An analysis of high‐flow sediment event data for evaluating model performance

Authors:
Publisher website
Google Scholar
Abstract
A methodology for utilizing high‐flow event data in model evaluation is presented. The methodology was applied to the Soil and Water Assessment Tool, version 2000 (SWAT) for the Cannonsville watershed, a rural, hilly watershed in northeastern USA, with a substantial winter snow pack. Thirty‐five high‐flow events that contained intensive total suspended solid sampling were selected in order to evaluate SWAT strengths and limitations. A comparison of calibrated model results to total sediment load data over 34 of the 35 events shows that the model underestimated the observed load by 29% (observed = 61 414 MT, predicted = 43 491 MT). In general, high sediment loading events (over 2000 metric tons) were underestimated. Most winter events were underestimated, likely due to limitations in SWAT for simulating erosion caused by snowmelt. In general, the model fit measured TSS concentrations better in the late summer/early fall than in other seasons. In addition, the model does not model erosion caused by flood waters eroding the flood plain. Additional analysis was performed for nine high‐flow events during the validation, but little conclusion could be drawn from so few events. The methodology can be used to evaluate the relative performance of the model's simulation of different processes and seasons (e.g., summer runoff events versus winter snowmelt events), provide guidance for expanding the model, and focus future calibration efforts towards controlling equations and/or parameters. Copyright © 2005 John Wiley & Sons, Ltd.

This publication has 10 references indexed in Scilit: