Vries et al. (2025) Precipitation disaster hotspots depend on historical climate variability

Identification

• Journal: Nature Communications
• Year: 2025
• Date: 2025-11-29
• Authors: Iris de Vries, Maybritt Schillinger, Erich Fischer, Sebastian Sippel, Reto Knutti
• DOI: 10.1038/s41467-025-66601-2

Research Groups

• Institute for Atmospheric and Climate Science, ETH Zürich, Zürich, Switzerland
• Department of Earth, Atmospheric and Planetary Sciences, MIT, Cambridge, MA, USA
• Seminar for Statistics, ETH Zürich, Zürich, Switzerland
• Leipzig Institute for Meteorology, Leipzig University, Leipzig, Germany

Short Summary

This study investigates how historical climate variability and climate change interact to shape the probability of future record-breaking precipitation events and identifies global disaster hotspots where high risk combines with potentially low societal preparedness. It reveals that regions with low historical precipitation records are currently most vulnerable, while those with high records face the steepest increase in risk due to climate change, exposing over a billion people to high record-breaking probabilities by 2100.

Objective

• To assess local future annual daily maximum precipitation (Rx1d) record-breaking probabilities conditional on historical observations, accounting for non-stationarity in the distribution of extreme precipitation.
• To identify regions with high disaster potential by combining quantitative record-breaking probabilities with a qualitative assessment of disaster preparedness influenced by historical record evolution and societal risk perception biases.

Study Configuration

• Spatial Scale: Global land areas, analyzed at grid cell levels of 1.875° × 1.25° (HadEX3) and 1° × 1° (REGEN, ERA5), corresponding to areas of a few hundred square kilometers.
• Temporal Scale: Historical observations/reanalysis from 1950 to 2015. Future projections from 2016 to 2100.

Methodology and Data

• Models used:
  • Multi-model ensemble of 8 Coupled Model Intercomparison Project Phase 6 (CMIP6) climate models (e.g., ACCESS-ESM1-5).
  • Generalized Extreme Value (GEV) distributions, fitted with non-stationary parameters (location µ and scale σ varying linearly with global mean surface temperature, shape ξ constant).
• Data sources:
  • Gridded in situ observations: HadEX3 (1.875° × 1.25° resolution), REGEN (1° × 1° resolution) for 1950–2015.
  • Reanalysis data: ERA5 (0.5° × 0.5° regridded to 1° × 1° resolution) for 1950–2015.
  • Future climate scenario: Shared Socioeconomic Pathway (SSP) 2-4.5.
  • Population data: Gridded population numbers for SSP2 from the Socioeconomic Data and Applications Centre (SEDAC).

Main Results

• Moderate climate change (SSP2-4.5) is projected to increase the average record-breaking probability of annual daily maximum precipitation (Rx1d) by 40% by 2050.
• Regions with low current historical Rx1d records are most at risk of near-term record-breaking events, exhibiting high initial record-breaking probabilities.
• Regions with high current historical Rx1d records, while having lower absolute record-breaking probabilities, show the steepest relative increase in risk due to climate change, with probabilities increasing by up to 75% by 2050 compared to a stationary climate.
• By 2100, over 1 billion people globally are projected to live in regions with a greater than 75% probability of experiencing a local Rx1d record-breaking event under the SSP2-4.5 scenario.
• Approximately 0.3 billion people are estimated to live in areas by 2100 where high record-breaking precipitation probabilities coincide with low disaster preparedness (defined by a state likelihood ≤0.5 and conditional cumulative record-breaking probability ≥0.75).
• Key regions identified with high disaster potential include eastern Asia, Brazil, and Australia, often characterized by rapid urbanization and vulnerability.

Contributions

• This study introduces a novel approach by focusing on local record-breaking precipitation probabilities conditional on observed historical record levels, integrating both natural variability and climate change effects, which differs from traditional marginal (history-unaware) assessments.
• It quantifies the non-linear, quantile-dependent effect of climate change on record-breaking probabilities, showing that the rarest extremes become more likely at the highest rate.
• The concept of "state likelihood" is introduced as a qualitative measure of disaster preparedness, linking historical record gaps and low record values to potential societal biases in risk perception and inadequate resilience building.
• It provides a global quantification of population exposure to combined high record-breaking probability and low disaster preparedness, highlighting areas of major imminent threat, particularly in developing economies.

Funding

• EU Horizon 2020 Project XAIDA (grant no. 101003469)

Citation

@article{Vries2025Precipitation,
  author = {Vries, Iris de and Schillinger, Maybritt and Fischer, Erich and Sippel, Sebastian and Knutti, Reto},
  title = {Precipitation disaster hotspots depend on historical climate variability},
  journal = {Nature Communications},
  year = {2025},
  doi = {10.1038/s41467-025-66601-2},
  url = {https://doi.org/10.1038/s41467-025-66601-2}
}