Surrogate modelling approach for modelling peak discharge in rainwater harvesting tanks: case study in Quito, Ecuador
4 affiliations
- 1Universidad Estatal Península de Santa Elena
- 2University of Cartagena
- 3Fundación Universitaria Tecnológico Comfenalco Cartagena
- 4University of the Coast
Abstract
Urbanisation increases surface runoff and recurrent flooding in cities, mainly by reducing infiltration capacity and shortening times of concentration. Rainwater harvesting tanks can store water to mitigate peak flows from urban development and support sustainable planning for new buildings. This study presents a hydrological routing approach for rainwater harvesting tanks, solved using the Rosenbrock method, combined with Monte Carlo analysis to generate a sufficiently large, representative set of hydrological scenarios. Based on these simulations, we developed multiple machine learning (ML) models to provide a simplified, efficient predictive framework. A total of 200 scenarios were generated using Latin Hypercube Sampling by varying key input parameters, including rainfall intensity, pipe length, internal pipe diameter, and tank surface area. In total, 28 ML regression models were evaluated to estimate peak outflow discharge. The Cubic Support Vector Machine (SVM) exhibited the best performance, achieving coefficients of determination ( R2 R 2 ) of 0.984 and 0.985 for the validation and testing stages, respectively. Sensitivity analysis revealed that the internal pipe diameter is the most influential parameter controlling peak discharge. The proposed methodology offers a computationally efficient, reliable tool for predicting the hydraulic performance of rainwater harvesting tanks, supporting decision-making for urban planners and local authorities in flood mitigation strategies.
Related
- Integrated machine learning and AHP-based framework for spatial delineation of rainwater harvesting potential
- Water-nitrogen Management under supplementary irrigation through rainwater harvesting technique for improving soil fertility, tomato yield and water productivity in the West African Sahel
- Leveraging satellite data and machine learning to quantify rooftop rainwater harvesting reliability and tank requirements for schools and health care facilities in Uganda
- Sequential rainwater chemistry constrains in-cloud and below-cloud scavenging at a high-elevation site under contrasting transport regimes
- Dataset on water quality and trophic status assessment of San Martín Lake (Ecuador) using NSF, OWQI, CCME, OECD, Carlson and TRIX approaches (2021-2022).
BibTeX
@article{guarda2026surrogate,
title = {{Surrogate modelling approach for modelling peak discharge in rainwater harvesting tanks: case study in Quito, Ecuador}},
author = {Teresa Guarda and Oscar E. Coronado-Hernández and Manuel Saba and Germán Herrera-Vidal and Jairo R. Coronado-Hernández},
journal = {Water Science},
year = {2026},
month = oct,
doi = {10.1007/s44533-026-00089-7},
url = {https://doi.org/10.1007/s44533-026-00089-7},
note = {CC BY},
}
RIS
TY - JOUR
TI - Surrogate modelling approach for modelling peak discharge in rainwater harvesting tanks: case study in Quito, Ecuador
AU - Teresa Guarda
AU - Oscar E. Coronado-Hernández
AU - Manuel Saba
AU - Germán Herrera-Vidal
AU - Jairo R. Coronado-Hernández
PY - 2026
DA - 2026/10/01
JO - Water Science
DO - 10.1007/s44533-026-00089-7
UR - https://doi.org/10.1007/s44533-026-00089-7
AB - Urbanisation increases surface runoff and recurrent flooding in cities, mainly by reducing infiltration capacity and shortening times of concentration. Rainwater harvesting tanks can store water to mitigate peak flows from urban development and support sustainable planning for new buildings. This study presents a hydrological routing approach for rainwater harvesting tanks, solved using the Rosenbrock method, combined with Monte Carlo analysis to generate a sufficiently large, representative set of hydrological scenarios. Based on these simulations, we developed multiple machine learning (ML) models to provide a simplified, efficient predictive framework. A total of 200 scenarios were generated using Latin Hypercube Sampling by varying key input parameters, including rainfall intensity, pipe length, internal pipe diameter, and tank surface area. In total, 28 ML regression models were evaluated to estimate peak outflow discharge. The Cubic Support Vector Machine (SVM) exhibited the best performance, achieving coefficients of determination ( R^2 R 2 ) of 0.984 and 0.985 for the validation and testing stages, respectively. Sensitivity analysis revealed that the internal pipe diameter is the most influential parameter controlling peak discharge. The proposed methodology offers a computationally efficient, reliable tool for predicting the hydraulic performance of rainwater harvesting tanks, supporting decision-making for urban planners and local authorities in flood mitigation strategies.
SN - 2357-0008
SN - 1110-4929
ER -