Ensemble machine learning (EML) based regional flood frequency analysis model development and testing for south-east Australia

IF 4.7 2区地球科学 Q1 WATER RESOURCES

Journal of Hydrology-Regional Studies Pub Date : 2025-03-19 DOI:10.1016/j.ejrh.2025.102320

Nilufa Afrin , Ataur Rahman , Ahmad Sharafati , Farhad Ahamed , Khaled Haddad

{"title":"Ensemble machine learning (EML) based regional flood frequency analysis model development and testing for south-east Australia","authors":"Nilufa Afrin , Ataur Rahman , Ahmad Sharafati , Farhad Ahamed , Khaled Haddad","doi":"10.1016/j.ejrh.2025.102320","DOIUrl":null,"url":null,"abstract":"<div><h3>Study region</h3><div>South-east Australia</div></div><div><h3>Study focus</h3><div>This study develops two ensemble-based regional flood frequency analysis (RFFA) techniques, Random Forest Regression (RFR) and Gradient Boosting Regression (GBR)) with a standalone method (Artificial Neural Network (ANN), for south-east Australia. A dataset from 201 catchments across south-east Australia is used in this study. It includes six Annual Exceedance Probabilities (AEPs), 1 in 2, 1 in 5, 1 in 10, 1 in 20, 1 in 50, and 1 in 100 to estimate design floods, which are widely used in the planning and design of water infrastructure. An independent test is adopted to compare the performance of the selected RFFA techniques.</div></div><div><h3>New hydrological insights for the region</h3><div>This study employs a random forest (RF) algorithm as a nonlinear feature selection method to select the important features/catchment characteristics (predictors) in the RFFA. Out of the eight candidate predictors, three are selected to develop and test the selected RFFA techniques. The findings indicate that ensemble methods (RFR and GBR) provide better performance than the standalone ANN technique. The median relative error values are found to be in the range of 33–44 % for the RFR, 34–46 % for the GBR, and 35–53 % for the ANN. The results of this study would be useful in upgrading RFFA methods in the Australian Rainfall and Runoff (national guideline).</div></div>","PeriodicalId":48620,"journal":{"name":"Journal of Hydrology-Regional Studies","volume":"59 ","pages":"Article 102320"},"PeriodicalIF":4.7000,"publicationDate":"2025-03-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Journal of Hydrology-Regional Studies","FirstCategoryId":"89","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S2214581825001442","RegionNum":2,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"WATER RESOURCES","Score":null,"Total":0}

引用次数: 0

Abstract

Study region

South-east Australia

Study focus

This study develops two ensemble-based regional flood frequency analysis (RFFA) techniques, Random Forest Regression (RFR) and Gradient Boosting Regression (GBR)) with a standalone method (Artificial Neural Network (ANN), for south-east Australia. A dataset from 201 catchments across south-east Australia is used in this study. It includes six Annual Exceedance Probabilities (AEPs), 1 in 2, 1 in 5, 1 in 10, 1 in 20, 1 in 50, and 1 in 100 to estimate design floods, which are widely used in the planning and design of water infrastructure. An independent test is adopted to compare the performance of the selected RFFA techniques.

New hydrological insights for the region

This study employs a random forest (RF) algorithm as a nonlinear feature selection method to select the important features/catchment characteristics (predictors) in the RFFA. Out of the eight candidate predictors, three are selected to develop and test the selected RFFA techniques. The findings indicate that ensemble methods (RFR and GBR) provide better performance than the standalone ANN technique. The median relative error values are found to be in the range of 33–44 % for the RFR, 34–46 % for the GBR, and 35–53 % for the ANN. The results of this study would be useful in upgrading RFFA methods in the Australian Rainfall and Runoff (national guideline).

查看原文本刊更多论文

求助全文

约1分钟内获得全文求助全文

来源期刊

Journal of Hydrology-Regional Studies Earth and Planetary Sciences-Earth and Planetary Sciences (miscellaneous)

CiteScore

6.70

自引率

8.50%

发文量

284

审稿时长

60 days

期刊介绍： Journal of Hydrology: Regional Studies publishes original research papers enhancing the science of hydrology and aiming at region-specific problems, past and future conditions, analysis, review and solutions. The journal particularly welcomes research papers that deliver new insights into region-specific hydrological processes and responses to changing conditions, as well as contributions that incorporate interdisciplinarity and translational science.