Aranyossy et al. (2025) Multi-annual predictions of hot, dry and hot-dry compound extremes

Identification

• Journal: Earth System Dynamics
• Year: 2025
• Date: 2025-12-17
• Authors: Alvise Aranyossy, Paolo De Luca, Carlos Delgado‐Torres, Balakrishnan Solaraju-Murali, Margarida Samsó, Markus G. Donat
• DOI: 10.5194/esd-16-2225-2025

Research Groups

• Barcelona Supercomputing Center (BSC), Barcelona, Spain
• Universitat de Barcelona, Barcelona, Spain
• Zurich Insurance, Barcelona, Spain
• Institució Catalana de Recerca i Estudis Avançats (ICREA), Barcelona, Spain

Short Summary

This study evaluates the multi-year predictability of hot, dry, and hot-dry compound extremes using CMIP6 decadal hindcast experiments for forecast years 2–5. It finds that hot-dry compound and hot extremes are skillfully predicted over most land regions, while skill for dry extremes is more limited, with most predictability stemming from external forcings and long-term trends rather than initialisation.

Objective

• To evaluate the ability of the CMIP6 multi-model decadal climate hindcast to predict hot, dry, and hot-dry compound climate extremes for forecast years 2–5.
• To investigate the improvement in prediction skill attributable to the initialisation of the forecast.
• To analyze and compare the representation of interconnections between compound extremes and their univariate counterparts in models and observations.

Study Configuration

• Spatial Scale: Global land regions.
• Temporal Scale: Multi-annual predictions (forecast years 2–5) derived from decadal hindcasts (1960–2009 initialisation, covering 1962–2014 for analysis). Extreme indices calculated with a 3-month accumulation period for dry conditions.

Methodology and Data

• Models used: Multi-model ensemble from the Coupled Model Intercomparison Project Phase 6 (CMIP6), specifically the Decadal Climate Prediction Project (DCPP MME) hindcasts and CMIP6 historical forcing simulations (Hist MME). Models include NorCPM1, EC-Earth, IPSL-CM6A-LR, MIROC6, MPI-ESM1.2-HR, CanESM5, and CMCC.
• Data sources:
  • Observation-based reference: GPCC-BEST (Global Precipitation Climatology Centre (GPCC) for monthly total precipitation and Berkeley Earth Surface Temperatures (BEST) for daily maximum and minimum temperatures).
  • Reanalysis reference: ERA5 (European Centre for Medium-Range Weather Forecasts (ECMWF) Reanalysis v5) for daily total precipitation, daily maximum, and daily minimum temperature.
  • Extreme indicators:
    • Hot extremes (TX90p): Days above the 90th percentile of daily maximum temperature.
    • Dry extremes: Standardized Precipitation Index (SPI3dry) and Standardized Precipitation-Evapotranspiration Index (SPEI3dry) with values less than or equal to -1, calculated over a 3-month accumulation period using the Hargreaves method for potential evapotranspiration.
    • Hot-dry compound extremes (HDSPI3, HDSPEI3): TX90p days co-occurring with SPI3dry or SPEI3dry months.
  • Forecast quality assessment: Spearman's rank correlation coefficient and residual correlation (to quantify skill improvement from initialisation).

Main Results

• The CMIP6 multi-model ensemble skillfully predicts hot-dry compound extremes and hot extremes (TX90p) over most land regions for forecast years 2–5.
• Prediction skill for dry extremes (SPI3dry and SPEI3dry) is more limited, though SPEI3dry generally shows higher skill than SPI3dry, particularly over mid-latitudes.
• Improvements in skill due to hindcast initialisation are minor and spatially limited, especially for hot-dry compound extremes. Most of the observed skill is attributed to external forcings, primarily long-term trends.
• The decadal hindcast systems are generally able to reproduce the observed connections between compound extremes and their univariate components, with dry conditions identified as a leading factor in the variability of hot-dry extremes.
• Sensitivity tests show that changing extreme thresholds (e.g., 95th percentile for hot, -1.5 for dry) does not significantly alter overall results. However, increasing the accumulation period for dry conditions (to 6 or 12 months) generally leads to a decrease in significant skill, except for specific regions like Northeast Asia.

Contributions

• Provides the first multi-model assessment of the predictability of hot-dry compound extremes on multi-annual timescales using CMIP6 decadal predictions.
• Quantifies the relative contributions of initialisation and external forcings (long-term trends) to the skill of multi-annual predictions for hot, dry, and hot-dry compound extremes.
• Demonstrates the ability of decadal prediction systems to reproduce the observed interconnections between compound and univariate extremes, highlighting the critical role of dry conditions.
• Identifies the limited skill in predicting the interannual variability of dry extremes as a bottleneck for improving hot-dry compound extreme predictions.

Funding

• European Union's Horizon Europe research and innovation programme (ASPECT project, grant no. 101081460).
• Departament de Recerca i Universitats de la Generalitat de Catalunya (Climate Variability and Change (CVC) Research Group, reference 2021 SGR 00786).
• AXA Research Fund.
• MCIN/AEI/10.13039/501100011033 and ESF Investing in Your Future (grant no. PRE2022-104391) for Alvise Aranyossy.
• EU Horizon Europe Marie Skłodowska-Curie Actions (grant no. 101059659) for Paolo De Luca.

Citation

@article{Aranyossy2025Multiannual,
  author = {Aranyossy, Alvise and Luca, Paolo De and Delgado‐Torres, Carlos and Solaraju-Murali, Balakrishnan and Samsó, Margarida and Donat, Markus G.},
  title = {Multi-annual predictions of hot, dry and hot-dry compound extremes},
  journal = {Earth System Dynamics},
  year = {2025},
  doi = {10.5194/esd-16-2225-2025},
  url = {https://doi.org/10.5194/esd-16-2225-2025}
}