Helfer et al. (2025) Enhanced baseflow separation in rural catchments: event-specific calibration of recursive digital filters with tracer-derived data

Identification

• Journal: Hydrology and earth system sciences
• Year: 2025
• Date: 2025-12-03
• Authors: Fernanda Helfer, Felipe Bernardi, Cláudia Alessandra Peixoto de Barros, Daniel Allasia, Jean Paolo Gomes Minella, Rutinéia Tassi, Néverton Scariot
• DOI: 10.5194/hess-29-6959-2025

Research Groups

• School of Engineering and Built Environment, Griffith University, Gold Coast, QLD, Australia
• Sanitary and Environmental Engineering Department, Federal University of Santa Maria, Santa Maria, RS, Brazil
• Soil Department, Federal University of Rio Grande do Sul, Porto Alegre, RS, Brazil
• Soil Department, Federal University of Santa Maria, Santa Maria, RS, Brazil

Short Summary

This study enhanced baseflow separation in a small rural catchment by developing an innovative event-specific calibration methodology for Recursive Digital Filters (RDFs) using silica tracer data, demonstrating that dynamic calibration significantly improves accuracy, particularly for the Eckhardt's filter.

Objective

• To calibrate the BFImax and Beta parameters for Eckhardt’s and Lyne and Hollick’s (LH) filters, minimizing the percentage bias between filter estimates and dissolved silica concentration analyses.
• To assess and contrast the performance of the calibrated Eckhardt’s and LH filters alongside the Chapman and Maxwell (CM) filter in generating baseflow hydrographs, evaluating shapes, peak timing, and baseflow to streamflow ratio.
• To examine the performance of the Eckhardt’s and LH filters using an “event calibration approach,” determining distinct BFImax and Beta values for different rainfall–runoff event intensity ranges.

Study Configuration

• Spatial Scale: Arvorezinha catchment, Brazil; Drainage area: 1.2 square kilometers.
• Temporal Scale: Hydrological monitoring from 2010–2015, with intensive dissolved silica and flow monitoring for 15 specific events.

Methodology and Data

• Models used:
  • Recursive Digital Filters (RDFs): Eckhardt's filter (two-parameter), Lyne and Hollick (LH) filter (one-parameter), Chapman and Maxwell (CM) filter (one-parameter).
  • Mass balance approach for dissolved silica (DSi) to derive observed baseflow.
  • Bisection method for parameter optimization.
• Data sources:
  • Precipitation: Pluviograph (10 minute interval data), two pluviometers (daily totals).
  • Streamflow: Water level encoder and flume (10 minute interval readings converted to flow rates).
  • Dissolved Silica (DSi): Water samples collected at catchment outlet using a USDH-48 sampler (sampling frequency 30 minutes to 4 hours during events).
  • 204 hydrographs (2010-2015) for recession constant determination.
  • 15 specific rainfall-runoff events with complete DSi and flow datasets for calibration and evaluation.

Main Results

• The recession constant for the catchment was determined to be 0.952 (decay rate ≈ 0.05 h⁻¹), indicating rapid baseflow recession.
• General Calibration:
  • Optimal BFImax for Eckhardt's filter was 0.653, and optimal Beta for the LH filter was 0.965.
  • Eckhardt's filter achieved a Percent Bias (PBias) near 0%, Nash–Sutcliffe Efficiency (NSE) of 0.85, Kling–Gupta Efficiency (KGE) of 0.76, and Normalized Root Mean Square Deviation (NRMSD) of 7.5%. Its baseflow (BF) ratio was 64%.
  • LH filter achieved a PBias near 0%, NSE of 0.79, KGE of 0.74, and NRMSD of 8.9%. Its BF ratio was 75%.
  • The uncalibrated CM filter showed a PBias of -28%, NSE of 0.86, KGE of 0.70, and NRMSD of 7.2%, significantly underestimating baseflow.
  • The Eckhardt's filter showed the most accurate BF ratio (64%) compared to the DSi-derived 66% and exhibited the lowest variability (coefficient of variation of 0.04).
• Event-based Calibration:
  • Eckhardt's BFImax values varied with event magnitude: 0.809 for low, 0.701 for medium, and 0.576 for high magnitude events.
  • LH Beta values varied with event magnitude: 0.921 for low, 0.957 for medium, and 0.970 for high magnitude events.
  • Event-based calibration significantly improved performance metrics for both Eckhardt's and LH filters. For Eckhardt's, PBias was < 0.50%, NSE = 0.92, KGE = 0.91, and NRMSD = 5.4%. For LH, PBias was < 1.20%, NSE = 0.84, KGE = 0.85, and NRMSD = 7.8%.
  • The average delay between baseflow and streamflow peaks was reduced from 1.5 hours to 1.3 hours for Eckhardt's filter and from 4.4 hours to 3.4 hours for LH filter with event-based calibration.
  • The Eckhardt's filter with event-based calibration emerged as the most robust and accurate method.

Contributions

• Introduced and validated an innovative event-specific calibration methodology for Recursive Digital Filters (RDFs) using tracer-derived baseflow data (dissolved silica).
• Demonstrated that RDF parameters (BFImax and Beta) are not static but vary significantly with rainfall-runoff event magnitude, challenging the "one-size-fits-all" approach.
• Provided optimal, catchment-specific, and event-magnitude-dependent parameters for Eckhardt's and LH filters in a small rural catchment in Southern Brazil.
• Showcased the superior performance of the Eckhardt's filter, especially when dynamically calibrated, for accurate baseflow separation.
• Highlighted the limitations of uncalibrated methods like the CM filter and the importance of tracer data for robust calibration.
• Offered practical recommendations for practitioners in data-rich and data-scarce environments.

Funding

• Coordenação de Aperfeiçoamento de Pessoal de Nível Superior – Brazil (CAPES) under Finance Code 001, through the Young Talent Grant (reference no. 88887.893273/2023-00).
• Federal University of Santa Maria (UFSM).
• Brazilian research agencies FAPERGS and CNPq.
• Griffith University (Australia) through the Griffith Academic Studies Program 2023–2024.

Citation

@article{Helfer2025Enhanced,
  author = {Helfer, Fernanda and Bernardi, Felipe and Barros, Cláudia Alessandra Peixoto de and Allasia, Daniel and Minella, Jean Paolo Gomes and Tassi, Rutinéia and Scariot, Néverton},
  title = {Enhanced baseflow separation in rural catchments: event-specific calibration of recursive digital filters with tracer-derived data},
  journal = {Hydrology and earth system sciences},
  year = {2025},
  doi = {10.5194/hess-29-6959-2025},
  url = {https://doi.org/10.5194/hess-29-6959-2025}
}