Zhang et al. (2025) Multi-scale assessment of ERA5 hourly pressure-level data on a global scale with in- situ observations

Identification

• Journal: Theoretical and Applied Climatology
• Year: 2025
• Date: 2025-10-13
• Authors: Qingbo Zhang, Jun Li
• DOI: 10.1007/s00704-025-05794-4

Research Groups

• Jiangsu Urban and Rural Construction Vocational College, Changzhou, China
• College of Geological Engineering and Geomatics, Chang’an University, Xi’an, China

Short Summary

This study globally evaluates the performance of ERA5 hourly pressure-level data (2020–2023) for temperature, pressure, specific humidity, and relative humidity against 9,957 in-situ observations, revealing variable accuracy influenced by seasonal, latitudinal, and altitudinal factors.

Objective

• To evaluate the global simulation performance of ERA5 hourly pressure-level data for temperature, pressure, specific humidity, and relative humidity using a dense network of in-situ observations.
• To analyze the variations in these meteorological variables across daily, monthly, and seasonal scales, with particular attention to the influences of altitude and latitude.

Study Configuration

• Spatial Scale: Global (0.25° × 0.25° grid for ERA5), utilizing 9,957 in-situ observation stations distributed across all continents, with specific analysis for Arctic (66.5°N–90°N) and Antarctic (66.5°S–90°S) regions, and various elevation ranges.
• Temporal Scale: January 1, 2020, to January 1, 2023 (3 years). Data extracted at 3-hour intervals for consistency, and evaluated on daily, monthly, and seasonal scales.

Methodology and Data

• Models used: ERA5 reanalysis system, which employs the IFS 41r2 system 4D-Var data assimilation method.
• Data sources:
  • ERA5 hourly data on pressure levels (37 isobaric layers, 0.25° × 0.25° spatial resolution, 1 hour temporal resolution).
  • HadISD dataset (version 3.4.1.202410p): A comprehensive surface database from the UK Met Office Hadley Centre, providing in-situ meteorological data from 9,957 stations globally at a minimum temporal resolution of 3 hours.
  • Earth Gravitational Model 2008 (EGM2008) for elevation reference unification.
  • Statistical methods: Root Mean Square Error (RMSE), Correlation Coefficient (CC), Bias, and Slope.
  • Interpolation/Extrapolation: Inverse Distance Weighted (IDW) for horizontal interpolation; univariate linear regression (using a tropospheric average temperature lapse rate of -0.0065 K/m) and linear extrapolation for vertical interpolation/extrapolation.

Main Results

• Temperature: RMSE ranges from 2.5 °C to 3.5 °C, CC between 94% and 98%, and Bias between -2 °C and 0.2 °C. Best agreement in summer (Bias 0.072 °C, RMSE 1.023 °C), worst in winter (Bias 1.136 °C, RMSE 1.862 °C). Errors show significant latitude and seasonal effects, with general overestimation in spring, summer, and winter (up to 4 °C in low-mid latitudes during winter).
• Pressure: RMSE ranges from 3.9 hPa to 4.3 hPa, CC between 99.72% and 99.78%, and Bias between -0.5 hPa and 0.1 hPa. Exhibits extremely high correlations across all seasons. Smallest bias (0.277 hPa) and RMSE (4.027 hPa) occur in spring. Errors increase with elevation, reaching up to 15 hPa bias in high-altitude regions (e.g., Himalayas, Andes), and are more significant in the Northern Hemisphere.
• Specific Humidity: RMSE ranges from 1.5 g/kg to 3.5 g/kg, CC between 93% and 99%, and Bias between 0 g/kg and 3 g/kg. Larger bias and RMSE are observed in summer (Bias -2.535 g/kg, RMSE 3.053 g/kg) and autumn (Bias -1.312 g/kg, RMSE 2.018 g/kg). Systematic underestimation at higher values (slopes below 0.8). Overestimated errors are primarily distributed in tropical regions (30°S to 30°N), peaking at 8 g/kg near the equator, and decreasing to -2 g/kg in polar regions. Errors decrease with increasing elevation.
• Relative Humidity: RMSE ranges from 10% to 15%, CC between 70% and 90%, and Bias between -4% and 4%. This is the worst-performing variable, with lower correlation than specific humidity and temperature, especially in winter (CC 0.856). Pronounced positive biases are found in spring, summer, and winter (1.827 %rh, 2.048 %rh, and 1.941 %rh, respectively), with near-zero bias in autumn (0.164 %rh). RMSE ranges from 5.024 %rh to 7.435 %rh (highest in winter). Exhibits systematic underestimation under high relative humidity conditions (slopes below 0.95). Markedly underestimated over the Tibetan Plateau, Arctic, and eastern Pacific coast (deficits up to 20 %rh), while substantially overestimated across the Mediterranean (up to 15 %rh).

Contributions

• Provides the first systematic global assessment of ERA5 hourly pressure-level data for key meteorological variables (temperature, pressure, specific humidity, relative humidity) using a dense network of in-situ observations, addressing a gap in existing regional studies.
• Offers detailed insights into the multi-scale (daily, monthly, seasonal) and spatial (latitude, altitude) variability of ERA5 accuracy, clarifying its strengths and weaknesses across different climate regimes and terrains.
• Delivers valuable guidance for the application of ERA5 data in diverse fields such as weather forecasting, climate research, geodesy, and satellite remote sensing.
• Establishes a comprehensive reference for the future performance evaluation of the next-generation reanalysis product, ERA6.

Funding

• National Natural Science Foundation of China (No. 61872095)

Citation

@article{Zhang2025Multiscale,
  author = {Zhang, Qingbo and Li, Jun},
  title = {Multi-scale assessment of ERA5 hourly pressure-level data on a global scale with in- situ observations},
  journal = {Theoretical and Applied Climatology},
  year = {2025},
  doi = {10.1007/s00704-025-05794-4},
  url = {https://doi.org/10.1007/s00704-025-05794-4}
}