You are here:
Great Lakes modeling: Are the mathematics outpacing the data and our understanding of the system?
Pauer, J., L. Lowe, B. Rashleigh, AND W. Melendez. Great Lakes modeling: Are the mathematics outpacing the data and our understanding of the system? IAGLR meeting, Toronto, ON, CANADA, June 18 - 22, 2018.
Water quality models are useful tools to address science questions and assist with management decisions in lakes and rivers; in the Great Lakes, for example, models have been applied to improve our understanding of nutrient dynamics and to set nutrient loading targets. There has been a trend towards increasing model complexity by adding equations to describe the biogeochemistry and using increasingly smaller computational grid sizes. Some believe that these models provide a more realistic representation of the system compared to simpler models. Others have cautioned about the difficulty of knowing and reporting model performance given the limitations of estimating model accuracy and uncertainty. Here we evaluate Great Lakes models of different complexity and discuss the advantages and limitations of using simple versus complex models in the Great Lakes. Our work demonstrates that a high level of model sophistication (many equations and smaller cells) does not necessarily improve performance, but decreases transparency, making it difficult for modelers and end users to determine model accuracy and uncertainty. We propose a path forward in model development that more gradually increases complexity, to an extent that it does not compromise our ability to estimate accuracy and uncertainty. We believe that this approach will improve the models’ reputation as consensus-building tools among modelers, scientists and stakeholders.
Mathematical modeling in the Great Lakes has come a long way from the pioneering work done by Manhattan College in the 1970s, when the models operated on coarse computational grids (often lake-wide) and used simple eutrophication formulations. Moving forward 40 years, we are now running models on extremely fine computational grid resolutions and using tens if not hundreds of equations to describe the biogeochemistry. Many will argue that today’s models enable a realistic representation of the transformation, transport and fate of nutrients and phytoplankton in the lakes. Here, we will show from our own work and from analyses of published models, the pitfalls of using such sophisticated models. We will show how this level of sophistication can lead to a false sense of model accuracy, lack of transparency, and difficulty in estimating model uncertainty, resulting in a lack of faith and consensus in model results. We will discuss the need to compromise between the level of sophistication, required model observations, and model reality. We will also offer an approach to developing models with better defined accuracy, improved transparency and credibility, and greater consensus among stakeholders, and we will discuss the advantages of using community models.