The ABCs of Marine Ecosystem Modeling

Any student of oceanography can tell you that the ocean is a complicated place. A model of the marine ecosystem can help us think about it, ask questions about it, and learn about it. A very simple model of marine ecosystem might include just a few main components, as shown below.

This example represents the ecosystem as simplified components with simplified functions:

• Phytoplankton absorbs sunlight and nutrients
• Phytoplankton are eaten by zooplankton
• When phytoplankton and zooplankton die, they are consumed by decomposers
• Decomposers release and recycle the stored nutrients
• Energy moving through the system is shown by black arrows

This model helps us answer simple questions such as: What happens if more nutrients are added to the water? The logic shown in this model suggests that additional nutrients will result in higher numbers of phytoplankton. It goes on to suggest that zooplankton will have more phytoplankton to eat and zooplankton population will increase. As the larger populations live and die, they produce larger amounts of detritus, which is decomposed thereby releasing larger amounts of nutrients. It may seem likely that larger amounts of nutrients will cause an increase in the phytoplankton population. However the zooplankton population is also higher now and they are eating a larger number of phytoplankton. This model can’t tell us the magnitude of these various processes. It doesn’t include enough detail to answer the question in depth.

To answer this question in depth, we need numerical modeling. The numerical model includes an equation for each arrow in the model. The equation incorporates more data and knowledge into the model so that the results more accurately represent reality. As a hypothetical example, let’s say that scientists know that adding X amount of nutrient A to the ocean will cause phytoplankton to increase by Y amount. This relationship is converted to an equation and is built into the model. The model also needs equations for other relationships and processes such as:

• The rate at which zooplankton consume phytoplankton
• The rates at which these organisms reproduce, die, and decompose.

To run the model, we enter data into the equations contained in the ecosystem model and compute the results. The input data consists of measured or estimated values for amount of nutrients, growth rate of phytoplankton, predation rate by zooplankton, and so on. This numerical model can provide a more detailed and accurate answer to the original question: What happens if more nutrients are added to the water?

For more complicated questions and even more detailed answers, we extend the model with more components and connections. The improved model (see image below) adds the following processes and components:

• Four types of phytoplankton
• Three groups of zooplankton, grouped by functional type (Heterotrophs, Micro-, Meso-)
• Bacteria component
• Cycling of carbon: Dissolved, Particulates, DIC (dissolved inorganic carbon)
• Cycling of other important nutrients: Phosphate, Nitrate, Silicon (Silicon is important because diatoms use it to build their skeletons)
• Cycling of gases between the ocean and atmosphere: Oxygen, Carbon dioxide, DMS (a biogenic sulfur compound, dimethylsulfide, produced mostly by phytoplankton)
• More components representing the benthic community

Image credit: Susan Kay, Plymouth Marine Laboratory

Remember that each arrow needs to have an equation to represent the relationship in the ecosystem model.

So far the model includes biological and chemical processes. Next we add physical processes of the ocean such as the movement of plankton and nutrients in the water column, and processes that control water temperature and salinity. We accomplish this by expanding our basic model to become a stack of basic models, where each model in the stack represents the status and properties of one cube-like unit of the water column. At the bottom of the stack, a unit represents the actions and properties of the benthos. Now we can track the movement of water and organisms in one-dimension, that is, up and down in the water column. For each cell in the stack, the model runs all the previous equations plus additional equations for interactions with the cell above and cell below.

It’s not enough to model up-and-down movement in the water column. We need to account for the effects of ocean currents, tides, and winds. We accomplish this by building out a three dimensional matrix of cells, and a corresponding matrix of calculations.

At this point, we have a biological and chemical model that is integrated into a three-dimensional physical model. Now we can apply this model at various scales. We can ask: How will warmer seawater influence Cook Inlet in South Central Alaska? How will it influence the entire continental shelf of the Gulf of Alaska? Or the entire North Pacific including the sea floor?

Further we can apply these modelling techniques to calculate a time sequence of snapshots, to show changes over time. We might want to know about seasonal changes in plankton. In the winter, when there are fewer hours of sunlight, how does that affect phytoplankton and zooplankton?

Or we might want to know how increasing levels of carbon dioxide in the atmosphere will affect the acidity of the ocean, including predictions until 2100. When we add the time element, the calculations increase by orders of magnitude. The model must run a massive number of calculations for each cell in the matrix, and then run the massive number of calculations repeatedly, once for each increment of time needed to create the timeline of the forecast.

Example of a long term forecast produced by a large scale ecosystem model

No problem. The super computers can do that and provide a forecast for ocean acidity in 2100. But is it correct? Why should we believe this forecast? How can we validate the model? The model is not going to be exactly right, but how do we know it’s close enough?

Answering these questions is a crucial part of ecosystem modeling. This usually involves checking the accuracy of the model’s predictions over time. For an ecosystem model created by researchers in 2000, we check how accurately the model predicted conditions over the next twenty years. We can compare the prediction for today with the reality for today, as evidenced in data from satellites, buoys, and other oceanographic research methods. The data that was forecast is compared to the data that actually happened, and then we apply advanced statistical methods to determine the confidence we can have in the forecast.

The validity checks of the model provide feedback for improvements. In addition, scientists are always bringing more innovations and insights that can help extend and improve the marine ecosystem model.

Concept and inspiration for this article came from Susan Kay, Numerical Modeler, Plymouth Marine Laboratory, Plymouth England