An Informal Survey of Watershed Model Users: Preferences, Applications, and Rationales

Making a selection from a smorgasbord of watershed models can be challenging, and it is interesting and potentially valuable to see how water resource science and engineering professionals make these choices. Introduction If you ask 20 hydrologists for their opinions on watershed models, you are likely to receive 20 different answers. This seems a well-known trait of the watershed modelling community, but it is perhaps not so widely understood that the situation is somewhat unique to the discipline. Though intellectual diversity is healthy and indeed needed for science to progress, most other fields of science and engineering seem a little more unified and coherent in their modelling philosophies and tools. Consider the closely allied field of groundwater modelling, for instance. There is a consensus on how groundwater models should be constructed, such as agreement that models should in general solve a governing differential equation, and agreement on what that equation is. Differences between groundwater modelling packages lie instead with important, but strictly secondary, issues such as the specific numerical solution methods adopted (e.g., analytical vs. finite-difference vs. finite-element), additional physical complications accommodated (e.g., wetlands), or preand post-processing utilities or graphical user interfaces (GUIs). In contrast, fundamental modelling concepts and methods may be quite different between watershed models (e.g., IHACRES vs. SHETRAN). Further, there has been a remarkable proliferation of process-oriented watershed modelling packages, each representing a different tack on the problem. The reason for this lack of a consistent view on how to model the transformation of meteorological forcing to streamflow seems straightforward. Such a tool is not a mathematicalcomputational representation of a process, but a representation of very many processes, both physical and biological; most of those individual constituent processes are themselves inherently difficult to model and/or constrain by data; and the operative, important processes vary from one region, and even one catchment, to the next. On top of this, watershed hydrology spans a wide range of goals, from fundamental scientific enquiry into the nature of the world around us to very practical, focused, applied questions. Each goal has its own set of requirements: a forest hydrology research model can afford to favour high detail over fast run times, for instance, whereas in general an operational river forecast model cannot. The net effect of all these considerations is a plethora of available models to choose from (e.g., Table 1 of Singh and Woolhiser 2002; see also Beckers et al. 2009a–c).