Skip to main content

An Ensemble Modeling Approach to Assess Lake Erie’s Response to Changes in Nutrient Loads

An Ensemble Modeling Approach to Assess Lake Erie’s Response to Changes in Nutrient Loads

An Ensemble Modeling Approach to Assess Lake Erie’s Response to Changes in Nutrient Loads

Lake Erie

Lake Erie. Credit: skypics.com

Program: Managing Nutrients in the Western Lake Erie Basin
Program details » | All Managing Nutrients in the Western Lake Erie Basin projects »

The Erb Family Foundation and the Great Lakes Fishery Commission, with in-kind efforts from Environment and Climate Change Canada and the National Oceanic and Atmospheric Administration, supported experts to use the most recent data to update models of Lake Erie harmful algal blooms (HABs) and hypoxia. These experts will combine the models into an ensemble and use it to determine how the system responds to changes in nutrient loads, weather, and climate.

 

Historical Background

The Canada-U.S. Great Lakes Water Quality Agreement (GLWQA) was originally signed in 1972 and updated several times in the decades since, including most recently in 2012. The GLWQA sets water quality goals and objectives for the Great Lakes. 

The related nutrient loading targets are meant to address the impacts of urban development, industrial growth, and agricultural land use practices that result in altered habitat and nutrients entering the lakes. In 2015, the two countries agreed to phosphorus loading targets for Lake Erie. 

To achieve HABs no greater than that observed in 2004 or 2012, 90% of the time, it recommended a total phosphorus (TP) spring (May-Jul) load of 860 MT and a dissolved reactive phosphorus (DRP) load of 186 MT from the Maumee River. For central basin hypoxia, the target total phosphorus load is 6000 MTA, designed to raise the average Aug - Sep hypolimnetic dissolved oxygen concentration to 2.0 mg/L or higher. 

A Decade Later

These phosphorus load reduction goals were developed to meet thresholds guided by a collection of models developed and evaluated with nutrient loading and environmental response data collected through 2014. There have been over 10 years of new data, a suite of new and updated HAB and hypoxia models, increased understanding of the impacts of weather and climate, and new information especially on hypoxia impact on fisheries. It is time to update the models and to participate in a new ensemble that explores any changes in the response. 

Ensemble Model Approach

Using the most recent data and building on the models, experts will update them to develop and compare environmental response curves for Lake Erie harmful algal blooms and hypoxia. They will compare their results and combine the models into a single ensemble model, following, for example, Fletcher et al. (2018), Scavia et al. (2017), and Walker et al. (2003). Using the ensemble model, the team will develop updated phosphorus-load response curves for how the system will respond to nutrient loads, weather, and climate conditions.

There are three 3D mechanistic models of the entire lake (LimnoTech, Environment and Climate Change Canada, and University of Waterloo) and one model that treats the whole lake in a mass balance (University of Waterloo). These models simulate western basin HABs and dissolved oxygen leading to central basin hypoxia. There is a statistical model that predicts central basin hypoxia as a function of loads and air temperature (University of Michigan/North Carolina State University). There are 3 statistical models that predict western basin HABs as a function of Maumee River loads (NOAA, Stanford University, University of Michigan). The University of Michigan model also explores the potential impact of temperature. There are also three models that explore the effects of loads, HABs, and hypoxia on fish and fisheries (Ohio State University, Eureka Aquatic Research). 

All of these models and their derivative publications are listed below.

After calibrating and validation, the models will be used to develop response curves for today’s climate and future climates. The ensemble will create a combined curve for each response variable. 

Fletcher, D. (2018). Model Averaging. Springer.

Scavia, D., Bertani, I., Obenour, D. R., Turner, R. E., Forrest, D. R., & Katin, A. (2017). Ensemble modeling informs hypoxia management in the northern Gulf of Mexico. Proceedings of the National Academy of Sciences, 114(33), 8823-8828.

Walker, W.E., Harremoes, P., Rotmans, J., van der Sluijs, J.P., van Asselt, M.B.A., Janssen, P., and Krayer von Krauss, M.P. (2003). Defining uncertainty: A conceptual basis for uncertainty management in model-based decision support. Integrated Assess 4:5–17.

Modeling Team

  • Don Scavia - University of Michigan
  • Stuart Ludsin - The Ohio State University
  • Daniel Obenour - North Carolina State University
  • Anna Michalak - Carnegie Institution for Science/Stanford University
  • Serghei Bocaniov - University of Waterloo
  • Richard Stumpf - National Oceanic and Atmospheric Administration
  • Reza Valipour - Environment and Climate Change Canada
  • Todd Redder - LimnoTech
  • Hongyan Zhang - Eureka Aquatic Research, LLC.

Policy-oriented Stakeholders

The University of Michigan Water Center will convene this group of stakeholders to ensure the modeling team addresses questions that will result in useful information relevant to Great Lakes policy decision-making. They will provide advice on framing results in meaningful contexts and formats for sharing with various groups in the public and private sectors as the project is completed.

