Scholarlink

<p>Due to climate change, the Majalaya Basin in West Java is becoming increasingly susceptible to flooding from irregular rainfall patterns, emphasizing the urgent need for accurate rainfall projections to better understand and mitigate flood risks in the region. A multi-step comparative and ranking approach was utilised in this study. A comparative assessment showed that Support Vector Machine (SVM) outperforms K-nearest neighbors (KNN), Random Forest (RF), and Extreme Gradient Boosting (XGBoost) in bias correcting the Climate Hazards Group InfraRed Precipitation with Station data (CHIRPS) dataset, achieving the lowest bias (-13.26 to 65.49), smallest root mean square error (RMSE) (68.95 to 139.70), and highest coefficient of determination (R²) values (0.582 to 0.723), resulting in an improved observed dataset simulating local monthly rainfall from 1981 to 2014 across 20 grid points at a 0.05-degree resolution. The ranking of 23 Coupled Model Intercomparison Project Phase 6 (CMIP6) rainfall global climate models (GCMs) was conducted based on their temporal reliability in simulating CHIRPS monthly rainfall. Model performance across 20 spatial grids was evaluated using seven relative importance metrics (RIMs). To consolidate the rankings across all grids, the compromise programming index (CPI) was calculated, and the results were further categorized using Jenks optimized classification (JOC). This multi-step ranking approach consistently identified EC-Earth3-Veg-LR as the top-performing GCM, supported by the lowest CPI value of 0.02918, confirming its superior performance among all 23 GCMs evaluated. The gamma quantile mapping (GQM) method, incorporating bias-corrected constructed analogues (BCCA), outperformed generalized quantile mapping (GEQM), linear scaling (LS), and power transformation (PT), achieving the highest Nash-Sutcliffe efficiency (NSE) of 0.221, the lowest RMSE of 133.90&#xa0;mm, and the highest R<sup>2</sup> of 0.559, reflecting its strong performance in correcting both the magnitude and variability of rainfall. Projected changes in mean monthly rainfall indicate significant reductions across the monsoons and transition periods, with the West monsoon showing a decrease, particularly in December, and the East monsoon expected to experience a more severe decrease. These projections align with a long-term trend of decreasing rainfall and increasing drought conditions, influenced by rising temperatures and shifting monsoonal patterns, particularly in the northern and central areas of the basin. This basin-scale study contributes to improving climate change modelling by refining the accuracy of bias correction and GCM selection, thereby enabling more reliable future rainfall projections to support targeted flood risk management and informed local planning efforts.</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

Assessing the temporal reliability of CMIP6 GCMs in projecting future rainfall for the Majalaya Basin, West Java

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年8月18日

查看全文
  • Zulfaqar Sa’adi,
  • Faizal I. W. Rohmat,
  • Ioanna Stamataki,
  • Shamsuddin Shahid,
  • Zafar Iqbal,
  • Zaher Mundher Yaseen,
  • Nor Eliza Alias,
  • Zulkifli Yusop,
  • Zainura Zainon Noor,
  • Ricky Anak Kemarau,
  • Mohammed Sanusi Shiru

摘要

Due to climate change, the Majalaya Basin in West Java is becoming increasingly susceptible to flooding from irregular rainfall patterns, emphasizing the urgent need for accurate rainfall projections to better understand and mitigate flood risks in the region. A multi-step comparative and ranking approach was utilised in this study. A comparative assessment showed that Support Vector Machine (SVM) outperforms K-nearest neighbors (KNN), Random Forest (RF), and Extreme Gradient Boosting (XGBoost) in bias correcting the Climate Hazards Group InfraRed Precipitation with Station data (CHIRPS) dataset, achieving the lowest bias (-13.26 to 65.49), smallest root mean square error (RMSE) (68.95 to 139.70), and highest coefficient of determination (R²) values (0.582 to 0.723), resulting in an improved observed dataset simulating local monthly rainfall from 1981 to 2014 across 20 grid points at a 0.05-degree resolution. The ranking of 23 Coupled Model Intercomparison Project Phase 6 (CMIP6) rainfall global climate models (GCMs) was conducted based on their temporal reliability in simulating CHIRPS monthly rainfall. Model performance across 20 spatial grids was evaluated using seven relative importance metrics (RIMs). To consolidate the rankings across all grids, the compromise programming index (CPI) was calculated, and the results were further categorized using Jenks optimized classification (JOC). This multi-step ranking approach consistently identified EC-Earth3-Veg-LR as the top-performing GCM, supported by the lowest CPI value of 0.02918, confirming its superior performance among all 23 GCMs evaluated. The gamma quantile mapping (GQM) method, incorporating bias-corrected constructed analogues (BCCA), outperformed generalized quantile mapping (GEQM), linear scaling (LS), and power transformation (PT), achieving the highest Nash-Sutcliffe efficiency (NSE) of 0.221, the lowest RMSE of 133.90 mm, and the highest R2 of 0.559, reflecting its strong performance in correcting both the magnitude and variability of rainfall. Projected changes in mean monthly rainfall indicate significant reductions across the monsoons and transition periods, with the West monsoon showing a decrease, particularly in December, and the East monsoon expected to experience a more severe decrease. These projections align with a long-term trend of decreasing rainfall and increasing drought conditions, influenced by rising temperatures and shifting monsoonal patterns, particularly in the northern and central areas of the basin. This basin-scale study contributes to improving climate change modelling by refining the accuracy of bias correction and GCM selection, thereby enabling more reliable future rainfall projections to support targeted flood risk management and informed local planning efforts.