USGS Grant Number: G21AP10178
Year Established: 2020 Start Date: 2020-09-01 End Date: 2023-08-31
Total Federal Funds: $44,193 Total Non-Federal Funds: $249,993
Principal Investigators: Trisha Moore
Abstract: We aim to couple two modeling approaches â€“ mechanistic and data-driven models â€“ to address shortcomings of both with respect to understanding and predicting toxin-producing cyanobacteria blooms. The overall goal of this approach is to provide a prediction framework that overcomes the site-specific nature of existing cyanoHAB models and that, by accounting for mechanisms that drive cyanobacterial growth and toxin production, enables causative factors to be more readily identified and managed. Mechanistic approaches aim to predict cyanobacterial blooms via mathematical models in which biological processes underlying phytoplankton growth and/or toxin release are represented and are thereby valuable for understanding a predicting toxic blooms. However, model parameter values are generally highly uncertain and may not accurately represent the myriad of interacting environmental and ecological variables that drive these blooms. On the other hand, data-driven models such as statistical regressions and machine learning approaches do not require prior knowledge of model structure, parameter values, or underlying ecological processes and have been applied more broadly by the HAB monitoring and modeling community. While data-driven methods have been successfully used to infer non-linear and potentially causal relationships between cyanobacterial growth, toxin production and environmental variables using only observed data, they lack a strong mechanistic basis and so are limited in their ability to predict cyanoHABs across systems or for conditions outside of the range of environmental variables observed in the local datasets from which these models were created. We contend that these apparently disparate modeling paradigms are actually complementary and aim to integrate them by (1) overlaying data-driven approaches to inform and refine an existing mechanistic model and (2) utilizing mechanistic models to validate models developed by data-driven methods. Because our proposed modeling framework is founded on mechanistic understanding, we aim to advance prediction of future states that are not constrained to current conditions. Resulting prediction tools would be of direct benefit to lake managers because, by more clearly elucidating causes underlying toxic cyanoHAB blooms, efforts to alter the physicochemical and biological characteristics of HAB-impacted water bodies via in-lake interventions and/or watershed management can be made with higher confidence that these efforts will influence cyanoHAB management targets. ________________________________________US EPA, 1985. Rates, Constants, and Kinetics Formulations in Surface WaterQuality Modelling. Report EPA/600/3e85/040, second ed. EnvironmentalResearch Laboratory, Office of Research and Development, U.S. EnvironmentalProtection Agency, Athens, GA, USA.