Journal of Advanced Engineering Technology and Management
ISSN (Online): 3049-3684
Volume: 1 Issue: 1 | Open Access | 09 April 2025
A predictive model for climate change using advanced machine learning algorithms
Vishnu Mathur, Student, RVR & JC College of Engineering
Abstract
We present CLIM-ML, a hybrid machine-learning framework combining graph neural networks (GNNs), spatio-temporal transformers, and physics-informed modules to predict regional and global climate variables (surface temperature, precipitation, and extreme indices) under multiple greenhouse-gas forcing scenarios. CLIM-ML is trained on reanalysis (ERA5), satellite-derived observational products, and CMIP6 simulations and evaluated using standard climate metrics and benchmark datasets. Our approach improves short-to-decadal predictive skill relative to baseline statistical and purely physical emulators while providing interpretable feature attributions for decision-makers. The design is informed by prior work showing deep learning’s promise in Earth system science and recent breakthroughs in AI-driven weather/climate forecasting.
Keywords: Machine Learning, Climate Change, ERAS, Machine Learning Algorithms
References
Reichstein, M. et al. Deep learning and process understanding for data-driven Earth system science. Nature (2019).
Rasp, S., Pritchard, M. S., & Gentine, P. WeatherBench: A benchmark data set for data-driven weather forecasting. (2020).
IPCC. Climate Change 2021: The Physical Science Basis. (AR6 WGI).
Survey: Machine Learning Methods in Climate Prediction: A Survey (2023).
Recent news and peer-reviewed advances demonstrating AI performance improvements in weather forecasting (GraphCast, NeuralGCM) and hybrid models.
Yang, R. et al. Interpretable machine learning for weather and climate (2024 review).