Flux Mapping: a new approach to evaluate process-based models and generate model-based hypotheses
Abstract
While scientific hypotheses are our best tools to represent and understand real world phenomena, we know a priori that all hypotheses are incomplete and uncertain. Particularly in domains such as Earth and environmental sciences, given the exuberance of uncertainty and inexactness associated with every step of the scientific inquiry. No single hypothesis represents all aspects of a real-world phenomenon, nor serves all purposes. Therefore, the most defensible approach of scientific inquiry is the paradigm of multiple working hypotheses (MWH): A range of plausible hypotheses are continuously generated, evaluated, and refined towards better approximating reality. A conceptual model is a hypothesis. Utilizing conceptual models within the MWH paradigm implies generating an ensemble of plausible models. Such ensembles could be produced through variations of model components, namely model structures and parameter sets, input data, model fluxes, objective functions, etc. The question then becomes how to evaluate these models to generate MWH? Here, we demonstrate that the conventional methods of evaluating model performance and parameter uncertainty are not insightful enough. When evaluating process-based models, we should develop process-based approaches to then generate model-based hypotheses. To this end, we developed the model evaluation method Flux Mapping. We map the dynamics of model internal fluxes to generate model-based hypotheses. We demonstrate how flux mapping gives new insights into model behavior that cannot be inferred from conventional methods. We show that even within a narrow margin of model error, different modes of model response can be equally active. That is, different combinations of model internal fluxes can equally reproduce a given observed hydrograph within a narrow margin of model error. Such model-based hypotheses can be tested against one another, and/or other hard/soft data of the real-world system. Hypotheses based on model internal behaviour, and not just the model output, hopefully encourage the collection and utilization of field data to evaluate the model process-representation, and better direct our efforts for improving the model realism. Flux Mapping can be extended to any domain that deals with conceptual modeling of open complex systems under uncertainty.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2019
- Bibcode:
- 2019AGUFM.H13Q1998P
- Keywords:
-
- 1847 Modeling;
- HYDROLOGY;
- 1873 Uncertainty assessment;
- HYDROLOGY;
- 1880 Water management;
- HYDROLOGY;
- 1916 Data and information discovery;
- INFORMATICS