A network of computer models is predicting the future of Puget Sound


A new $4.8 million dollar project led by the Puget Sound Institute links together a series of computer models to explore future scenarios across the watershed.

Some of Puget Sound’s biggest concerns hold the greatest uncertainties. 

Will we have clean water? Can the ecosystem sustain species like endangered salmon? How can the region continue to grow and still maintain healthy habitats for wildlife and people? What, in other words, is the future of Puget Sound?

No one can travel through time to answer these questions firsthand (science has its limits), but increasingly, scientists can study what might happen virtually. Advances in computer modeling are allowing researchers to examine all kinds of scenarios in more detail than ever before. 

This month, the Puget Sound Institute (PSI) announced a three-year, $4.8 million dollar project to study the dynamics of Puget Sound’s changing ecosystem. The Puget Sound Integrated Modeling Framework (PSIMF) combines a network of computer models to look at how different factors like urban growth and climate change will influence the health of Puget Sound. The project is funded in part by a $2.5 million dollar grant from the Paul G. Allen Family Foundation with additional support in-kind from EPA-ORD and NOAA Fisheries, and grants from the Puget Sound Partnership.

The project is the most comprehensive endeavor so far undertaken to simulate environmental conditions connecting Puget Sound’s land and water. The result is a virtual ecosystem of sorts, combining five different models that integrate data across the watershed.

“The Pacific Northwest is home to cutting-edge terrestrial, estuarine, and marine ecosystem models that each have their own use in supporting specific decisions and advancing understanding,” said Dr. Tessa Francis, lead ecosystem ecologist with PSI and Principal Investigator on the project. “So far, we haven’t been able to link these models together to provide a unified platform with insights into the interconnected nature of our ecosystems.”

Collaborators on the project include EPA’s Office of Research and Development (EPA-ORD), UW’s Salish Sea Modeling Center, NOAA’s Northwest Fisheries Science Center (NWFSC), Long Live the Kings (LLTK), the Commonwealth Scientific and Industrial Research Organization of Australia (CSIRO), and Pacific Northwest National Laboratory (PNNL). PSI will lead the effort, which held its first formal meetings last month. 

Models linked in the project include: 

  • The VELMA ecohydrological watershed model developed by EPA-ORD
  • The Salish Sea Model, a hydrodynamic circulation model from the Salish Sea Modeling Center and PNNL
  • The Atlantis ecosystem model,  a food web model developed by CSIRO and implemented by LLTK and NOAA’s NWFSC
  • A high-resolution land cover change model from PSI
  • A qualitative socio-ecological model developed by NOAA’s NWFSC

Read a press release about the project from the University of Washington Tacoma.

Scientists have known for some time that what happens on the land affects the health of the water. Toxic pollutants from stormwater are a classic example, as they move from impervious surfaces like city streets into the marine food web, affecting Puget Sound’s salmon and the people who eat them. 

Models like EPA’s VELMA (which stands for Visualizing Ecosystem Land Management Assessments) can track the movements of chemicals like PCBs and make predictions about where these chemicals might end up. That can help planners identify the best places to place rain gardens and other green infrastructure, for example, but it doesn’t always help scientists follow the chemicals after they enter Puget Sound. 

Now, within PSIMF, analyses from VELMA will be fed into the Salish Sea Model, an advanced computer simulator that shows how the water circulates through Puget Sound and the Salish Sea. The Atlantis ecosystem model of Puget Sound developed by Long Live the Kings and CSIRO in Australia completes the circle by analyzing how species might react. Additional models from PSI and NOAA Fisheries will analyze land cover change and some of the ways that ecological processes can affect human social and economic wellbeing.

“These are tools for systems thinking,” said the EPA’s Dr. Robert McKane who leads development of the VELMA model, and is one of the originators of the idea to create Puget Sound’s integrated modeling framework. “Thinking about systems and how they work is really the ultimate goal for ecology. The science of ecology is to understand how things fit together.”

While salmon recovery will play an important part in the modeling framework, the potential of these linkages extends far beyond salmon and toxic chemicals. The ability to evaluate so many factors and data points will help scientists make a range of potential predictions about the ecosystem. Those might include smaller analyses — concerns like the abundance of herring or zooplankton — or billion-dollar questions about the impacts of regional population growth on wastewater, providing a powerful tool for managers and planners. 

The project comes at a time when computer models are increasingly important to the biological and physical sciences.

“I think that like every other field on the planet the use of technology in the biological sciences is increasing,” said Francis. “Quantitative analysis is becoming more and more accessible and more common as a skill. And access to high-capacity supercomputing is increasing as well.  So, I think we see more modeling happening not because we need models more and more but because we can do more and more modeling owing to technological advances in the world in general.”

