HR: 0800h
AN: H21C-0700 [Abstracts]
TI: Effective Use of Time-Series Data to Calibrate a Coupled Ground Water/Surface Water Model in a Small Headwater Watershed, Northern Wisconsin
AU: * Walker, J F
EM: jfwalker@usgs.gov
AF: U.S. Geological Survey, 8505 Research Way, Middleton, WI 53562, United States
AU: Hunt, R J
EM: rjhunt@usgs.gov
AF: U.S. Geological Survey, 8505 Research Way, Middleton, WI 53562, United States
AU: Doherty, J
EM: johndoherty@ozemail.com.au
AF: Watermark Numerical Computing, 336 Cliveden Avenue
Corinda, Brisbane, 4075, Australia
AB:
A major focus of the U.S. Geological Survey's Trout Lake Water, Energy and
Biogeochemical Budgets (WEBB) project is the development of a watershed model to allow predictions of
hydrologic response to future conditions including land-use and climate change. Because of the highly
conductive nature of the outwash sand aquifer and the topography of the watershed, stream flow is dominated by
groundwater contributions; however, runoff does occur during intense rainfall periods and spring snowmelt. The
coupled ground water/surface water model GSFLOW was chosen because it could easily incorporate an existing
ground-water flow model and provides for simulation of surface-water processes.
Data collected from 1992 to 2006 in the study area include lake levels, ground-water levels and streamflow. The
frequency of data collection varies from monthly to daily; in general the more frequent data was collected over a
shorter period and during the latter portion of the period monitored. The time-series processing software
TSPROC (Doherty, 2003) was used to distill the large time-series data set to a smaller set of observations and
summary statistics that captured the salient hydrologic information. The TSPROC software was also used to
process model output, thus providing equivalent comparisons of modeled and observed variables.
Calibration targets for lake and ground-water levels included the mean and range for the entire simulation period
as well as incremental differences in monthly measurements. For wells with daily water levels, the time series
was first smoothed with a digital filter and then resampled at the middle of each month. Targets for streamflow
included monthly volumes, comparison of points on the flow-duration curve, and total streamflow smoothed with
a digital filter and then resampled at specific dates to capture the inherent variability of the observed time series.
Baseflow separation was also carried out using a recursive digital filter to separate the quick-flow component
from total streamflow; the smoothed baseflow values were resampled at specific dates to capture the seasonal
variation in baseflow. The time-series processing reduced hundreds of thousands of observations to less than
5000. This provided a distilled set of calibration targets that effectively represented the components of the system
response important for model prediction.
Reference: Doherty, J., 2003, TSPROC - A Time-Series Processor Utility: Watermark
Numerical Computing, Brisbane, Australia.
DE: 1816 Estimation and forecasting
DE: 1836 Hydrological cycles and budgets (1218, 1655)
DE: 1846 Model calibration (3333)
DE: 1847 Modeling
DE: 1872 Time series analysis (3270, 4277, 4475)
SC: Hydrology [H]
MN: 2007 Fall Meeting