Kurugama et al. (2026) Augmenting observation network design and assimilation frequency in distributed hydrological models: insights from the LISFLOOD-based hydrological data assimilation framework

Identification

• Journal: Journal of Hydrology
• Year: 2026
• Date: 2026-01-04
• Authors: Kumudu Madhawa Kurugama, So Kazama, Yusuke Hiraga
• DOI: 10.1016/j.jhydrol.2025.134853

Research Groups

• Department of Civil and Environmental Engineering, Tohoku University, Japan
• Joint Research Centre (JRC), European Commission, Italy (for LISFLOOD model development)

Short Summary

This study developed a LISFLOOD-based hydrological data assimilation framework (LISFLOOD-HDAF) coupling LISFLOOD with an Ensemble Kalman Filter (EnKF) to evaluate the impact of assimilation frequency and observation network design on streamflow prediction. It found that EnKF consistently improved predictions, with non-monotonic frequency effects and strong dependence on gauge density and placement, enabling cost-effective network design.

Objective

• To investigate the combined influence of streamflow observation network design (gauge density and spatial placement) and assimilation frequency on hydrological data assimilation performance using a LISFLOOD-based Ensemble Kalman Filter framework across contrasting hydrological regimes.

Study Configuration

• Spatial Scale: A 4,975 km² sub-catchment of the Upper Po River Basin, Italy, upstream of Pieve del Cairo, with a spatial resolution of 5 km × 5 km.
• Temporal Scale: Six representative wet and dry hydrological events from 2013 to 2017, using meteorological forcing data from January 2, 2010, to December 31, 2017. Three assimilation frequencies (6-hour, 12-hour, and 24-hour intervals) and eight contrasting observation network configurations were tested.

Methodology and Data

• Models used: LISFLOOD hydrological model (version 4.3.1) coupled with an Ensemble Kalman Filter (EnKF).
• Data sources: Synthetic twin experiments using a "nature run" generated by LISFLOOD. Input data included the official LISFLOOD sample dataset, comprising 6-hourly meteorological forcing (precipitation, temperature, potential evapotranspiration) and static physiographic maps (land cover, soil type, local drainage direction network, elevation, channel geometry).

Main Results

• EnKF assimilation consistently improved streamflow simulation performance compared to the open-loop run, with the most significant gains observed during wet periods and at the downstream outlet.
• At the downstream outlet during wet periods, the normalized Kling–Gupta efficiency assimilation index (KGEAI) reached 0.80–0.90, and root mean square error (RMSE) reductions exceeded 50 %.
• The impact of assimilation frequency was non-monotonic: 12-hour intervals yielded the most robust improvements at midstream and downstream locations. Upstream headwaters benefited from 6-hour updates in wet conditions and 24-hour updates in dry conditions.
• Assimilation performance strongly depended on gauge density and placement, with dense or strategically distributed networks performing best, though with diminishing returns beyond a certain density.
• Performance-cost analysis showed that five gauges with 12-hour updates provided near-optimal skill in wet regimes, while three to four gauges sufficed in dry conditions with 24-hour updates.
• Strategically locating gauges at tributary junctions, mid-basin control points, and downstream confluences can achieve nearly the same benefits as full high-density networks.

Contributions

• Development of the first open-source implementation enabling real-time streamflow assimilation within the LISFLOOD modeling environment (LISFLOOD-HDAF).
• Systematic factorial assessment of the isolated and joint impacts of gauge-network density, spatial configuration, and assimilation interval across contrasting hydrological regimes using synthetic twin experiments.
• Provides practical, decision-oriented guidance for cost-effective gauge network design, supporting rationalization of dense networks in data-rich regions and strategic design of minimum-viable networks in poorly gauged basins.
• Introduces a transparent and transferable methodological pathway (blocked experimental design, nonparametric inference, normalized indices, and knee-point synthesis) for deriving defensible design recommendations for hydrological data assimilation.

Funding

• "Big Data Platform for Water Environment Monitoring and Data-Driven Policy Implementation (CREWS)" project under the Science and Technology Research Partnership for Sustainable Development (SATREPS), jointly funded by the Japan Science and Technology Agency (JST) and the Japan International Cooperation Agency (JICA).

Citation

@article{Kurugama2026Augmenting,
  author = {Kurugama, Kumudu Madhawa and Kazama, So and Hiraga, Yusuke},
  title = {Augmenting observation network design and assimilation frequency in distributed hydrological models: insights from the LISFLOOD-based hydrological data assimilation framework},
  journal = {Journal of Hydrology},
  year = {2026},
  doi = {10.1016/j.jhydrol.2025.134853},
  url = {https://doi.org/10.1016/j.jhydrol.2025.134853}
}