Catchment Classification via Hydrologic Modeling: Evaluating the Relative Importance of Model Selection, Parameterization and Classification Techniques

Abstract:

Classification has emerged as an important tool for evaluating the runoff generating mechanisms in catchments and for providing a basis on which to group catchments having similar characteristics. These methods are particularly important for transferring models from one catchment to another in the case of data scarce regions or paired catchment studies .In many cases, the goal of catchment classification is to be able to identify models or parameter sets that could be applied to similar catchments for predictive purposes. A potential impediment to this goal is the impact of error in both the classification technique and the hydrologic model.

In this study, we examine the relationship between catchment classification, hydrologic models, and model parameterizations for the purpose of transferring models between similar catchments. Building on previous work using a data set of over 100 catchments from south-east Australia, we identify several hydrologic model structures and calibrate each model for each catchment. We use clustering to identify groups of catchments with similar hydrologic response (as characterized through the calibrated model parameters). We examine the dependency of the clustered catchment groups on the pre-selected model, the uncertainty in the calibrated model parameters, and the clustering or classification algorithm. Further, we investigate the relationship between the catchment clusters and certain catchment physical characteristics or signatures, which are more typically used for catchment classification. Overall, our work is aimed at elucidating the potential sources of uncertainty in catchment classification, and the utility of classification for improving hydrologic predictions.