Nguyen‐Duy et al. (2025) Performance and added value of a high-resolution (2 km) rainfall product based on WRF-downscaled ERA5 for Ho Chi Minh City, Vietnam

Identification

• Journal: Theoretical and Applied Climatology
• Year: 2025
• Date: 2025-11-17
• Authors: Tung Nguyen‐Duy, Thanh Ngo‐Duc, Marc Choisy, I. Fernández, Sarah Sparrow
• DOI: 10.1007/s00704-025-05894-1

Research Groups

• Oxford University Clinical Research Unit, Ho Chi Minh, Vietnam
• Department of Space and Applications, University of Science and Technology of Hanoi (USTH), Vietnam Academy of Science and Technology (VAST), Hanoi, Vietnam
• Centre for Tropical Medicine and Global Health, Nuffield Department of Medicine, University of Oxford, Oxford, United Kingdom
• Department of Engineering Science, University of Oxford, Oxford, UK

Short Summary

This study dynamically downscaled ERA5 reanalysis data using the WRF model and applied bias correction to generate a 2-km resolution rainfall product for Ho Chi Minh City, Vietnam. The bias-corrected product (WRFC-HCM) significantly improved daily rainfall accuracy and representation of extreme events compared to original reanalysis datasets, despite limited improvement at the monthly scale.

Objective

• To generate and evaluate a high-resolution (2-km) rainfall product for Ho Chi Minh City, Vietnam, by dynamically downscaling the ERA5 reanalysis dataset using the Weather Research and Forecasting (WRF) model and applying quantile mapping bias correction.

Study Configuration

• Spatial Scale: Ho Chi Minh City, Vietnam (approximately 2,061 km²). WRF model with two nested domains: d01 at 6 km resolution and d02 at 2 km resolution. Analysis focused on the innermost d02 domain (2 km).
• Temporal Scale: Simulation period from 2000 to 2016. Bias correction training period: 2000-2009. Validation period: 2010-2016. Evaluations performed at monthly and daily scales.

Methodology and Data

• Models used:
  • Weather Research and Forecasting (WRF) model, version 4.4, with specific physical parameterizations: SBU-YLin cloud microphysics, Kain-Fritsch convective scheme (d01 only), Dudhia shortwave radiation, Rapid Radiative Transfer Model (RRTM) longwave radiation, and Yonsei University planetary Boundary Layer scheme.
  • Empirical Quantile Mapping (QMBC) for bias correction.
• Data sources:
  • Initial and lateral boundary conditions for WRF: European Centre for Medium-Range Weather Forecasts (ECMWF) ERA5 reanalysis (0.25° horizontal resolution, 38 pressure levels, updated every 6 hours).
  • Comparison datasets: ERA5 (25 km resolution) and ERA5-Land (10 km resolution).
  • Observational data for calibration and validation: Daily rainfall data from 29 meteorological stations around Ho Chi Minh City.

Main Results

• At the monthly scale, ERA5 generally outperformed WRF-HCM, WRFC-HCM, and ERA5-Land, exhibiting the lowest Mean Absolute Error (MAE = 1.90 mm/day) and Root Mean Square Error (RMSE = 2.55 mm/day). WRFC-HCM reduced biases compared to raw WRF-HCM but did not significantly improve temporal correlation.
• At the daily scale, WRFC-HCM significantly improved accuracy, particularly in representing dry events. It showed a lower False Alarm Ratio (FAR, average 0.57) and higher overall Accuracy (ACC, 0.74) compared to ERA5/ERA5-Land (FAR ~0.65-0.7, ACC ~0.63).
• Seasonal probability density function (PDF) analysis revealed that ERA5 and ERA5-Land significantly underestimated dry days and overestimated light rain events, while WRFC-HCM substantially reduced these biases and provided better representation of moderate and heavy rainfall events (> 20 mm/day).
• For extreme climate indices, WRFC-HCM demonstrated the lowest errors for most indices. It closely matched observed total precipitation (PRCPTOT: 1565.82 mm/year vs. 1556.56 mm/year observed) and showed improved representation of maximum 1-day (Rx1day: 107.99 mm) and 5-day (Rx5day: 173.97 mm) precipitation compared to ERA5/ERA5-Land, which significantly underestimated these extremes.
• ERA5 and ERA5-Land substantially overestimated Consecutive Wet Days (CWD, 56.34-68.32 days vs. 8.59 days observed) and underestimated the Simple Daily Intensity Index (SDII, 8.37-8.79 mm/day vs. 15.68 mm/day observed), indicating an overestimation of wet days and underestimation of rainfall intensity.

Contributions

• This study is the first to generate and evaluate a high-resolution (2-km) rainfall product for Ho Chi Minh City, Vietnam, by dynamically downscaling ERA5 using the WRF model and applying quantile mapping bias correction.
• It demonstrates the significant added value of combining dynamical downscaling and bias correction for improving daily rainfall accuracy and the representation of extreme precipitation events in a flat, tropical monsoon region.
• The research highlights critical limitations of widely used reanalysis datasets (ERA5 and ERA5-Land) for fine-scale daily and extreme precipitation analysis in the study area, specifically their biases in dry event frequency, light rain overestimation, and heavy rain underestimation.

Funding

• Wellcome Trust [226052/Z/22/Z]

Citation

@article{NguyenDuy2025Performance,
  author = {Nguyen‐Duy, Tung and Ngo‐Duc, Thanh and Choisy, Marc and Fernández, I. and Sparrow, Sarah},
  title = {Performance and added value of a high-resolution (2 km) rainfall product based on WRF-downscaled ERA5 for Ho Chi Minh City, Vietnam},
  journal = {Theoretical and Applied Climatology},
  year = {2025},
  doi = {10.1007/s00704-025-05894-1},
  url = {https://doi.org/10.1007/s00704-025-05894-1}
}