# Demystifying modeling

Want to predict the population of a particular whale species 50 years into the future? There’s a model for that. Want to know exactly how much water is moving around one spot of the ocean at any given time? There’s a model for that too.

Modeling has a long history in science, and advancements in technology have significantly improved the capabilities in recent years. Yet, despite our fondness for some new technology – smartphnes, for instance – many people seem to greet scientific models with more skepticism than fascination.

To find out more about modeling and how it can help researchers, Oregon Sea Grant talked with some of the scientists we fund and collaborate with who specialize in modeling.

In its simplest form, a model is a mathematical way of estimating variables that can’t readily be measured in the field.

When laypeople express skepticism or mistrust about models, it may be that they’re nervous or uncertain about the arithmetic.

“Most people don’t think that they can do math,” said Selina Heppell, a Fisheries and Wildlife professor at Oregon State University who specializes in population models. “When in fact they can do math. They use math all of the time although they don’t necessarily realize that they’re doing it.”

Another way to think about a model is as a laboratory experiment where you hold one variable constant and see what happens to the others.

“The point of doing a lab experiment isn’t to know what’s going to happen in the real world, it’s to control factors that you can’t control in the real world so you can see the effect of a couple of variables,” explained Julie Alexander, a postdoctoral researcher studying aquatic invertebrates. “That’s the same goal of a model, to see the effect of variables that you can’t manipulate in the lab.”

MODELS FEEDING MODELS

If you were a scientist trying to study the presence of particular larvae in Yaquina Bay, you would need information on tides, currents and more. Many of these data can be found in come from existing models, and they are combined with field data to answer research questions.

Moreover, there is a tendency to add additional factors into your system (precipitation, for example) in an attempt to make the model more accurate. In fact, Heppell explains, this approach can make the models less reliable.

“Making a more complicated model adds more parameters which adds more uncertainty,” she said. “That uncertainty can be accounted for, but adding too many details that you don’t know much about can make the model hard to understand and not very useful.”

Each model has its own level of uncertainty based on the data that went into making it. That problem only expands as you combine multiple models with the uncertainty already present in your own data.

To account for this, scientists spend a lot of time analyzing model outputs to ensure the results are reasonable. Microbiology professor Jerri Bartholomew is the lead biologist in her lab studying pathogens, and she constantly checks that the data correlates with her prior knowledge of the species.

“I think transparency is very important. You have to be very honest about what you can say with your model,” she said, adding that her lab also calibrates its models annually against new field data to ensure accuracy.

PROJECTING THROUGH TIME

Technological advancements are improving our ability to reduce uncertainty and run multiple simulations in a short period of time. But new technology does little to help explain models to the general public or decision-makers.

A large portion of Heppell’s work is reviewing the models used to set fisheries harvest regulations and explaining the outputs to fishermen and coastal leaders. As a modeler, she puts fish life cycle information into equations and simulations to show how various species will be impacted by new policies. She uses Microsoft Excel to help managers see how the model was created and how the outputs change with new information.

“The reason I use Excel is because it’s a platform that everybody has,” she said. “I create modeling tools that I can then give to a manager and they can manipulate it and look at what if this changes and what if that changes.

As models become more widely used in science, it’s important for those who make them know where the data came from, and for those who use them to understand their limitations. Whether field data or computer-generated values are fueling the model, the strength of the source makes all the difference in the usefulness of the model.

YOU ARE A MODELER

Let’s look at a simple model. The link below will take you to an Excel worksheet with information on whale populations. Through this model you can estimate changes in whale abundance over 50 years in the face of changing survival or reproduction affected by stressors like pollution, ship traffic and climate change. By tweaking simple variables such as lifespan and number of offspring, you will be able to see first hand how we can get a sense of the impact our policies have on animals with lifespans as long as your own.

You can find the model here: Modeling Practice