Scholarlink

<p>The prediction of monthly rainfall-runoff time series has a significant influence in planning and developing water resources projects. Thus, in this research, a novel advanced coupled predictive disintegration-optimization-based model is developed to improve the forecasting exactness of the mean monthly river’s runoff. The suggested estimation model is a coupled version of the feature mode decomposition (FMD) algorithm and support vector regression (SVR) model optimized with artificial&#xa0;bee colony (ABC) metaheuristic algorithm, <i>i.e.,</i> hybrid FMD-SVR-ABC model. Its performance is tested on monthly Barandouzchay River’s runoff (<i>BCRR</i><sub><i>m</i></sub>) watershed in Urmia City, West Azerbaijan Province from Sep&#xa0;1971 to Aug 2022. In the FMD-based approaches, the optimal amount of mode number for the rainfall time series measured is 5. Using the partial autocorrelation function (PACF) technique, the number of predictor variables is determined as 9. Comparison plots and performance assessment criteria attest that the recommended model under the optimum predictor and meta-parameters tuned, provides better forecasting results with coefficient of determination (<i>R</i><sup><i>2</i></sup>) of 0.82, root mean square error (<i>RMSE</i>) of 2.67 (m<sup>3</sup>/s), mean bias error (<i>MBE</i>) of 0.22 (m<sup>3</sup>/s), Nash–Sutcliffe efficiency (<i>NSE</i>) of 0.8. Comparatively, the individual SVR model leads to the <i>R</i><sup><i>2</i></sup> of 0.36, <i>RMSE</i> of 5.39 (m<sup>3</sup>/s), <i>MBE</i> of 2.23 (m<sup>3</sup>/s), and <i>NSE</i> of 0.23. Integrating with FMD and ABC algorithms lessens the <i>RMSE</i> value in the single SVR (as the benchmark model) by 27.8% and 15.7%, respectively. Therefore, the suggested hybrid model can be operated as an ingenious, sensible, and precise predictive model for the evaluation of the sequential rainfall-runoff rivers data, mainly the peak flows in different hydro-climatic regimes.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Advancement in forecasting rainfall-runoff process: application of a novel hybrid FMD-SVR-ABC modeling technique

  • 期刊论文

  • 发表日期：

    2026年4月29日

查看全文
  • Redvan Ghasemlounia,
  • Amin Gharehbaghi,
  • Farshad Ahmadi

摘要

The prediction of monthly rainfall-runoff time series has a significant influence in planning and developing water resources projects. Thus, in this research, a novel advanced coupled predictive disintegration-optimization-based model is developed to improve the forecasting exactness of the mean monthly river’s runoff. The suggested estimation model is a coupled version of the feature mode decomposition (FMD) algorithm and support vector regression (SVR) model optimized with artificial bee colony (ABC) metaheuristic algorithm, i.e., hybrid FMD-SVR-ABC model. Its performance is tested on monthly Barandouzchay River’s runoff (BCRRm) watershed in Urmia City, West Azerbaijan Province from Sep 1971 to Aug 2022. In the FMD-based approaches, the optimal amount of mode number for the rainfall time series measured is 5. Using the partial autocorrelation function (PACF) technique, the number of predictor variables is determined as 9. Comparison plots and performance assessment criteria attest that the recommended model under the optimum predictor and meta-parameters tuned, provides better forecasting results with coefficient of determination (R2) of 0.82, root mean square error (RMSE) of 2.67 (m3/s), mean bias error (MBE) of 0.22 (m3/s), Nash–Sutcliffe efficiency (NSE) of 0.8. Comparatively, the individual SVR model leads to the R2 of 0.36, RMSE of 5.39 (m3/s), MBE of 2.23 (m3/s), and NSE of 0.23. Integrating with FMD and ABC algorithms lessens the RMSE value in the single SVR (as the benchmark model) by 27.8% and 15.7%, respectively. Therefore, the suggested hybrid model can be operated as an ingenious, sensible, and precise predictive model for the evaluation of the sequential rainfall-runoff rivers data, mainly the peak flows in different hydro-climatic regimes.