Increasing physical realism in hydrological models
Martyn Clark is a senior scientist at NCAR (National Centre for Atmospheric Research)
Although enshrined in hydrology many of the issues Martyn is tackling on model realism and uncertainty is relevant to other disciplines in environmental modelling.
The hydrological modeling community has organized themselves into two camps: The "conceptual" modelers, who strive for parsimonious descriptions of observed processes and seek to retain flexibility in model parameters and structure to suit the idiosyncrasies of individual catchments; and "physics-based" modelers, who strive for a physically realistic representation of all dominant hydrological processes albeit with limited flexibility in model structure and parameters. Such differences in modeling philosophies are manifest in a different degree of prior knowledge: physics-based modelers often rely on physical understanding and established process descriptions, whereas conceptual modelers often rely on use of inverse methods to infer appropriate model structure and parameters. In short, physics-based modelers typically (and implicitly) assume they know more about system function than conceptual modelers.
This presentation will explore the development of physically realistic hydrological models from an uncertainty perspective. We will carefully delineate the major (subjective) decisions made when building a model, and evaluate different modeling alternatives available at each decision-point. Based on this analysis, we will introduce a new "master modeling template", from which existing physics-based models can be reconstructed and new models derived. We suggest the new modeling template can unify the conceptual and physics-based modeling approaches, by providing scope to scrutinize physics-based modeling assumptions in more detail and scope to infer the most suitable modeling approaches using different signatures of hydrological behavior.