394868 - Target Calibration Parameters for Flood Frequency Distribution Simulation using SWAT Model
Monday, June 4
4:00 PM - 5:30 PM
Location: Greenway IJ
Sara Mille, Houghton, MI – Michigan Technological University
Changes in climate and landuse impact flood risk, however, standard techniques for flood frequency analysis are based on the assumption that annual maximum flood (AMF) series are stationary. A reasonable alternative to traditional statistical analysis is use of a distributed hydrologic model to understand flood risk under altered climate and landscape conditions. Unfortunately, these physical models have been shown to exhibit bias with a tendency to underestimate peak flows, thereby increasing uncertainty in flood risk assessment. This study seeks to identify the parameters with most influence on the flood risk distribution resulting from simulation using the Soil and Water Assessment Tool (SWAT). Hydrologic models of case study watersheds were built using SWAT with calibration and validation based on daily historical observations of climate and streamflow for the period spanning 2000-2009, then employed to simulate daily streamflows for water years 1970-1999. The daily streamflows were converted to instantaneous AMF series using the method of Fill and Steiner (2003). The characteristics of the simulated AMF series were compared against those of gauged AMF series for the same period, and the impact of calibration on the simulated flood risk distribution for each case study watershed was investigated by perturbing each parameter over a reasonable range of values. Overall, sensitivity analysis reveals that AMFs in watersheds that are predominantly used for agriculture are highly influenced by runoff related parameters and thus would be more sensitive to changes in land use and land cover. On the other hand, AMFs in heavily forested watersheds are influenced more by parameters related to climate. In both cases, effects were more significant on flows corresponding to larger return periods. Moreover, significant differences are observed for even the 25-year event, a relatively small magnitude flow, but of interest for multiple design purposes.