Manfreda et al. (2026) Zero-flow and hydrotype estimation with karst-SWAT and Sentinel-2 data in the Keritis Basin, Crete

Identification

• Journal: Journal of Hydrology Regional Studies
• Year: 2026
• Date: 2026-04-10
• Authors: G. Manfreda, C. Cavallo, M.A. Lilli, A. Maragkaki, G. Negro, M.N. Papa, N.P. Nikolaidis, P. Vezza
• DOI: 10.1016/j.ejrh.2026.103431

Research Groups

• Department of Environment, Land and Infrastructure Engineering, Polytechnic University of Torino, Torino, Italy
• Department of Civil Engineering, University of Salerno, Fisciano, SA, Italy
• School of Chemical and Environmental Engineering, Technical University of Crete, Chania, Greece
• INGENIO CSIC-UPV, Universitat Politecnica de Valencia, Val`encia, Spain

Short Summary

This study developed a novel methodology integrating Karst-SWAT hydrological modeling with Sentinel-2 satellite imagery to provide daily forecasts of flow conditions and estimate hydrotypes in two non-perennial reaches of the Keritis River, Greece. The approach significantly improved the understanding of flow intermittency patterns, revealing that deep aquifer recharge and evapotranspiration are key controls on zero-flow events.

Objective

• To develop and apply a novel methodology integrating Karst-SWAT hydrological modeling with Sentinel-2 satellite imagery to provide daily forecasts of flow conditions (flowing, ponding, dry) and estimate the hydrotype of non-perennial river reaches.

Study Configuration

• Spatial Scale: Keritis Basin, Crete Island, Greece (278 km²). Specifically, two non-perennial reaches of the Keritis River: Reach 1 (R1, 1480 m length, 30 m average width) and Reach 2 (R2, 1530 m length, 40 m average width).
• Temporal Scale: Satellite imagery analysis from March 2019 to December 2023. Hydrological model simulation for daily streamflow from 1971 to 2023. Meteorological data from 1961 to 2023. Streamflow data for calibration/validation from various periods between 1971 and 2015.

Methodology and Data

• Models used:
  • Karst-SWAT model (Soil and Water Assessment Tool coupled with a karst-specific model)
  • Classification and Regression Tree (CART), specifically Decision Stump (DS), for flow rate threshold determination
  • Synthetic Minority Over-sampling Technique (SMOTE) for handling unbalanced datasets
• Data sources:
  • Satellite: Sentinel-2 (S2) imagery (March 2019 – December 2023, Level-2A, using B11, B8, B4 for classification). PlanetScope (PS) imagery (2020 and 2023, 3 m spatial resolution) for error assessment. Google Earth Pro high-resolution images for shadow identification.
  • Observation/Reanalysis: Hydro-meteorological data (rainfall, temperature from four meteorological stations, 1961-2023). Streamflow data (temporary gauging station on Keritis River, Meskla spring, Agia springs, Kalamionas spring, various periods between 1971-2015). Digital Elevation Model (DEM, 0.025 km spatial resolution). CORINE land use map. Harmonized World Soil Database (HWSD) digital soil map (1 km resolution).

Main Results

• The integrated approach successfully provided daily forecasts of flow conditions (Flowing, Ponding, Dry) and hydrotype estimations for the two non-perennial reaches (R1, R2) of the Keritis River.
• The average Effective Revisit Time (ERT) for S2 imagery was 8.3–8.7 days, with 417 usable cloud-free images out of 699 total acquisitions.
• The Karst-SWAT model, despite achieving satisfactory performance for low-flows in some karst springs (e.g., Meskla spring validation: Nash-Sutcliffe efficiency (NSE) = 0.54, logarithmic NSE (LNSE) = 0.61, root mean square error observations standard deviation ratio (RSR) = 0.68, percent bias (PBIAS) = -31.8%), consistently failed to reproduce zero-flow conditions when used alone.
• Annual flow rate thresholds (T1 for Flowing to zero-flow, T2 for Ponding to Dry) were determined, showing interannual variability. T1 classification accuracy ranged from 0.90 to 1.00, while T2 accuracy ranged from 0.83 to 0.95.
• Both studied reaches (R1 and R2) were classified as Intermittent–Fluent hydrotypes over the 2019–2023 period, but exhibited pronounced interannual variability. R1 shifted to an Ephemeral hydrotype in 2023 due to drought conditions.
• Water balance analysis indicated that deep aquifer recharge (approximately 30-35% of annual precipitation) primarily controls the annual occurrence of drying events, while evapotranspiration (ET, 19-40% of annual precipitation) primarily modulates the frequency and magnitude of zero-flow peaks.
• The methodology demonstrated robustness with high Overall Accuracy (OA > 0.90) when compared to independent PlanetScope data, and classification errors did not alter the assigned hydrotype.

Contributions

• Developed a novel, replicable, and robust integrated methodology combining Karst-SWAT hydrological modeling and Sentinel-2 satellite imagery for daily flow condition forecasting and hydrotype estimation in non-perennial rivers.
• Significantly improved the understanding of flow intermittency patterns at a daily scale, especially in complex karst systems where traditional models and gauging stations are limited.
• Provided new hydrological insights for the Keritis Basin, identifying deep aquifer recharge as the primary control for annual drying events and evapotranspiration for modulating zero-flow peak frequency and magnitude.
• Demonstrated a method to overcome data scarcity and structural model limitations in complex hydrogeological settings, offering a valuable tool for hydro-ecological assessments and long-term water resource management.
• Showcased the potential for scaling up this approach to regional or continental scales for water resource management and compliance with directives like the Water Framework Directive (WFD).

Funding

• RIVERTEMP project (2022–1-IT02-KA220-HED-000086223), co-funded by the Erasmus+ program of the European Commission.
• Research project ‘Implementation of Ecological Flows in the Watercourses of the Po River Basin District’ financed by the Po River Basin Authority (Italy).

Citation

@article{Manfreda2026Zeroflow,
  author = {Manfreda, G. and Cavallo, C. and Lilli, M.A. and Maragkaki, A. and Negro, G. and Papa, M.N. and Nikolaidis, N.P. and Vezza, P.},
  title = {Zero-flow and hydrotype estimation with karst-SWAT and Sentinel-2 data in the Keritis Basin, Crete},
  journal = {Journal of Hydrology Regional Studies},
  year = {2026},
  doi = {10.1016/j.ejrh.2026.103431},
  url = {https://doi.org/10.1016/j.ejrh.2026.103431}
}