Kim et al. (2026) Improved river transmission loss modelling for environmental flow releases during droughts

Identification

• Journal: Environmental Modelling & Software
• Year: 2026
• Date: 2026-01-20
• Authors: Shaun Sh Kim, Russell S. Crosbie, Ursula Zaar, Jai Vaze, Bill Wang, Cherry May R. Mateo, Rebekah May, Sudeep Nair, Jahangir Alam
• DOI: 10.1016/j.envsoft.2026.106880

Research Groups

• CSIRO, Environment, Black Mountain, ACT, Australia
• CSIRO, Environment, Waite, SA, Australia
• CSIRO, Environment, Waterford, WA, Australia
• eWater Group, Bruce, ACT, Australia
• MDBA, Canberra, ACT, Australia

Short Summary

This study develops and evaluates two novel river transmission loss models, Dynamic Maximum Alluvium as River Storage (DMAARS) and DMAARS coupled with river dead storage (DMAARSDS), demonstrating their significantly improved performance over a benchmark model in simulating environmental flow events during droughts in the northern Murray-Darling Basin and their potential for long-term water resource planning.

Objective

• To develop and describe parsimonious transmission loss models that consider antecedent conditions and can conveniently be adopted and calibrated in current basin-scale models.
• To test the new models against the 2018/2019 environmental flow events and assess these based on resulting environmental flow metrics.
• To perform an ex post facto analysis of the 2018/2019 Northern Murray-Darling Basin flow events that explores the sensitivity to the tested models when applying alternate flow releases.

Study Configuration

• Spatial Scale: Basin-scale, specifically the northern Murray-Darling Basin (Border Rivers, Gwydir, and Barwon-Darling catchments, covering 44,000 km², 22,000 km², and 105,000 km² respectively). River reaches range from 10 km to 1000 km in length.
• Temporal Scale: Daily time steps, focusing on environmental flow events during the 2018/2019 drought. Calibration and validation periods included short-term event-specific periods (e.g., April-June 2018/2019) and longer-term periods (e.g., 1972-2017) for split-sample tests, with warm-up periods of 3 to 10 years.

Methodology and Data

• Models used:
  • Dynamic Maximum Alluvium as River Storage (DMAARS)
  • DMAARS coupled with river dead storage (DMAARSDS)
  • Piecewise Linear (PWL) loss model (benchmark)
  • Australian Water Resource Assessment River (AWRA-R) model (base platform)
  • GR4J (rainfall-runoff component)
  • Muskingum and lag routing equations (river routing)
  • Autoregressive error models (for stochastic simulations)
• Data sources:
  • Climate data: Australian Water Availability Project (AWAP) for daily mean rainfall and Morton's Wet evapotranspiration.
  • Streamflow data: Bureau of Meteorology's (BoM's) Water Data Online for observed streamflow.
  • Reservoir data: BoM's Water Data Online for reservoir volume and area (Glenlyon, Pindari, Coolmunda, Copeton dams).
  • Irrigation diversion data: Previously calibrated irrigation module parameters and NSW government's Allocations dashboard for announced allocation data.
  • NSW Macquarie IQQM model for patching missing inflow data in the Barwon-Darling.

Main Results

• DMAARS and DMAARSDS significantly outperformed the benchmark PWL model in 8 out of 12 fit metrics (including MSE, NSE, NSE(sqrt), PD(5 %tile), RMSE(<10 %tile), PD(0 %tile), POD, and ETS) when calibrated directly to the 2018/2019 environmental flow events. DMAARSDS also showed significant improvement for NSE(log).
• The new models provided more realistic estimates of environmental flow metrics, particularly for baseline conditions and ecological benefits like peak water height and flow extent, compared to PWL which often underestimated losses during "no release" scenarios.
• Scenario testing revealed that model choice significantly influences predictions of ecological benefits. For example, in a hypothetical intense release scenario (NFFb), PWL predicted flows reaching the end-of-system (0.04 GL at Wilcannia), while DMAARS and DMAARSDS only extended past Bourke.
• Validation on independent data (predicting 2018/2019 events from prior data) showed DMAARS and DMAARSDS performed similarly to PWL in overall fit metrics, but time series plots indicated better prediction of event ranges and less overfitting for the new models.
• Longer-period split-sample tests consistently demonstrated that DMAARS and DMAARSDS outperformed the PWL model over extended simulation periods, with significant improvements in several key metrics.
• The new models have been integrated into eWater Source as a community plugin, facilitating their adoption in existing basin-scale water resource models.

Contributions

• Development of parsimonious transmission loss models (DMAARS and DMAARSDS) that explicitly incorporate antecedent conditions and can be readily adopted and calibrated in basin-scale river system models.
• Novel integration of alluvium geometry and a water balance to dynamically compute depth to groundwater, reducing groundwater data requirements and simplifying implementation.
• Introduction of a river dead storage model (DMAARSDS) to account for water capture by instream pools, a critical process during droughts often overlooked in basin-scale modeling.
• Demonstrated superior predictive performance and more realistic ecological benefit assessments for environmental flow events during droughts compared to conventional piecewise linear loss models.
• Highlighted the significant impact of model choice on predictions of baseline conditions and ecological outcomes, providing a more robust tool for environmental flow management.
• Facilitated practical adoption by integrating the alluvium store model into eWater Source as a community plugin.
• Provided a framework for adapting long-term water resource management models to enable more reliable event-scale predictions across diverse hydrological conditions.

Funding

• Murray-Darling Basin Authority through the Murray-Darling Water and Environment Research Program (MD-WERP).

Citation

@article{Kim2026Improved,
  author = {Kim, Shaun Sh and Crosbie, Russell S. and Zaar, Ursula and Vaze, Jai and Wang, Bill and Mateo, Cherry May R. and May, Rebekah and Nair, Sudeep and Alam, Jahangir},
  title = {Improved river transmission loss modelling for environmental flow releases during droughts},
  journal = {Environmental Modelling & Software},
  year = {2026},
  doi = {10.1016/j.envsoft.2026.106880},
  url = {https://doi.org/10.1016/j.envsoft.2026.106880}
}