Abstract
Abstract. Hydrological models are important tools for effective management, conservation and restoration of forested wetlands. The objective of this study was to test a distributed hydrological model, MIKE SHE, by using bi-criteria (i.e., two measurable variables, streamflow and water table depth) to describe the hydrological processes in a forested watershed that is characteristic of the lower Atlantic Coastal Plain. Simulations were compared against observations of both streamflow and water table depth measured on a first-order watershed (WS80) on the Santee Experimental Forest in South Carolina, USA. Model performance was evaluated using coefficient of determination (R2) and Nash-Sutcliffe's model efficiency (E). The E and root mean squared error (RMSE) were chosen as objective functions for sensitivity analysis of parameters. The model calibration and validation results demonstrated that the streamflow and water table depth were sensitive to most of the model input parameters, especially to surface detention storage, drainage depth, soil hydraulic properties, plant rooting depth, and surface roughness. Furthermore, the bi-criteria approach used for distributed model calibration and validation was shown to be better than the single-criterion in obtaining optimum model input parameters, especially for those parameters that were only sensitive to some specific conditions. Model calibration using the bi-criteria approach should be advantageous for constructing the uncertainty bounds of model inputs to simulate the hydrology for this type of forested watersheds. R2 varied from 0.60–0.99 for daily and monthly streamflow, and from 0.52–0.91 for daily water table depth. E changed from 0.53–0.96 for calibration and 0.51–0.98 for validation of daily and monthly streamflow, while E varied from 0.50–0.90 for calibration and 0.66–0.80 for validation of daily water table depth. This study showed that MIKE SHE could be a good candidate for simulating streamflow and water table depth in coastal plain watersheds.
Highlights
Computer models are effective tools for understanding and quantifying watershed hydrology, but may be limited by various constraints of different model types
This study showed that MIKE SHE could be a good candidate for simulating streamflow and water table depth in coastal plain watersheds
The calibration results indicated that surface detention storage was a critical calibration parameter with substantial in1-9 fluence on water table depth and streamflow (Fig. 3a and b)
Summary
Computer models are effective tools for understanding and quantifying watershed hydrology, but may be limited by various constraints of different model types. Most hydrological models are lumped using spatially averaged conditions for the study sites (Singh et al, 1999). Geological and hydrological conditions in a large catchment or watershed may exhibit considerable spatial and temporal variability such that it can be difficult to accurately describe their hydrology using lumped hydrological models. Distributed models consider spatial variability in watersheds and are widely used. Because of the high uncertainties, distributed models may perform poorly even if they are calibrated well using data from another time period (Kirchner, 2006); similar problems can occur when models are tested against data from different study sites. Distributed models are most likely to perform better than lumped models because of their capability to utilize spatial and temporal characteristics of watersheds (Refsgaard, 1997)
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
