Beyene et al. (2025) Comparison of bias correction methods to enhance CHIRP rainfall estimates for improved streamflow simulation at Ziway-Shalla catchment, Ethiopia

Identification

• Journal: Journal of Hydrology Regional Studies
• Year: 2025
• Date: 2025-11-11
• Authors: Terhas Legese Beyene, Alemseged Tamiru Haile, Demelash Wondimagegnehu Goshime, Tigabie Setu Birhan
• DOI: 10.1016/j.ejrh.2025.102920

Research Groups

• Aksum University, Shire Faculty of Water Technology (SFWT), Shire, Ethiopia
• International Water Management Institute (IWMI), Addis Ababa, Ethiopia
• Arba Minch University Water Technology Institute (AWTI), Arba Minch, Ethiopia
• Wollo University, Kombolcha Institute of Technology (KIoT), Kombolcha, Ethiopia

Short Summary

This study compares five bias correction methods for the CHIRP satellite rainfall product in Ethiopia's Ziway-Shalla catchment, finding that Quantile Mapping based on Gamma distribution (QMG) and Power Transformation (PT) significantly improve streamflow simulations by the HBV model compared to raw CHIRP data.

Objective

• To evaluate the Climate Hazards Group Infrared Precipitation (CHIRP) rainfall product in capturing rainfall statistics in the Ziway-Shalla sub-basin.
• To identify the best bias correction method for the study site.
• To quantify the propagation of satellite rainfall error to simulated streamflow using the Hydrologiska Byråns Vattenbalansavdelning (HBV) rainfall-runoff model.

Study Configuration

• Spatial Scale: Ziway-Shalla Catchment, Rift Valley Basin, Ethiopia (7°20’00" to 8°20’00" N latitude and 38°00’00" to 39°00’00" E longitude). Elevation ranges from 1608 meters to 4213 meters above mean sea level, with a mean elevation of 2275 meters.
• Temporal Scale:
  • Observed meteorological data: 1981–2019 (38 years).
  • Daily river discharge data: 1984–2000.
  • CHIRP product: 1984–2010 (approximately 26 years).
  • Bias correction applied at daily, 14-day, monthly, and seasonal periods.
  • HBV model calibration: 1986–1991.
  • HBV model validation: 1992–1996.
  • HBV warm-up period: 1984–1985.

Methodology and Data

• Models used:
  • Bias Correction Methods: Power Transformation (PT), Quantile Mapping based on Gamma distribution (QMG), Daily Translation (DAT), Distribution Transformation (DT), Linear Scaling (LS).
  • Hydrological Model: Hydrologiska Byråns Vattenbalansavdelning (HBV) rainfall-runoff model (Integrated Hydrological Modeling System (IHMS) version 6.3).
  • Spatial Interpolation: Inverse Distance Weight (IDW) for distributing bias-correction parameters.
• Data sources:
  • Satellite Rainfall: Climate Hazards Group Infrared Precipitation (CHIRP) product (daily time scale, 0.05° x 0.05° spatial resolution).
  • Observation Data: Daily observed meteorological data (rainfall, minimum and maximum temperature, relative humidity, solar radiation, wind speed) from 15 reliable rain gauge stations obtained from the Ethiopian Meteorological Institute (EMI). Daily river discharge data for five rivers (Meki, Katar, Kerekaristu, Harekelo, and Gedemso) from the Ethiopian Ministry of Water and Energy (MoWE).

Main Results

• The raw CHIRP satellite rainfall product exhibited large biases and a weak correlation with observed rainfall across various spatial and temporal scales in the Ziway-Shalla catchment, generally overestimating rainfall at most stations except Asela and Tora.
• The accuracy of CHIRP data improved significantly when aggregated from daily to 14-day, monthly, and seasonal time scales.
• All five bias correction methods improved the accuracy of CHIRP satellite rainfall estimates, but the degree of improvement varied substantially.
• Quantile Mapping based on Gamma distribution (QMG) and Power Transformation (PT) methods consistently outperformed Daily Translation (DAT), Linear Scaling (LS), and Distribution Transformation (DT), showing lower Mean Error (ME), Root Mean Error (RME), Pbias, and Root Mean Squared Error (RMSE), along with higher correlation (R²). DT performed the worst.
• When used as input for the HBV streamflow model, raw CHIRP data resulted in a large Relative Volume Error (RVE) of approximately 11%.
• Bias-corrected CHIRP rainfall estimates significantly improved the HBV model's performance in capturing both the volume and pattern of the observed streamflow hydrograph, with QMG and PT-based simulations showing the best agreement (NSE > 0.7 for calibration, > 0.6 for validation, and RVE < 5%).
• Calibrated HBV model parameter values changed notably depending on whether raw or bias-corrected rainfall data were used, indicating that the model calibration compensated for rainfall input errors.

Contributions

• Provides a comprehensive comparative evaluation of five bias correction methods for the CHIRP satellite rainfall product in the data-scarce Ziway-Shalla catchment, Ethiopia.
• Identifies the most effective bias correction methods (QMG and PT) for enhancing CHIRP data utility in hydrological modeling within the specific study region.
• Quantifies the significant improvement in streamflow simulation accuracy achieved by applying bias correction to satellite rainfall data, highlighting its critical role in water resource management in regions with limited ground observations.
• Demonstrates how hydrological model parameters adjust to compensate for biases in rainfall input, emphasizing the importance of accurate precipitation data for robust model calibration and reliable simulations.

Funding

• Ministry of Water and Energy (MOWE)
• Ethiopian Meteorological Institute (EMI)
• Climate Hazards Group (CHG)
• Swedish Meteorological and Hydrological Institute (for providing the HBV model)

Citation

@article{Beyene2025Comparison,
  author = {Beyene, Terhas Legese and Haile, Alemseged Tamiru and Goshime, Demelash Wondimagegnehu and Birhan, Tigabie Setu},
  title = {Comparison of bias correction methods to enhance CHIRP rainfall estimates for improved streamflow simulation at Ziway-Shalla catchment, Ethiopia},
  journal = {Journal of Hydrology Regional Studies},
  year = {2025},
  doi = {10.1016/j.ejrh.2025.102920},
  url = {https://doi.org/10.1016/j.ejrh.2025.102920}
}