*Related: Six things that people should know about ecosystem modeling and virtual experiments*

Much of what PSIMF will do in its first year will be to figure out “how to get the models to talk with each other,” Francis said. “It’s developing that shared language at the points of intersection between the models. Much of the work of this project is developing that communication and that dialogue.”

To test this, early efforts will focus on the Snohomish watershed, “just to start with a smaller piece of geography,” Francis said. “And then once we know we can connect the models we can expand that to the rest of Puget Sound.” Once the models are linked, the team will then start asking the big questions. “We’re going to be asking questions about how human choices will affect the ecosystem,” she said. The answers may still be uncertain, but Puget Sound modeling has just taken a big step into the future.

Ask a Scientist: A Q&A with PSI’s Tessa Francis

It’s hard to overstate the importance of mathematical models to science. Models show how planets move and how diseases spread. They track the paths of hurricanes and the future of climate change. Models allow scientists to look at systems or scenarios that they could never view otherwise. Increasingly, mathematical models are also helping scientists understand Puget Sound. We spoke with PSI’s Tessa Francis about the new Puget Sound Integrated Modeling Framework and some of what we need to understand about the uses of computer models in ecosystem science. 

Q: Help us understand what you mean by models. What is a model?

A model is really a simplification of some part of the real world. And we build a model because when we have questions or hypotheses that we want to test, we can’t always replicate the actual world at full scale, because that would just be the world. So, sometimes, to test hypotheses, we build models, simplified descriptions of the real world. 

It’s helpful to think about an analogy of a 3D model, like a model airplane. A model airplane is not just a miniature version of an actual airplane, but it’s simplified. You’ve cut out a lot of important parts of an airplane to build the model, but it still replicates reality enough that you recognize it. You believe that the important parts and functions are still there. It looks like an airplane; it still has wings. You can see the doors, windows and wheels, and some model airplanes even fly. But you would never use a model airplane to make a prediction or to test a hypothesis about how an airplane will function with people on it and a full tank of fuel in an ice storm.

Q: But the models used in PSIMF are a little different than 3D fabrications, right? These are mathematical models. Can you describe what that means?

A mathematical model can be a simple equation that describes some sort of linear process. Like a line, y = mx+b. That’s a model; it is a mathematical representation of a line. Usually, however, when we talk about models we are talking about a system of many mathematical equations; very complicated models have hundreds and hundreds of equations in them. Those equations can be very complicated, for example if they are describing a complicated process that’s not linear. You can think of the models that we’re talking about here as being a set of many mathematical equations.

Q: Are these computer models?

The models we are talking about are complicated enough that you cannot solve all the equations by hand. You have to use a computer, and there are so many equations that it actually requires a lot of computing power to solve them all. 

Q: Where did the idea for the PSIMF project come from?

It’s an idea that Bob McKane [of EPA’s Office of Research and Development] has been working on for many years. He and his team have developed a watershed model called VELMA, but watersheds really sit in the middle of an integrated system. A watershed sits in the middle of a lot of human processes, like where people build houses and roads, and climate influences, such as how much rain is falling and where.  VELMA predicts where water goes and where the chemicals carried by that water go and how much of those chemicals go out from streams and rivers into the Puget Sound. But then if you want to know what happens after that you need a separate model of the estuary where the water circulates. And if you want to know about effects on the food web, you need a food web model. So, Bob saw VELMA as being a model that sits in the middle of the social ecological ecosystem. And he also saw that many of the questions that we have in our region about how to manage growth, and how to plan for climate change, require this type of integration. Answering those questions requires linking multiple models together. He couldn’t answer those questions just using his watershed model alone. So, he developed this concept of a coupled modeling framework for Puget Sound, and he’s been shopping that around for years. 

Q: Would you call what you are doing predicting the future?

It’s not always the future. Models are predictions or guesses about the way something works. There’s not always a time component. We could be using a model to ask questions about the present. We can build a model that asks where fish are likely to be found, based on shoreline characteristics. Model results might suggest that fish are more likely to aggregate near shorelines that have certain characteristics. And that’s not a future prediction, that’s a current prediction.

Q: Computers are incredibly powerful now. Will that ever diminish the need to get real-world data? 

You still need field observations. We still haven’t finished collecting the data used to drive the models. When you cannot conduct an experiment at the relevant scale, that’s when you can use a model to do the experiment for you. We cannot do a real experiment to simulate the climate we’re going to see in 30 years, not at the ecosystem scale. We can’t do an experiment and observe how the fish respond, for example. So, we have to do that experiment using math if we want to make predictions. But there’s still a lot we don’t understand about the way the world works, even about simple things. So there’s still a role for field experimentation. We still need that. We still need observations in the field because models can’t tell us everything. 

Interviewed by Jeff Rice

Visit the PSIMF web page

Read more about the project from University of Washington Tacoma News and Information.