Changes over time in the relationship between rainfall and catchment runoff pose a significant challenge for hydrological models, which are often calibrated under the assumption that the future relationship will be consistent with that of the past. In a recent paper, Westra et al. outlined a method for diagnosing, interpreting, and improving the capacity of models to develop predictions under such conditions.
Using data from a 29-square-kilometer area of land in southern Australia known as the Scott Creek catchment as a case study, the authors investigated 22 model parameterizations. They calibrated the flow data from 1985 to 1999; then, they used the models to predict flows from 2000 to 2009—a much drier decade. The standard hydrological model overpredicted the amount of flow in the 2000s by 18%.
To better predict long-term changes, the authors identified parameters that varied seasonally and interannually and identified ways to better represent the physical catchment processes in the model. They also used a range of statistical tools to select the most accurate models for a changing climate. Through this method, they were able to select and adjust the models so that they overestimated the flow during the dry period by only 3%–7%. The authors suggest that an ensemble of models should be used to capture the range of possibilities in an uncertain future. (Water Resources Research, doi:10.1002/2013WR014719, 2014)
—Shannon Palus, Writer
Citation: Palus, S. (2014), Improving hydrology models for a changing climate, Eos Trans. AGU, 95(50), 484, doi:10.1002/2014EO500012.