Scinocca et al. (2025) Runtime bias correction of regional climate model driving data and its continental-scale impacts

Identification

• Journal: Climate Dynamics
• Year: 2025
• Date: 2025-11-25
• Authors: John Scinocca, Viatcheslav Kharin, Dominic Matte, Yanjun Jiao, Marie-Pier Labonté, Minwei Qian, Dominique Paquin, Ayodeji Akingunola, Michael Lazare
• DOI: 10.1007/s00382-025-07814-5

Research Groups

• Canadian Centre for Climate Modelling and Analysis, Environment and Climate Change Canada
• Ouranos Consortium, Organization

Short Summary

This study introduces and evaluates an Empirical Runtime Bias Correction (ERBC) method for Earth System Model (ESM) driving data, demonstrating its effectiveness in significantly reducing global model biases in regional climate model (RCM) downscaling products and inducing statistically significant changes in climate-change circulation responses.

Objective

• To present and evaluate a novel Empirical Runtime Bias Correction (ERBC) method for Earth System Model (ESM) driving data, assessing its impact on reducing climatological biases in regional climate model (RCM) downscaling products and its potential to improve and reduce uncertainty in future climate change circulation responses.

Study Configuration

• Spatial Scale: North American NA-CORDEX domain with a horizontal resolution of approximately 50 kilometers (0.44 degrees angular resolution). Global models employ T63 spectral truncation. Regional climate models (CanRCM5, CRCM5) use 49 or 56 vertical levels extending from the surface up to approximately 100 Pascals or 1000 Pascals, respectively. Spectral nudging is applied on scales larger than approximately 1100 kilometers.
• Temporal Scale: Historical analysis period from 1981 to 2010. Evaluation simulations cover 1950-2014 (CanRCM5) and 1979-2014 (CRCM5). Future climate change projections for the period 2071-2100, following the Coupled Model Intercomparison Project phase 6 (CMIP6) SSP3-7.0 scenario.

Methodology and Data

• Models used:
  • Global Earth System Model (ESM): CanESM5.1 (referred to as CanESM5).
  • Atmospheric component of ESM: CanAM5.1 (referred to as CanAM5), run in AMIP mode.
  • Regional Climate Models (RCMs): CanRCM5 (employs the same physical parameterizations as CanAM5), CRCM5 (employs a distinct physics package based on the 33-kilometer meso-global GEM model, including CLASS 3.5 and FLAKE land surface/lake models).
  • Bias Correction Method: Empirical Runtime Bias Correction (ERBC) applied to winds, temperature, and specific humidity in proxy AMIP simulations of CanESM5, derived using the Climatological Adaptive Bias Correction (CABCOR) method. Diagnostic bias correction applied to Sea Surface Temperature (SST) and Sea Ice Concentration (SIC). CABCOR adaptation simulation used a vertical timescale varying from 1.157 x 10^-5 s^-1 at the surface to 3.858 x 10^-6 s^-1 at the model top.
• Data sources:
  • Reanalysis data: ERA5 (used as observational reference for SST, SIC, and atmospheric driving data for RCM evaluation runs).
  • Future forcing scenario: CMIP6 SSP3-7.0.

Main Results

• Historical Period Bias Reduction:
  • ERBC significantly reduces climatological biases in the global model (CanESM5). For instance, the root-mean-square error (RMSE) of 50000 Pascals geopotential height over the full domain is reduced by 71% in DJF and 34% in JJA.
  • Indirect improvements are observed for non-targeted variables: normalized RMSE bias of total precipitation over land is reduced by 50% in JJA and 7% in DJF. Screen temperature normalized RMSE bias is reduced by 47% in JJA and 51% in DJF over the full domain.
  • The Driving Bias Influence (DBI) on RCMs, which quantifies the impact of driving model biases, is substantially mitigated. When both atmospheric driving data and sea-surface forcings are bias-corrected, normalized RMSE DBI for 50000 Pascals geopotential height is reduced to 0.31 in CanRCM5 and 0.40 in CRCM5.
  • Overall, normalized DBI reductions exceed 60% for many variables and regions. However, the impact on precipitation can be complex and non-linear, sometimes leading to local degradations due to interactions with RCM internal physics.
• Future Projections and Uncertainty Reduction:
  • ERBC induces statistically significant changes in climate change circulation responses in both global and regional models, fulfilling a necessary condition for potential uncertainty reduction.
  • For JJA mean screen temperature (tas), ERBC causes reductions in the projected increase across Northern Canada and the south-central United States, and increases along the central East and West coasts of the continent in the global model. These patterns are largely preserved in the RCMs.
  • Statistically significant ERBC-induced differences in tas response cover roughly 70-75% of the domain.
  • ERBC also impacts the climate-change response of interannual variability. For JJA screen temperature, ERBC causes a statistically significant increase in variability over western North America and the eastern United States in the global model, with this pattern largely preserved in the RCMs.
  • Statistically significant non-zero response differences for interannual variability cover approximately 40-60% of the area.
  • ERBC induces robust differences (significantly exceeding 5% areal coverage) in the climate change response of all tested variables across different seasons.

Contributions

• Presents and evaluates a novel Empirical Runtime Bias Correction (ERBC) approach for Earth System Model (ESM) driving data, which maintains physical consistency and influences internal model dynamics, offering advantages over traditional diagnostic bias correction methods.
• Demonstrates that ERBC significantly mitigates global model biases in regional climate model (RCM) downscaling products during the historical period, quantified by the novel Driving Bias Influence (DBI) diagnostic.
• Shows that ERBC induces statistically significant changes in climate change circulation responses in both global and regional models, addressing a necessary condition for reducing uncertainty in future projections.
• Highlights the capacity of ERBC to improve both the realism of historical simulations and the robustness of future projections by altering large-scale circulation patterns and modifying regional responses.
• Introduces the Climatological Adaptive Bias Correction (CABCOR) method for deriving ERBCs, which is shown to produce larger climatological bias reductions compared to previous methods.

Funding

• Ouranos Consortium through grant GCXE23M023 as part of the Grants & Contributions program under the Flood Hazard Identification and Mapping Program, a national collaborative initiative led by Natural Resources Canada.
• Open access funding provided by Environment & Climate Change Canada library.

Citation

@article{Scinocca2025Runtime,
  author = {Scinocca, John and Kharin, Viatcheslav and Matte, Dominic and Jiao, Yanjun and Labonté, Marie-Pier and Qian, Minwei and Paquin, Dominique and Akingunola, Ayodeji and Lazare, Michael},
  title = {Runtime bias correction of regional climate model driving data and its continental-scale impacts},
  journal = {Climate Dynamics},
  year = {2025},
  doi = {10.1007/s00382-025-07814-5},
  url = {https://doi.org/10.1007/s00382-025-07814-5}
}