Sofokleous et al. (2026) Assessing future streamflow in Cyprus through hydrological model calibration under non-stationary climate and regional climate model ensemble selection

Identification

• Journal: Hydrology and earth system sciences
• Year: 2026
• Date: 2026-06-24
• Authors: Ioannis Sofokleous, George Zittis, Gerald Dörflinger, Adriana Bruggeman
• DOI: 10.5194/hess-30-3903-2026

Research Groups

• Energy, Environment & Water Research Center (EEWRC), The Cyprus Institute, Nicosia, Cyprus
• Climate and Atmosphere Research Centre (CARE-C), The Cyprus Institute, Nicosia, Cyprus
• Water Development Department, Nicosia, Cyprus

Short Summary

This study developed a framework for hydrological model parameterization under non-stationary climate conditions and applied it to 38 mountain watersheds in Cyprus, projecting a median streamflow reduction of 17% by 2030–2060, with up to 70% reduction during driest years. These findings highlight significant future water scarcity and inform adaptation strategies for the region.

Objective

• To develop a framework for the parameterization of conceptual hydrological models for robust streamflow simulations under non-stationary climate conditions.
• To bias-correct, downscale, and evaluate the performance of an 18-member Regional Climate Model (RCM) ensemble from CMIP5 models for streamflow simulations.
• To apply the optimized hydrological model parameterization with selected RCMs for 38 mountain watersheds in Cyprus, and assess the mid-term future (2030–2060) impact on the island's water resources.

Study Configuration

• Spatial Scale: 38 mountain watersheds in Cyprus (eastern Mediterranean island), ranging from 16 km² to 86 km² in area. Regional Climate Models (RCMs) at 0.11° (approximately 12 km) horizontal spatial resolution, downscaled to 1 km.
• Temporal Scale:
  • Historical/Reference Period: 1980–2015 for observations, 1980–2010 for RCMs.
  • Calibration/Validation: Multiple 5-year moving windows within 1980–1998 (calibration) and 1998–2015 (validation).
  • Future Projections: 2030–2060.
  • Data and model simulations performed at a daily time step.

Methodology and Data

• Models used:
  • Hydrological Model: GR4J (Génie Rural à 4 paramètres Journalier), a conceptual, four-parameter, daily rainfall-runoff model.
  • Climate Models: 18-member ensemble of CMIP5-driven Regional Climate Models (RCMs) from EURO-CORDEX, utilizing the RCP8.5 scenario.
  • Bias Correction and Downscaling: Quantile Delta Mapping (QDM) algorithm (implemented using the R MBC Package).
• Data sources:
  • Daily gridded precipitation and temperature observations (minimum and maximum) for Cyprus (CY-OBS) at 1 km spatial resolution (1980–2015).
  • Daily reference evapotranspiration (ET) computed using the Hargreaves equation.
  • Daily streamflow observations at the outlets of the 38 watersheds from the Water Development Department of Cyprus (1980–2015).

Main Results

• Optimal Objective Function: The Nash-Sutcliffe Efficiency (NSE) with square-root transformed streamflow values (NSEsqrt) was identified as the most robust objective function for hydrological model calibration under non-stationary climate conditions, particularly for transitions to drier periods. For drier validation periods (ΔP ≤ -5%), NSEsqrt achieved an NSE of 0.48, KGE of 0.54, and a total streamflow bias of 9%.
• RCM Ensemble Selection: Nine out of 18 RCMs were selected for future projections based on their ability to accurately reproduce the observed seasonal distribution of precipitation (ratio of precipitation in the five wettest months to annual total, W5R) and minimize streamflow bias (absolute error 1%–11%) during the 1980–2010 reference period. RCMs that underestimated wet period precipitation (W5R 60%–73% vs. observed 82%) were excluded due to high streamflow biases (up to 40%).
• Future Streamflow Projections (2030–2060, RCP8.5):
  • Median precipitation reduction: 6%.
  • Worst-case precipitation reduction: 16%.
  • Median streamflow reduction: 17%.
  • Worst-case streamflow reduction: 39%.
  • During the driest years, streamflow reductions could reach 70% relative to historical dry years, significantly exceeding long-term average reductions.
  • Watersheds with permanent (P) and intermittent pools (I-p) flow regimes are projected to experience the highest streamflow reductions (up to -47% and -42% respectively).
  • 18 out of the 38 watersheds are projected to experience a shift in their flow regime type, indicating negative impacts on aquatic ecosystems.

Contributions

• Developed and validated a novel framework for selecting optimal hydrological model parameterizations under non-stationary climate conditions, demonstrating the superior robustness of NSEsqrt for Mediterranean drying trends.
• Provided a comprehensive comparative evaluation of six objective functions (NSE, KGE, and their transformed versions) using a composite scaled score (CSS) across multiple calibration and validation periods and varying climate conditions.
• Highlighted the critical importance of evaluating Regional Climate Models (RCMs) for their ability to reproduce observed seasonal precipitation patterns, not just annual totals, to ensure reliable hydrological impact assessments.
• Delivered detailed, mid-term (2030–2060) streamflow projections for 38 mountain watersheds in Cyprus, a climate change hotspot, revealing amplified reductions in streamflow compared to precipitation, particularly during dry years, and potential shifts in flow regimes.
• Emphasized the necessity of considering extreme dry year projections for effective water resource management and climate adaptation planning, as long-term averages may mask acute water shortages.

Funding

• Cyprus Research and Innovation Foundation (3PRO-TROODOS Project; grant no. INTEGRATED/0609/061)
• EU HORIZON EUROPE Climate, Energy and Mobility program (PREVENT project; grant no. 101081276)

Citation

@article{Sofokleous2026Assessing,
  author = {Sofokleous, Ioannis and Zittis, George and Dörflinger, Gerald and Bruggeman, Adriana},
  title = {Assessing future streamflow in Cyprus through hydrological model calibration under non-stationary climate and regional climate model ensemble selection},
  journal = {Hydrology and earth system sciences},
  year = {2026},
  doi = {10.5194/hess-30-3903-2026},
  url = {https://doi.org/10.5194/hess-30-3903-2026}
}