Uncertainty Assessment of Percolation Groundwater Recharge Estimated with the SWAT Hydrological Model: A Case Study for the Fetrek Watershed (Izmir, Turkiye)


Elçi A., Çakır F., Somay Altaş A. M.

2023 SWAT Conference, Arhus, Denmark, 26 - 30 June 2023, pp.1

  • Publication Type: Conference Paper / Summary Text
  • City: Arhus
  • Country: Denmark
  • Page Numbers: pp.1
  • Dokuz Eylül University Affiliated: Yes

Abstract

Hydrological models play an important role in the management of water resources. However, their predictions are inherently prone to multiple sources of uncertainty, which can significantly impact the accuracy and reliability of model results. Various methods, such as uncertainty analyses and ensemble modeling are developed to quantify uncertainty. This paper presents an assessment of uncertainties in groundwater recharge rates that were simulated using the SWAT model. This assessment is different from the usual approach taken within the SWAT-CUP calibration framework and resembles a simplified ensemble modeling approach. The model is applied to the Fetrek Stream watershed in Izmir (Turkiye), which is under environmental stress due to excessive groundwater abstraction and pollution from numerous wastewater discharges. The model includes flow contributions from 41 point sources and two inflowing tributaries. The watershed is partitioned into 8 sub-basins and 484 hydrologic response units. Monthly hydrological fluxes are obtained for a 30-year simulation period. The model is calibrated using the SUFI-2 algorithm while model performance is evaluated with the Kling-Gupta Efficiency (KGE), the regression coefficient (R2), and the bias percentage (PBIAS). Modeled groundwater recharge rates are presented with their associated prediction uncertainties in the form of uncertainty bands, which were determined based on an ensemble of model outcomes. Median annual groundwater recharge rates vary between 160 – 350 mm/yr. The lower and upper bounds of the estimation uncertainty were on average -13% and +38.7% around the median rate for the largest sub-basin, respectively. The assessment shows that the presented approach can be more flexible than the conventional p-factor and r-factor statistics approach when the number of behavioral solutions is not sufficient.