Rainfall (precipitation) prediction remains challenging in a warming world primarily because of uncertainties in how large-scale atmospheric circulation patterns will shift, even as basic thermodynamic (moisture-related) changes are more predictable.

A major 2026 study led by the University of Oxford and ETH Zurich (published in Nature) analyzed Northern Hemisphere winter rainfall (1950–2022) and separated these influences. Models capture thermodynamic changes well but significantly underrepresent circulation-driven trends—for example, simulating only about 10% of the observed circulation-driven rainfall trend in Southern Europe. This gap limits confidence in regional forecasts of floods and droughts.

Regional details (e.g., exact shifts in monsoon strength, mid-latitude storm tracks, or subtropical drying) carry lower confidence, especially for specific locations and seasons. This is why IPCC reports often show wider uncertainty ranges or hatching (low model agreement) for precipitation projections compared to temperature.

_____________________________________________________________________________________

Uncertain dynamic response of mid-latitude winter precipitation

“Uncertain dynamic response of mid-latitude winter precipitation” is the title of a peer-reviewed paper published in Nature on April 29, 2026, led by Lei Gu (University of Oxford, formerly ETH Zurich) along with co-authors including Dominik L. Schumacher, Erich M. Fischer, and Reto Knutti.

Understanding changes in precipitation is crucial for society and ecosystems. Previous studies have examined contributions from anthropogenic forcing versus internal variability, but discrepancies remain between observed and simulated patterns. In Northern Hemisphere winter, mismatches are often blamed on unforced internal variability. However, evidence suggests models may underestimate the total precipitation response to human forcings.

The study shows that thermodynamic contributions (related to moisture and temperature changes) are broadly reproduced by climate models. In contrast, dynamic contributions (shifts in large-scale atmospheric circulation) diverge more substantially. The authors disentangle forced thermodynamic and dynamic components from internal variability in winter precipitation trends (1950–2022).

Key finding in the Mediterranean:

The forced dynamic signal in model simulations explains only about 10% of the observed dynamic trend, making detection challenging. Under continued emissions, projected circulation responses intensify and increasingly resemble observed patterns—though internal variability may still dominate observations. Improving the representation of forced large-scale circulation changes in models is essential for better regional projections.

Core Approach and Methods

• Decomposition: Researchers separated precipitation trends into:
  • Thermodynamic: Moisture-related effects from warming (e.g., higher atmospheric water-holding capacity).
  • Dynamic: Changes in circulation patterns (e.g., jet stream, storm tracks, pressure systems like the North Atlantic Oscillation) that steer where moisture is delivered.
  • Internal variability: Natural fluctuations unrelated to long-term forcing.
• They used observations (e.g., ERA5, MSWEP, other reanalyses and datasets), CMIP6 models, large ensembles (e.g., CESM2-LE), and nudged simulations to isolate signals.
• Analysis focused on Northern Hemisphere mid-latitude winter trends over ~70+ years.

This work directly addresses why regional rainfall projections, especially in mid-latitudes during winter, carry high uncertainty despite clearer global thermodynamic signals.

Dynamic uncertainties stem from:

• Strong natural decadal variability in circulation.
• Potential model underestimation of how circulation responds to greenhouse gases, aerosols, or other forcings.
• Challenges in distinguishing forced trends from noise in relatively short observational records. bioengineer.org

The study suggests an emerging forced dynamic signal under further warming, but confidence remains limited. This has implications for predicting shifts in storm tracks, Mediterranean drying, European rainfall patterns, and risks of extremes (floods/droughts).

It aligns with prior findings on circulation biases in models (e.g., North Atlantic jet stream) and complements research on why mean precipitation and extremes are harder to project regionally than temperature. Data and code are available via Zenodo and OSF repositories for reproducibility.

The full paper is behind a paywall on Nature.com, but the abstract, figures (showing observed vs. simulated trends and decompositions), and supplementary information provide substantial detail. News summaries from Oxford, EurekAlert, and others offer accessible overviews.

This paper advances attribution science and highlights priorities for model improvement ahead of future CMIP phases.

Uncertain dynamic response of mid-latitude winter precipitation

Published: Nature (2026)

Journal information: Nature

peer-reviewed

DOI: 10.1038/s41586-026-10474-y

Provided by University of Oxford

Authors: Lei Gu, 
Dominik L. Schumacher, 
Sebastian Sippel, 
Erich M. Fischer, 
Istvan Dunkl, 
Robin Noyelle, 
Jitendra Singh, 
Lorenzo Pierini & 
Reto Knutti

Abstract

Understanding changes in precipitation is crucial for society and ecosystems^1,2. Studies have documented the respective contributions of anthropogenic forcing and internal variability to precipitation trends^3,4, yet discrepancies persist between observed and simulated patterns. In Northern Hemisphere winter, these mismatches are often attributed to unforced internal variability that dominates observed trends⁵. However, growing evidence also indicates that climate models underestimate the total response of precipitation to human forcings^6,7,8. Here we show that the thermodynamic contribution is broadly reproduced by climate models, whereas the dynamic contribution can diverge more substantially. Our approach disentangles the anthropogenic forced thermodynamic and dynamic components from internal variability in winter precipitation trends (1950–2022) to investigate their contribution to the trend discrepancies. In the Mediterranean, the forced dynamic signal from model simulations explains only about 10% of the observed dynamic trend, making detection challenging. Under continued anthropogenic emissions, the projected circulation response intensifies and more closely resembles observed trend patterns. Although internal variability in the observed record may contribute to this similarity, the results indicate an uncertain yet potentially emerging role of dynamic response in shaping regional winter precipitation trends. A reliable representation of the forced large-scale circulation response in climate models remains key for increasing confidence in regional precipitation projections.