A Systematic Assessment of the Relationship Between the Complexity and Fidelity of Hydrological Models

Abstract:

The relationship between the complexity and fidelity of hydrological models is challenging to investigate in a systematic way using current modeling frameworks. Its characterization has so far principally relied on the comparison of different models or of different modules within the same model. Shortcomings of these approaches include the difficulty to pinpoint model features that contribute to good simulations, given the small number of models or modeling hypotheses that are usually evaluated. We use the newly-developed Structure for Unifying Multiple Modeling Alternatives (SUMMA) to comprehensively and systematically explore modeling alternatives across the continuum of model complexity. We use SUMMA’s flexibility to evaluate the impacts of explicitly representing or lumping physical processes and hydrological landscapes. Starting from conceptual models based on the Framework for Understanding Structural Errors (FUSE), we progressively increase model complexity and assess corresponding model fidelity. We scrutinize models’ ability to reproduce observed events and the stability of their performance under changing climatic conditions (robustness). We will show preliminary results for catchments in different hydroclimatic regimes simulated using models of varying complexity. As a first step, model complexity will be quantified using computing time and the number of state variables; model robustness will be quantified using differential split-sample tests; and model performance will be quantified using a suite of multivariate and multi-scale diagnostic metrics. With this modeling approach we seek to uncover trade-offs between realism and practicality. A particular aim is to explore to which extent the replacement of conceptual formulations by physically explicit ones improves model performance, and whether this may lead to a reduction of uncertainty in hydrological simulations.