Satyapragyan et al. (2026) Bias corrections of ERA5 and ERA5-land temperature using automatic weather station data in the Higher Central Himalaya: implications for hydro-meteorological and glaciological research

Identification

• Journal: Journal of Hydrology Regional Studies
• Year: 2026
• Date: 2026-01-06
• Authors: Soumya Satyapragyan, Jairam Singh Yadav, Rakesh Bhambri
• DOI: 10.1016/j.ejrh.2025.103079

Research Groups

• Wadia Institute of Himalayan Geology, Dehradun, Uttarakhand, India
• AcSIR, Ghaziabad, Uttar Pradesh, India

Short Summary

This study evaluates and corrects biases in ERA5 and ERA5-Land mean temperature data for the Dokriani Glacier Catchment, Central Himalaya, using high-resolution daily observations from three Automatic Weather Stations. It found that Linear Regression and Generalized Additive Models were most effective, reducing biases to near zero and Root Mean Square Error by up to 86 % at the seasonal scale, significantly improving data reliability for hydro-meteorological and glaciological research.

Objective

• To compare the raw mean temperature (TMEAN) of ERA5 and ERA5-Land with observed Automatic Weather Station (AWS) data.
• To assess the effectiveness of Delta Change (DC), Linear Regression (LR), Empirical Quantile Mapping (EQM), Quantile Delta Mapping (QDM), and Generalized Additive Models (GAM) bias correction methods by comparing simulated TMEAN of ERA5 and ERA5-Land with observed data at daily, monthly, and seasonal timescales.
• To validate the effectiveness of bias correction techniques on raw ERA5 and ERA5-Land datasets for 2014–2015 based on the derived model equations.

Study Configuration

• Spatial Scale: Dokriani Glacier Catchment (DGC), Central Himalaya. Three AWSs located at different elevations and settings: glacierized (4364 m), proglacial (3763 m), and forested (2540 m). ERA5 data at approximately 27 km × 27 km resolution; ERA5-Land data at approximately 9 km × 9 km resolution.
• Temporal Scale: Analysis period from December 2011 to November 2014 for bias correction, and December 2014 to November 2015 for validation. Data analyzed at daily, monthly, and seasonal (winter: December–February, pre-monsoon: March–May, monsoon: June–September, post-monsoon: October–November) timescales.

Methodology and Data

• Models used: Delta Change (DC), Linear Regression (LR), Empirical Quantile Mapping (EQM), Quantile Delta Mapping (QDM), and Generalized Additive Models (GAM).
• Data sources:
  • ERA5 reanalysis mean temperature (TMEAN) data (2 m above ground).
  • ERA5-Land reanalysis mean temperature (TMEAN) data (2 m above ground).
  • High-resolution daily near-surface air temperature observations from three Automatic Weather Stations (AWSs) in the Dokriani Glacier Catchment.

Main Results

• Raw ERA5 and ERA5-Land TMEAN showed substantial spatio-temporal biases compared to AWS observations, with ERA5 overestimating temperatures during pre-monsoon and monsoon, and ERA5-Land underestimating during winter, pre-monsoon, and post-monsoon.
• Linear Regression (LR) and Generalized Additive Models (GAM) consistently outperformed other methods, reducing biases to near zero and Root Mean Square Error (RMSE) by up to 86 % at the seasonal scale.
• For ERA5 at the Advance Base Camp (ABC), the 3-year mean bias of 1.056 °C was reduced to approximately 0 °C, and RMSE decreased from 2.94 °C to 1.70 °C (42.18 % reduction) using GAM.
• For ERA5-Land at ABC, the 3-year mean bias of -2.179 °C was reduced to approximately 0 °C, and RMSE decreased from 3.46 °C to 1.64 °C (52.6 % reduction) using GAM.
• At the Base Camp (BC) for ERA5-Land, the 3-year mean bias of -5.22 °C was reduced to approximately 0 °C, and RMSE decreased from 5.88 °C to 1.74 °C (70.40 % reduction) using GAM.
• At the Tela Camp (TC), GAM reduced ERA5 RMSE from 7.44 °C to 1.57 °C (78.90 % reduction) and ERA5-Land RMSE from 7.81 °C to 1.72 °C (77.98 % reduction).
• Empirical Quantile Mapping (EQM) and Quantile Delta Mapping (QDM) showed improvements but were less effective than LR and GAM, especially at high-altitude stations.
• The Delta Change (DC) method showed the least improvement, often failing to account for temporal variability and sometimes increasing biases.
• Validation with independent 2014–2015 data confirmed that LR, EQM, QDM, and GAM effectively minimized errors, with LR and GAM showing particularly low residual errors during the monsoon.
• ERA5-Land generally performed better than ERA5, attributed to its finer spatial resolution (9 km vs. 27 km) and lapse-rate adjustments.
• Using the smallest possible grid size for ERA5 data further reduced errors, making it more efficient for glaciological and hydro-meteorological studies.

Contributions

• First study to use high-resolution daily temperature data from three AWSs in distinct glacierized, proglacial, and forested settings within the Dokriani Glacier Catchment, Central Himalaya, for rigorous evaluation and validation of bias correction methods across daily, monthly, and seasonal scales.
• Provides robust validation of reanalysis products using direct observations from glacierized areas, addressing gaps in previous studies that relied on coarser temporal resolutions or partial ground observations.
• Highlights that seasonal drivers differentially influence dataset-specific corrections, advocating for season-specific bias correction strategies to enhance accuracy in glacio-hydrological and climate modeling.
• Emphasizes the importance of utilizing ERA5 data with the smallest possible grid size and prioritizing ERA5-Land for more reliable simulation models in complex mountainous terrain.
• Enhances the reliability of reanalysis data for glacio-hydrological modeling and climate impact assessments in the Himalaya and other alpine regions with similar topography and climatic conditions.

Funding

• Wadia Institute of Himalayan Geology (WIHG), Dehradun, India (technical and infrastructural support).

Citation

@article{Satyapragyan2026Bias,
  author = {Satyapragyan, Soumya and Yadav, Jairam Singh and Bhambri, Rakesh},
  title = {Bias corrections of ERA5 and ERA5-land temperature using automatic weather station data in the Higher Central Himalaya: implications for hydro-meteorological and glaciological research},
  journal = {Journal of Hydrology Regional Studies},
  year = {2026},
  doi = {10.1016/j.ejrh.2025.103079},
  url = {https://doi.org/10.1016/j.ejrh.2025.103079}
}