Scholarlink

As climate change become more tangible in recent decades, assessing, simulating, and predicting the impact of climate change on the surrounding environment is vital for policymakers and sustainable development. National Centres for Environmental Prediction (NCEP) are globally gridded data sets representing the Earth’s atmosphere, incorporating observations and numerical weather prediction output the high resolution of the NCEP data needs downscaling to make the relationship between predictors and predictands. In this study, precipitation and mean temperature of Nicosia are simulated using Convolutional-Long Short-Term Memory (CNN-LSTM) and classical Feed Forward Neural Network (FFNN) downscaling models. The CNN-LSTM model can get the power of feature extraction from the Convolutional Neural Network (CNN) and predictive power from the LSTM network. Furthermore, in this research, Quantile Mapping (QM)-based bias correction was employed to remove systematic biases, and Mutual Information (MI) feature selection was used as a pre-processing step to the classical FFNN model. Then, the models were evaluated based on Root Mean Squared Error (RMSE) and Coefficient of Determination (DC). It was concluded that the CNN-LSTM had superiority over classical FFNN due to the deep-based inheritance.

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

Simulating the Precipitation and Temperature of Nicosia via a Hybrid CNN-LSTM Model

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Vahid Nourani,
  • Hadi Pourali,
  • Mohammad Bejani,
  • Aida Hosseini Baghanam

摘要

As climate change become more tangible in recent decades, assessing, simulating, and predicting the impact of climate change on the surrounding environment is vital for policymakers and sustainable development. National Centres for Environmental Prediction (NCEP) are globally gridded data sets representing the Earth’s atmosphere, incorporating observations and numerical weather prediction output the high resolution of the NCEP data needs downscaling to make the relationship between predictors and predictands. In this study, precipitation and mean temperature of Nicosia are simulated using Convolutional-Long Short-Term Memory (CNN-LSTM) and classical Feed Forward Neural Network (FFNN) downscaling models. The CNN-LSTM model can get the power of feature extraction from the Convolutional Neural Network (CNN) and predictive power from the LSTM network. Furthermore, in this research, Quantile Mapping (QM)-based bias correction was employed to remove systematic biases, and Mutual Information (MI) feature selection was used as a pre-processing step to the classical FFNN model. Then, the models were evaluated based on Root Mean Squared Error (RMSE) and Coefficient of Determination (DC). It was concluded that the CNN-LSTM had superiority over classical FFNN due to the deep-based inheritance.