Scholarlink

<p>Multi-model ensembles are commonly employed to improve the performance of General circulation models (GCMs) projections. The design and practical implementation of ensemble techniques over the Tibetan Plateau (TP) are not yet well understood. In this study, we assess the capability of various ensemble methods to simulate climate extremes over the TP, using 20 GCMs from the Coupled Model Intercomparison Project Phase 6, including the arithmetic mean (AM), Bayesian model averaging (BMA), support vector machine (SVM), random forest (RF), artificial neural networks (ANN), and long short-term memory (LSTM) models. Subsequently, climate extremes over the TP under global warming scenarios of 1.5&#xa0;°C, 2&#xa0;°C, and 3&#xa0;°C above pre-industrial levels are projected using the most effective ensemble method. The results show that SVM, RF, ANN, and LSTM outperform the AM and BMA approaches. Among these, RF performs best, demonstrating superior performance in capturing the spatial distribution of temperature and precipitation indices. Most climate extreme indices, except for CDD, TN10p, TX10p, FD, ID, and CSDI, exhibit an increasing trend at global warming levels of 1.5&#xa0;°C, 2&#xa0;°C, and 3&#xa0;°C, with more pronounced changes occurring at higher warming levels. In terms of geographic distribution, precipitation and temperature extreme indices exhibit substantial spatial variability. This study suggests that machine learning techniques offer a novel perspective for extracting deeper insights from large datasets and can enhance the accuracy and reliability of climate projections.</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

Enhancing multi-model ensemble simulations of climate extremes over the Tibetan Plateau using machine learning

  • 期刊论文

  • 发表日期：

    2025年6月27日

查看全文
  • Tong Cui,
  • JiaZhong Zheng,
  • Zupeng Zhang,
  • Yuping Han,
  • Dongjing Huang,
  • Liyan Yang,
  • Shuai Chen

摘要

Multi-model ensembles are commonly employed to improve the performance of General circulation models (GCMs) projections. The design and practical implementation of ensemble techniques over the Tibetan Plateau (TP) are not yet well understood. In this study, we assess the capability of various ensemble methods to simulate climate extremes over the TP, using 20 GCMs from the Coupled Model Intercomparison Project Phase 6, including the arithmetic mean (AM), Bayesian model averaging (BMA), support vector machine (SVM), random forest (RF), artificial neural networks (ANN), and long short-term memory (LSTM) models. Subsequently, climate extremes over the TP under global warming scenarios of 1.5 °C, 2 °C, and 3 °C above pre-industrial levels are projected using the most effective ensemble method. The results show that SVM, RF, ANN, and LSTM outperform the AM and BMA approaches. Among these, RF performs best, demonstrating superior performance in capturing the spatial distribution of temperature and precipitation indices. Most climate extreme indices, except for CDD, TN10p, TX10p, FD, ID, and CSDI, exhibit an increasing trend at global warming levels of 1.5 °C, 2 °C, and 3 °C, with more pronounced changes occurring at higher warming levels. In terms of geographic distribution, precipitation and temperature extreme indices exhibit substantial spatial variability. This study suggests that machine learning techniques offer a novel perspective for extracting deeper insights from large datasets and can enhance the accuracy and reliability of climate projections.