You are here:
HYDROLOGIC MODEL CALIBRATION AND UNCERTAINTY IN SCENARIO ANALYSIS
Citation:
SEMMENS, D. J. AND M. HERNANDEZ. HYDROLOGIC MODEL CALIBRATION AND UNCERTAINTY IN SCENARIO ANALYSIS. Presented at American Water Resources Association Annual Conference, Albuquerque, NM, November 12 - 15, 2007.
Description:
A systematic analysis of model performance during simulations based on
observed land-cover/use change is used to quantify error associated with water-yield
simulations for a series of known landscape conditions over a 24-year period with the
goal of evaluating methodologies for forecasting future hydrologic response from
baseline conditions (scenario assessment). Simulations were carried out using the Soil
and Water Assessment Tool (SWAT) with parameter inputs derived from the Automated
Geospatial Watershed Assessment (AGWA) modeling interface. Calibrated and
uncalibrated assessments of absolute and relative change in water yield over different
lengths of time, and based on different precipitation inputs, are presented and compared
to observed values. Results indicate that climatic uncertainty is the largest source of error
in assessments of future landscape conditions, and for the study area natural climatic
variability had a substantially greater impact on basin water yield than land-use change.
Quantitative predictions of absolute water-yield change are only possible with a model
calibrated to initial, baseline conditions and when future climatic conditions are precisely
known. The use of statistically derived rainfall input intended to approximate the
observed "future" conditions improved trend predictions, but failed to result in
acceptable, quantitative predictions of water yield. For predictions of change in average
annual water yield relative to baseline conditions, however, uncalibrated simulations
performed equally well as the initially calibrated simulations in all cases. Analyses
designed to forecast for comparison the hydrologic impacts associated with a range of
different possible future land-cover/use scenarios can thus be reliably carried out when
calibration is not possible.