Scholarlink

<p>Accurate prediction of climate variables is fundamental for mitigating the risks of climate change and supporting sustainable management of water, energy, and environmental resources. This study presents a systematic review of 4,276 peer-reviewed studies (2014–2024) covering a spectrum of forecasting methodologies, including classical models, machine learning (ML) techniques, deep learning (DL) approaches, hybrid frameworks, and physics-based general circulation models (GCMs). Classical statistical models such as ARIMA and SARIMA remain widely applied due to their simplicity and interpretability, but they are limited in handling nonlinear dynamics. ML methods (e.g., ANN, SVM, RF) and DL architectures (e.g., LSTM, GRU, CNN) demonstrate superior capability in capturing complex spatiotemporal dependencies, while hybrid models increasingly combine their strengths for enhanced accuracy and robustness. Meanwhile, GCMs, particularly those developed under the CMIP5 and CMIP6 initiatives, provide indispensable long-term projections, though their resolution and regional applicability remain challenging. Our review highlights recent trends, methodological gaps, and the comparative effectiveness of different approaches, emphasizing that the integration of ML/DL with high-resolution GCM outputs, coupled with advances in high-performance computing, will be central to the next generation of climate forecasting.</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

Future climate prediction and projection: A systematic review of classical and advanced methodologies

  • 期刊论文

  • 发表日期：

    2025年9月27日

查看全文
  • Reza Sepaspour,
  • Pouria Hajikarimi,
  • Fereidoon Moghadas Nejad

摘要

Accurate prediction of climate variables is fundamental for mitigating the risks of climate change and supporting sustainable management of water, energy, and environmental resources. This study presents a systematic review of 4,276 peer-reviewed studies (2014–2024) covering a spectrum of forecasting methodologies, including classical models, machine learning (ML) techniques, deep learning (DL) approaches, hybrid frameworks, and physics-based general circulation models (GCMs). Classical statistical models such as ARIMA and SARIMA remain widely applied due to their simplicity and interpretability, but they are limited in handling nonlinear dynamics. ML methods (e.g., ANN, SVM, RF) and DL architectures (e.g., LSTM, GRU, CNN) demonstrate superior capability in capturing complex spatiotemporal dependencies, while hybrid models increasingly combine their strengths for enhanced accuracy and robustness. Meanwhile, GCMs, particularly those developed under the CMIP5 and CMIP6 initiatives, provide indispensable long-term projections, though their resolution and regional applicability remain challenging. Our review highlights recent trends, methodological gaps, and the comparative effectiveness of different approaches, emphasizing that the integration of ML/DL with high-resolution GCM outputs, coupled with advances in high-performance computing, will be central to the next generation of climate forecasting.