AI reveals hidden climate extremes in Europe

With over 30,000 weather stations worldwide collecting daily data on temperature, precipitation, and other climate indicators, analyzing this immense dataset is a monumental task for researchers. 

Traditionally, climate scientists have relied on statistical methods to interpret these datasets, but a recent breakthrough demonstrates the power of artificial intelligence (AI) to revolutionize this process.

Previously unrecorded climate extremes

A team led by Étienne Plésiat of the German Climate Computing Center in Hamburg, alongside colleagues from the UK and Spain, applied AI to reconstruct European climate extremes.

The research not only confirmed known climate trends but also revealed previously unrecorded extreme events.

As climate change accelerates, understanding how temperature and precipitation extremes are evolving is critical for adaptation and planning. 

Recent studies have shown that heat and rainfall extremes are increasing dramatically. For instance, heavy rainfall events in some regions are now classified as “far outside the historical climate,” while heat extremes have expanded, affecting over 30% of global land areas annually compared to just 1% in 1950.

AI offers a solution to data gaps

Despite the increasing prevalence of climate extremes, analyzing historical temperature data is challenging due to gaps in records. 

Weather stations – particularly in the early 20th century – often faced interruptions, whether due to damaged equipment, station closures, or the lack of replacements. 

Sparse data from regions like Africa and the polar areas further complicate comprehensive analysis.

Data homogenization, a research area focused on standardizing and filling gaps in datasets, has long been used to address these issues. 

However, traditional statistical methods often struggle in areas with scarce weather stations, leaving uncertainties in extreme climate reconstructions.

Recognizing the limitations of traditional methods, Plésiat and his colleagues turned to AI, specifically neural network techniques, to reconstruct temperature extremes. 

Climate extremes in Europe 

Europe, with its dense network of weather stations and long historical records – such as the Hadley Central England Temperature dataset, which dates back to 1659 – offered an ideal testing ground.

The team’s AI, dubbed CRAI (Climate Reconstruction AI), focused on extreme events such as unusually warm or cold days and nights.

CRAI outperformed traditional interpolation methods like Kriging and Inverse Distance Weighting, which predict temperatures at unmeasured locations using nearby weather station data. 

These traditional methods often falter when station density is low, a gap that CRAI bridged effectively.

Reconstructing climate history

Using historical simulations from the CMIP6 archive (Coupled Model Intercomparison Project), the team trained CRAI to reconstruct past climate data. 

The experts validated their results using standard metrics such as root mean square error and Spearman’s rank-order correlation coefficient, which measure accuracy and association between variables.

The results were promising. CRAI outperformed conventional methods in reconstructing the frequency of extreme events—such as the percentage of days with temperatures above the 90th percentile (warm days) or below the 10th percentile (cool days).

The researchers then applied CRAI to the HadEX3 dataset, which includes over 80 indices of extreme temperature and precipitation from 1901 to 2018. CRAI not only reconstructed past extremes with high accuracy but also uncovered long-forgotten events.

Unearthing unknown climate extremes

One of the most striking findings was CRAI’s ability to reveal previously unrecorded climate extremes. For instance, the AI uncovered a severe cold spell in 1929 and a major heatwave in 1911, events that were only hinted at anecdotally due to sparse data at the time.

“Our research demonstrates both the necessity and the potential benefits of applying this approach to the global scale or other regions with scarce data,” the authors explained. 

They emphasized that AI-based reconstruction enhances accuracy, particularly in data-poor regions, and opens new possibilities for climate research.

The potential of AI in climate science

The study highlights AI’s transformative potential to improve our understanding of climate extremes and their long-term changes. 

By training models like CRAI on larger datasets, researchers can achieve even greater accuracy and uncover hidden patterns in global climate history.

“This work underscores the transformative potential of AI to improve our understanding of climate extremes and their long-term changes,” the researchers concluded.

As climate change continues to intensify, AI tools like CRAI offer a powerful means to analyze past data, predict future trends, and guide adaptation strategies in an increasingly extreme world.

The study is published in the journal Nature Communications.

—–

Like what you read? Subscribe to our newsletter for engaging articles, exclusive content, and the latest updates.

Check us out on EarthSnap, a free app brought to you by Eric Ralls and Earth.com.

—–