Li et al. (2025) Generative Downscaling and Bias Correction of Multivariable Earth System Model Simulations

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: Geophysical Research Letters
• Year: 2025
• Date: 2025-09-13
• Authors: Haijie Li, Ya Wang, Weichen Tao, Pengfei Lin
• DOI: 10.1029/2025gl117397

Research Groups

Not specified in abstract.

Short Summary

This study introduces a multivariate generative downscaling model (MVGDM) to simultaneously downscale global climate simulations from 100 km to 25 km resolution and correct inherent biases, significantly improving the simulation of key climate variables, phenomena like ENSO and IOD, and climate extremes.

Objective

• To introduce a multivariate generative downscaling model (MVGDM) that downscales global climate simulations from a 100 km to a 25 km resolution while simultaneously correcting climate simulation biases.

Study Configuration

• Spatial Scale: Downscaling from 100 km to 25 km resolution.
• Temporal Scale: Not explicitly specified, but implied to be climatological and relevant for future climate projections.

Methodology and Data

• Models used: Multivariate Generative Downscaling Model (MVGDM).
• Data sources: Global climate simulations (as input to MVGDM).

Main Results

• The MVGDM successfully downscales global climate simulations from 100 km to 25 km resolution.
• It reduces climatological biases for sea surface temperature (SST) by 72%, 2-meter temperature (T2M) by 79%, and 500-hPa geopotential height (Z500) by 71%.
• The model mitigates the common westward bias in El Niño-Southern Oscillation (ENSO)-related SST anomalies.
• It significantly improves the simulation of the Indian Ocean Dipole.
• The MVGDM substantially enhances the simulation of climate extremes.

Contributions

• Introduction of a novel multivariate generative downscaling model (MVGDM) capable of simultaneous high-resolution downscaling and bias correction for climate simulations.
• Demonstrated significant improvements in the accuracy of simulated climate variables and the representation of multi-scale physical processes.
• Enhanced simulation of critical climate phenomena (ENSO, IOD) and climate extremes.
• Offers potential for improving future climate projections.

Funding

Not specified in abstract.

Citation

@article{Li2025Generative,
  author = {Li, Haijie and Wang, Ya and Huang, Gang and Tao, Weichen and Lin, Pengfei},
  title = {Generative Downscaling and Bias Correction of Multivariable Earth System Model Simulations},
  journal = {Geophysical Research Letters},
  year = {2025},
  doi = {10.1029/2025gl117397},
  url = {https://doi.org/10.1029/2025gl117397}
}