a major influence on the streamflow of the Goulburn River and can be used for streamflow forecasting for other streams in the Victoria region. [ABSTRACT FROM AUTHOR] |
