Scholarlink

<p>Climate change poses significant risks to Sarawak, Malaysia, a region renowned for its ecological diversity, where accurate projections of temperature patterns are essential for effective adaptation and resource management. Existing studies often rely on limited global climate models (GCMs) without systematic evaluation, and the high variability in the tropics further complicates model selection. This underscore the need for machine-learning approaches to identify top-ranked GCMs for projecting future temperature changes in Sarawak. Climatic Research Unit (CRU) datasets were used as observational references, and machine learning-based ranking methods, including compromise programming (CP), entropy gain (EG), gain ratio (GR), random forest (RF), and symmetrical uncertainty (SU), were employed to assess and rank GCM performance. Group decision-making (GDM) and an envelope approach were then applied to identify representative GCMs, followed by the development of a RF-based ensemble for robust projection, featuring IPSL-CM5A-MR, IPSL-CM5A-LR, FIO-ESM, HadGEM2-AO, HadGEM2-ES, MIROC-ESM-CHEM, and GISS-E2-R as key contributors for different temperature indices. The results show that the successive Coupled Model Intercomparison Project Phase 6 (CMIP6) models, such as HadGEM3-GC31-LL, HadGEM3-GC31-MM, and FIO-ESM2-0, demonstrate better alignment with observed records compared to CMIP5, with notable improvements in reducing biases and capturing temporal variability. Ensemble projections reveal consistent warming across Sarawak, with CMIP6 generally indicating stronger warming trends than CMIP5, especially during the Southwest Monsoon (SWM). Spatially, the warming is uneven, highlighting persistent thermal contrasts across different parts of the region. These findings underscore the importance of systematic model evaluation and ensemble approaches in improving regional climate projections, providing critical guidance for climate adaptation, water resource management, and biodiversity conservation in Sarawak.</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

Integrating Envelope Approach with Machine Learning Ranking Process for Temperature Projection in Sarawak, Malaysia

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年11月20日

查看全文
  • Zulfaqar Sa’adi,
  • Shamsuddin Shahid,
  • Mohammed Sanusi Shiru,
  • Kamal Ahmed,
  • Mahiuddin Alamgir,
  • Mohamad Rajab Houmsi,
  • Lama Nasrallah Houmsi,
  • Ricky Anak Kemarau,
  • Stanley Anak Suab,
  • Muhamad Azahar Abas,
  • Zainura Zainon Noor

摘要

Climate change poses significant risks to Sarawak, Malaysia, a region renowned for its ecological diversity, where accurate projections of temperature patterns are essential for effective adaptation and resource management. Existing studies often rely on limited global climate models (GCMs) without systematic evaluation, and the high variability in the tropics further complicates model selection. This underscore the need for machine-learning approaches to identify top-ranked GCMs for projecting future temperature changes in Sarawak. Climatic Research Unit (CRU) datasets were used as observational references, and machine learning-based ranking methods, including compromise programming (CP), entropy gain (EG), gain ratio (GR), random forest (RF), and symmetrical uncertainty (SU), were employed to assess and rank GCM performance. Group decision-making (GDM) and an envelope approach were then applied to identify representative GCMs, followed by the development of a RF-based ensemble for robust projection, featuring IPSL-CM5A-MR, IPSL-CM5A-LR, FIO-ESM, HadGEM2-AO, HadGEM2-ES, MIROC-ESM-CHEM, and GISS-E2-R as key contributors for different temperature indices. The results show that the successive Coupled Model Intercomparison Project Phase 6 (CMIP6) models, such as HadGEM3-GC31-LL, HadGEM3-GC31-MM, and FIO-ESM2-0, demonstrate better alignment with observed records compared to CMIP5, with notable improvements in reducing biases and capturing temporal variability. Ensemble projections reveal consistent warming across Sarawak, with CMIP6 generally indicating stronger warming trends than CMIP5, especially during the Southwest Monsoon (SWM). Spatially, the warming is uneven, highlighting persistent thermal contrasts across different parts of the region. These findings underscore the importance of systematic model evaluation and ensemble approaches in improving regional climate projections, providing critical guidance for climate adaptation, water resource management, and biodiversity conservation in Sarawak.