These agencies and organizations are represented as stakeholders:

  • National Oceanic and Atmospheric Administration
  • Environmental Protection Agency
  • Environment and Climate Change Canada
  • Ontario Ministry of the Environment, Conservation and Parks
  • Michigan Department of Agriculture and Rural Development
  • Michigan Department of Natural Resources
  • Great Lakes Fishery Commission
  • Ohio Lake Erie Commission
  • Ohio Department of Natural Resources
  • National Wildlife Federation
  • Alliance for the Great Lakes
  • The Nature Conservancy

Contact

Alison Bressler - [email protected] - (864) 243-7914

Details About Models

3D HAB/Hypoxia Models:

This is a 3D linked hydrodynamic, wind-wave, and water quality model. It uses a hydrodynamic model to simulate water transport, mixing, and thermal regimes; a wave model for wind-driven waves and sediment resuspension; and an advanced aquatic ecosystem model to simulate lower food web dynamics, water quality, sediment diagenesis, and sediment transport. The model includes nutrient loads from more than 70 tributary sources, including direct drainage inputs, and more than 20 point sources. Meteorological inputs include air temperature, solar radiation, cloud cover and wind forcings at the air-water interface. The model was calibrated for 2014-2016, and its performance confirmed for 2011-2020. Calibration was focused on nutrients, suspended solids, dissolved oxygen, and chlorophyll-a at several long-term monitoring locations in the lake.

Publications: 

This is a coupled hydrodynamic (ELCOM) and biogeochemical (CAEDYM) model. The hydrodynamic model predicts the velocity, transport, mixing, temperature, and salinity distribution subjected to inflows, outflows, wind stress, surface heating or cooling. CAEDYM simulates inorganic particles, dissolved oxygen, organic and inorganic nutrients, phytoplankton, macroalgae and macrophytes, zooplankton, fish, mussels and clams, bacteria, and metals. The model was calibrated for 2002 and then validated for 2005, 2008 and 2014. It has been applied to Lake Erie for estimates of seasonal and spatial dynamics of water quality and phytoplankton, impacts of mussel grazing on phytoplankton biomass and its seasonal dynamics, effects of external nutrient loads and atmospheric forcings on seasonal hypoxia.

Links to Typical Publications:

This coupled hydrodynamic-ecosystem model has been applied successfully to several small and large lakes and shown to resolve the predominant nearshore and offshore physical processes such coastal upwelling events and internal waves. It has also been shown to be capable of simulating water quality conditions in Lake Erie. The model includes nutrient cycling, exchange of dissolved oxygen to and from the atmosphere, decomposition of organic material, abiotic and biotic oxidation of reduced species, and photosynthetic oxygen production and respiratory oxygen by phytoplankton community.

This model is used to derive total phosphorus load–response curves for annual, spring, and summer conditions. The relationships predict Chlorophyll-a, water transparency, primary production, fish production, and standing biomass. These relationships and matrices provide a simple but robust framework to gauge the potential long-term changes in response to total phosphorus reduction interventions.

Statistical Models:

This model predicts maximum summer harmful algal bloom (HAB) size as a function of bioavailable phosphorus loads from the Maumee River and added a lake temperature term. It includes a linear relationship with the cumulative load and a change-point-linear relationship with the spring load of the prediction year. This current model explains 84% of interannual HAB variability. This model continues to be used as part of the NOAA ensemble annual HAB forecasts.

  • Scavia, D., Wang, Y. C., & Obenour, D. R. (2023). Advancing freshwater ecological forecasts: Harmful algal blooms in Lake Erie. Science of The Total Environment, 856, 158959.
  • Link to information about the model (Don Scavia's Website)

This model uses April-July and 9-year cumulative dissolved reactive phosphorus loading from the Maumee River to predict the seasonal maximum bloom area. This simple two-term model explains 75% of the observed variability over 1985-2015 and has been used in seasonal forecasting since that time and continues to perform well.

This model predicts, with uncertainty, the summer seasonal average hypoxic extent as a function of March-April average air temperature and the cumulative total phosphorus load from the Detroit, Maumee, Raisin, Sandusky, and Cuyahoga rivers converted to the central basin load. This model augments similar models (Del Giudice et al. 2018, Scavia et al. 2024) by adding the Detroit River loads and extending the calibration period to 2024. In addition to predicting hypoxic extent as a function of phosphorus load, it considers the influence of climate change through the spring air temperature term.

This model determines the maximum harmful algal bloom (HAB) biomass associated with the March - July phosphorus load from the Maumee River. It was initially developed in 2012, and then refined to consider seasonal temperature as a factor. The model uses measurements of HAB biomass and bioavailable phosphorus loads from the Maumee River as a key input.

Fish and Fisheries Models:

This model is capable of identifying potential hypoxic area and nutrient load thresholds above which they negatively impact commercial harvests. The model has suggested that Lake Whitefish were most negatively affected by increased total phosphorus loads and hypoxic extent and that Walleye had slightly higher thresholds. Yellow Perch showed the opposite effects, with little negative impact on extensive hypoxic areas.

This is a food web model for the central basin of Lake Erie, including dynamic simulations of detritus, bacteria, algae, zooplankton, benthic invertebrates, and fish. The model was calibrated to the biomass dynamics of 23 model groups spanning the years 1996 to 2020. The model uses phosphorus levels and hypoxia conditions as forcing factors and steering the dynamics of the food web over time. By changing the phosphorus and hypoxia conditions as forcings, the model generates the responses of fish biomass to varying nutrient and hypoxia.