Attribution and Detection of Climate Change Impacts on Extreme Weather

Detection and attribution research has matured into a robust framework for linking human activities to changes in the frequency and intensity of extreme weather. By combining climate model ensembles with observational records, researchers isolate anthropogenic signals from natural variability in heatwaves, heavy precipitation and droughts. Machine-learning approaches, notably convolutional neural networks, have enhanced the detection of emerging fingerprints of warming in daily precipitation and temperature fields. Statistical fingerprinting methods employ regression against modelled responses to greenhouse gases and aerosols, yielding spatial patterns of forced change that become discernible as signal-to-noise ratios improve. Results consistently demonstrate that greenhouse-gas forcing has already amplified heavy rainfall and heat extremes at regional to global scales, while aerosol emissions can offset or enhance these trends seasonally. Advances in explainable artificial intelligence and refined attribution techniques now support more detailed mapping of climate sensitivities, informing impact assessments and resilience planning worldwide.

Research from Nature Portfolio

Recent studies have employed deep-learning algorithms to reveal anthropogenic fingerprints in daily precipitation fields. A convolutional neural network trained on climate-model simulations and observational records detected a clear deviation in daily rainfall variability from natural background since the mid-2010s, with particularly strong signals over the tropical eastern Pacific and mid-latitude storm tracks. Work on regional precipitation in the United States has shown that greenhouse-gas emissions have increased both mean and extreme rainfall across all seasons, while anthropogenic aerosols partly offset these increases in winter and spring but enhance summer and autumn rainfall. Another study applied artificial intelligence to infer spatial sensitivity of extreme precipitation by learning parameters of a generalised extreme value model from thousands of station records, generating high-resolution maps that reveal marked regional variability in how extreme rainfall responds to surface warming.

Attribution and Detection of Climate Change Impacts on Extreme Weather publication trend

The graph below shows the total number of articles in attribution and detection of climate change impacts on extreme weather across all publications each year (not limited to Nature Index journals).

Technical terms

Detection and attribution: The process of identifying a statistically significant change in climate (detection) and determining the relative contributions of human and natural drivers (attribution).

Convolutional neural network: A machine-learning model that extracts spatial features from data fields, here applied to gridded precipitation or temperature observations.

Fingerprint: A characteristic spatial or temporal pattern of climate response to a specific forcing, used to distinguish anthropogenic signals from natural variability.

Generalised extreme value model: A statistical distribution used to describe the behaviour of extreme events, parameterised here as a function of surface temperature changes.

Signal-to-noise ratio: A metric comparing the strength of a forced climate signal to the background variability, determining when anthropogenic impacts become distinguishable.

References

1. Anthropogenic fingerprints in daily precipitation revealed by deep learning. Nature (2023).
2. Anthropogenic aerosols mask increases in US rainfall by greenhouse gases. Nature Communications (2024).
3. Sensitivity of extreme precipitation to climate change inferred using artificial intelligence shows high spatial variability. Communications Earth & Environment (2023).
4. Human-induced intensification of terrestrial water cycle in dry regions of the globe. npj Climate and Atmospheric Science (2024).
5. Robust global detection of forced changes in mean and extreme precipitation despite observational disagreement on the magnitude of change. Earth System Dynamics (2023).
6. Attribution of observed changes in extreme temperatures to anthropogenic forcing using CMIP6 models. Weather and Climate Extremes (2023).