Frei et al. (2025) Characterizing and correcting for global climate models’ biases in multiyear extreme precipitation scenarios

Identification

• Journal: Climatic Change
• Year: 2025
• Date: 2025-11-27
• Authors: Allan Frei, Rakesh K. Gelda, Rajith Mukundan
• DOI: 10.1007/s10584-025-04033-z

Research Groups

• Department of Geography and Environmental Science, Hunter College, City University of New York, New York, NY, USA
• Center for Sustainable Cities, Hunter College, City University of New York, New York, NY, USA
• Bureau of Water Supply, New York City Department of Environmental Protection, Kingston, NY, USA

Short Summary

This study introduces the Multiyear Precipitation Variability Bias Correction (MPVBC) method to address the systematic underestimation of multiyear extreme precipitation variability by Global Climate Models (GCMs). The method successfully corrects GCMs to match observed variability, leading to significantly more realistic future extreme drought and pluvial scenarios for water supply system resiliency assessments.

Objective

• To characterize and correct for Global Climate Models’ (GCMs) biases in multiyear precipitation variability, which tend to underestimate extreme events, by presenting and applying the Multiyear Precipitation Variability Bias Correction (MPVBC) method to provide more realistic multiyear extreme precipitation scenarios.

Study Configuration

• Spatial Scale: West-of-Hudson (WOH) basins of the New York City water supply system, located in the Catskill Mountain region of New York State, USA. Precipitation is averaged across the entire WOH region.
• Temporal Scale:
  • Calculation period (historical observations and GCM evaluation): 1960–2016.
  • Future period (GCM projections): 2017–2099.
  • Multiyear event durations analyzed: 1–20 years.

Methodology and Data

• Models used:
  • Global Climate Models (GCMs) from CMIP5 and CMIP6.
  • Downscaling methods applied to GCMs: Multivariate Constructed Adapted Analog (MACA), NASA method (quantile mapping + spatial interpolation), Localized Constructed Analogs (LOCA), and Seamless System for Prediction and Earth System Research (SPEAR-HI) (high-resolution GCM).
  • Bias correction: Initial bias correction for mean and sub-annual variability (Gelda et al., 2019) followed by the novel Multiyear Precipitation Variability Bias Correction (MPVBC) method.
• Data sources:
  • Daily station-based precipitation observations from Global Historical Climate Network Daily (GHCND) for 11 stations in the WOH region.
  • Gridded, downscaled GCM output from CMIP5 and CMIP6.

Main Results

• Pre-MPVBC GCMs underestimate multiyear precipitation variability in the study region by factors ranging from approximately 1.5 to 7 for 1–20 year durations, with median GCMs underestimating by a factor of 2 to 3.
• The MPVBC method successfully corrects GCM multiyear precipitation variability during the 1960–2016 calculation period to exactly match observed values.
• Pre-MPVBC GCMs provide no future drought scenarios comparable to the observed historical drought of record and only slightly wetter future pluvial scenarios.
• The MPVBC correction changes median GCM precipitation magnitude during the most extreme dry and wet events in the future period by 10% to 25%.
• Post-MPVBC future GCM droughts are comparable in magnitude to historical precipitation droughts, with the correction decreasing mean drought precipitation by 5% to 50% for 5–15 year durations.
• Post-MPVBC future GCM pluvials are significantly wetter than any precedent in the historical station record, with the correction increasing mean pluvial precipitation by 2% to 40% for 5–15 year durations.
• The choice of downscaling method affects the magnitude of multiyear precipitation variability and extremes, with LOCA models showing pre-MPVBC extremes closer to observed values compared to other methods.

Contributions

• Developed and validated the Multiyear Precipitation Variability Bias Correction (MPVBC) method, a novel cascading bias correction specifically designed to address the underestimation of multiyear precipitation variability in GCMs.
• Demonstrated that conventional downscaling and bias correction techniques, applied at annual or sub-annual timescales, are insufficient for correcting multiyear precipitation variability, highlighting a critical gap in climate impact assessments.
• Provided a practical and robust approach to generate more plausible and extreme multiyear precipitation scenarios (droughts and pluvials) for future climate impact and water supply system resiliency studies.
• Quantified the significant underestimation of future extreme precipitation events by GCMs and the substantial impact of the MPVBC method in rectifying these biases, offering a more comprehensive range of potential future conditions.

Funding

• New York City Department of Environmental Protection and the City University of New York (contracts WQMODEL-19 and WQMODEL-25).
• World Climate Research Programme’s Working Group on Coupled Modelling (CMIP).
• NASA Earth Exchange (NEX) for LOCA and NASA downscaled products.
• GFDL/NOAA for SPEAR-HI climate model output.
• Regional Approaches to Climate Change (REACCH) project and the South East Climate Science Center (SECSC) for MACAv2-METDATA.
• Support from Hazen & Sawyer, Jordan Gass, and Sijal Dangol.

Citation

@article{Frei2025Characterizing,
  author = {Frei, Allan and Gelda, Rakesh K. and Mukundan, Rajith},
  title = {Characterizing and correcting for global climate models’ biases in multiyear extreme precipitation scenarios},
  journal = {Climatic Change},
  year = {2025},
  doi = {10.1007/s10584-025-04033-z},
  url = {https://doi.org/10.1007/s10584-025-04033-z}
}