Streams in the southeastern U.S. are among the most ecologically rich in the world, but climate change, land cover change, and withdrawals threaten the health of their aquatic ecosystems.
“Understanding how changes in streamflow affect aquatic wildlife is critical,” says U.S. Forest Service scientist Peter Caldwell. “Many states in the Southeast recognize the urgency of the issue and are taking steps to manage water uses in the interest of protecting aquatic ecosystems.”
Hydrologic models are important tools that managers can use to predict how streamflow may change as a result of changes in climate, land cover, and withdrawals. There is a wide range of models available, from highly detailed theoretical models that can be applied to small watersheds to simpler conceptual models that can be applied to large regions.
“However, there haven’t been many studies comparing these models,” says Caldwell. Caldwell is a research hydrologist at the Forest Service Southern Research Station (SRS) Coweeta Hydrologic Laboratory, and lead author of a recent study that compares differences in streamflow prediction among six hydrologic models. The study was recently published in Ecohydrology.
Caldwell and his colleagues compared the Water Supply Stress Index model (WaSSI) developed by the SRS Eastern Forest Environmental Threat Assessment Center; the Hydrological Simulation Program-Fortran; the Monthly Water Balance Model developed by the U.S. Geological Service; the Soil and Water Assessment Tool developed by the U.S. Department of Agriculture; the Precipitation-Runoff Modelling System developed by the U.S. Geological Service; and WaterFall®, which was developed by the Research Triangle Institute.
Some of the models were intensively adjusted, or calibrated, to achieve the best match to measured streamflow while others (including WaSSI) were not calibrated and instead relied on the accuracy of publicly available inputs such as weather and soil information to predict streamflow. “We weren’t trying to identify which model was ‘best,’” says Caldwell. “Differences in model accuracy are as likely to be related to how the model was set up and calibrated as they are to differences in model structure.” Instead, the scientists wanted to quantify and compare the magnitude and potential causes of error among the various models.
The scientists compared the models across five study sites in the Apalachicola-Chattahoochee-Flint basin, which drains over 19,000 square miles across Alabama, Florida, and Georgia. The basin supplies drinking water to millions of people, and water shortages have led to conflicts between states that rely on the basin. “The basin has been intensely studied over the last decade,” says Caldwell. “Multiple modeling efforts have evaluated drought and global change effects on aquatic ecosystems and water supply in the basin, providing a unique opportunity to compare models at the same sites but developed to answer a wide range of management questions.”
The scientists had expected that simpler uncalibrated models would not be able to predict streamflow as well as more complicated models. However, when predicting monthly streamflow, the scientists found that all the models were comparable and fairly accurate. Based on the study, the scientists don’t recommend any one model over the others.
“We found that model calibration is as important as the specific type of model used,” says Caldwell. “It appears that the person doing the modeling, and the data they have available, is as important for predicting streamflows as the specific model they use.”
For more information, email Peter Caldwell at firstname.lastname@example.org.