Can et al. (2025) Evaluating the impact of subsurface hydraulic barriers on Qanat flow rates using quantile regression forest

Identification

• Journal: Scientific Reports
• Year: 2025
• Date: 2025-11-22
• Authors: Murat Can, Babak Vaheddoost, Mir Jafar Sadegh Safari
• DOI: 10.1038/s41598-025-28693-0

Research Groups

• State Hydraulic Works of Turkey, 1st District, Bursa Office
• Department of Civil Engineering, Bursa Technical University, Bursa, Turkey
• Department of Geography and Environmental Studies, Toronto Metropolitan University, Toronto, Ontario, Canada
• Department of Civil Engineering, Yaşar University, Izmir, Turkey

Short Summary

This study evaluated the impact of a subsurface dam on Qanat flow rates using machine learning models, finding that the dam significantly and positively influences discharge, with Quantile Regression Forest (QRF) demonstrating superior predictive performance (Nash–Sutcliffe Efficiency = 0.818).

Objective

• To evaluate the hydraulic performance and discharge capacity of the Dirsak Qanat system in Sufi Village, Iran, following the installation of a subsurface dam, using machine learning models.

Study Configuration

• Spatial Scale: Dirsak Qanat in Sufi Village, Maku, West Azerbaijan Province, Iran, within the Sufi River Basin (approximately 45 km²).
• Temporal Scale: Monthly data from April 2002 to March 2008. Modeling period from April 2003 to March 2008, with 48 months for training (April 2003 - March 2007) and 12 months for testing (April 2007 - March 2008).

Methodology and Data

• Models used: Quantile Regression Forest (QRF), Random Forest (RF), Support Vector Regression (SVR).
• Data sources:
  • Climate (Meteorological): Monthly precipitation (mm), temperature (°C), evaporation (mm), and humidity (%). Data from meteorological stations, spatially averaged using Thiessen Polygons.
  • Groundwater (Hydrogeological): Water levels (mASL) from three observation wells (P1, P2, P3).
  • Qanat's Discharge (Hydraulic): Dirsak Qanat discharge (l/s) measured using a rectangular sharp-edged weir.
  • Runoff and Infiltration Depths (Hydrological): Runoff depth (mm) calculated via the Soil Conservation Service (SCS) Curve Number (CN) method. Infiltration depth (mm) calculated using a simple budget approach (Precipitation - (Runoff - Potential Evapotranspiration)), with Potential Evapotranspiration (PET) from the Thornthwaite equation.
  • Dummy Variable (Mathematical): A binary variable (0 or 1) representing the absence or presence of the subsurface dam.

Main Results

• The subsurface dam (SD) and evaporation were identified as the most influential factors affecting Qanat discharge.
• The dummy variable (representing SD presence) was the most important variable in the QRF model, followed by evaporation and runoff depth.
• The QRF model outperformed SVR and RF in predicting Qanat discharge, achieving a Nash–Sutcliffe Efficiency (NSE) of 0.818, Root Mean Square Error (RMSE) of 1.173, and Akaike Information Criterion (AIC) of 23.823 in the testing stage.
• The SD positively influences the Qanat discharge rate, with post-construction data showing increasing trends compared to pre-construction declines.
• QRF provided robust predictions with prediction intervals (5th to 95th percentiles), effectively capturing uncertainty.
• Multicollinearity analysis led to the exclusion of temperature from the model inputs, as evaporation was retained and showed a strong correlation with temperature.

Contributions

• Utilized rare and valuable Qanat data from the Sufi site, one of the few documented and still-functioning systems.
• Employed advanced artificial intelligence methods (SVR, RF, QRF) to assess Qanat hydraulic behavior and performance.
• Incorporated a dummy variable approach to simulate boundary conditions and analyze the interaction between the Qanat and the constructed SD.
• Presented an applicable and replicable methodological framework for the conservation and modernization of historical aqueducts.
• Evaluated the lessons learned from the study based on the UN Sustainable Development Goals (SDGs) for the first time.

Funding

The authors received no funding for this work.

Citation

@article{Can2025Evaluating,
  author = {Can, Murat and Vaheddoost, Babak and Safari, Mir Jafar Sadegh},
  title = {Evaluating the impact of subsurface hydraulic barriers on Qanat flow rates using quantile regression forest},
  journal = {Scientific Reports},
  year = {2025},
  doi = {10.1038/s41598-025-28693-0},
  url = {https://doi.org/10.1038/s41598-025-28693-0}
}