The Model Lake

One of the world’s most respected ecological thinkers sounds a warning for Lake Mendota

by

When Lake Mendota turned the color of a bad Gatorade experiment in June, you should have seen it through Steve Carpenter’s eyes.

Carpenter, who is retiring this month after 28 years at the UW Center for Limnology, talks about Lake Mendota with a subtly relaxed sense of time. He’s been studying the Madison lakes since he began his doctorate in 1974. His specialty is environmental change: understanding it, predicting it, manipulating it. And Lake Mendota has been his laboratory, his lens, his living model.

Just like you, he saw that unnerving blue, the dead and dying fish. He knew the smell would probably get worse. No swimming allowed.

But he could also see the lake in 1971, when we finally diverted human sewage from the lakes. He could see it in 1987, when an extra large batch of walleye were released to initiate a grand ecological experiment to clean the lakes. He also sees decades into the future, divergent scenarios ranging from ecological recovery to stinking death spiral.

Ecological change is often slow, but people tend to accelerate its pace. When European settlers first broke sod in Dane County, they quickly sullied Lake Mendota’s clear water with excess nutrients and the lake became eutrophic: overfed and overly green.

Status quo for the last 50 years has been a tolerable level of eutrophication. But now, in less than a decade, four major events have sent our signature ecosystem careening on a potentially disastrous course.

Record rains and floods came in 2008. In 2009, the spiny water flea invaded, joined in 2015 by the zebra mussel. Between these invasions, groundwater flow reversed: Instead of the lake being recharged by cool, filtered groundwater, the water you think twice about swimming in now recharges your municipal water supply.

Any one of these changes would mark a permanent turning point in the life of any lake. None improve water quality. And we let it happen. For decades now, Carpenter and his colleagues have been building increasingly powerful models of how land and water interact to create Lake Mendota. They have continually warned us that bad things were coming.

The naysayers of environmental protection distrust such models and say the science is inconclusive. But our Lake Mendota is one of the best understood ecosystems on the planet. How did we let this happen?

And more importantly, what’s next? “These lakes are changing tremendously right now,” says Carpenter. “This should really be a canary in Madison’s coal mine for the pace of environmental change.”

Carpenter watched the torrents of June 2008 thinking: “Here comes a hell of a lot of phosphorus. I was also thinking here’s just another click showing that the climatology models are right.” Of the 11 record summer rainfalls in the region, seven have occurred since 2000. “This isn’t a fluke,” he says. “This is a change in the pattern.”

This casual observation shows how deeply models inform modern ecological thinking. He’s confident the phosphorus is coming because of models he’s helped build and perfect. And the rain doesn’t surprise him because of models climatologists have built.

You’ve probably heard climate change deniers disparage these models. Models are complicated and abstract, and part of the denial strategy is to portray models as a black-box technology powered by voodoo instead of veracity.

But good science is smarter than this. Yes, models can be used to predict things. And the better those models get, the more valuable they are for both scientists and the general public.

But models are far more valuable as a tool for inquiry and exploration. “The main reason to use models is that they are a great way to ask what-if questions,” says Carpenter. “It’s a way to expand your thinking. Using models you can very cheaply do experiments that would be expensive, impossible or unethical in nature.”

“I’ve got one Lake Mendota. One year in Lake Mendota happens every year,” he says. “I’ve got all the Lake Mendotas I want in this little computer. And one year in one of those Lake Mendotas takes place in a second.”

These models help generate ideas, and then he checks his work in the natural world. “What makes it science is that we close the loop,” he says. “There is this continual dialectic between field observations and model expectations. The models are wrong most of the time and that’s great, because that means you learn something about why your understanding is wrong.”

Models show up early in the history of Lake Mendota’s malaise. Just two months ago marked the 50th anniversary of the first global symposium on eutrophication, sponsored by the National Academy of Sciences. After World War II, the proliferation of phosphorus as a fertilizer and as a detergent additive was choking waterways. Nearly 600 people from 11 nations met in Madison in 1967 to discuss eutrophication.

“Speakers agreed that the prevention of further damage to water resources is a matter of great urgency,” says the report from the proceedings. “Models of simple ecosystems suggest that a reduction in nutrient supply is the only management program that has the cooperation of the system.”

They were speaking generally, yet their analysis fit the Madison lakes 50 years ago, and today. “It’s like being on a treadmill,” says Carpenter. “We’ve just been running in place with phosphorus management. Now it looks like we’re slipping behind.”

You could say that the spiny water flea is one thing that the models got wrong. Originally from eastern Europe, it was first found in Lake Huron in 1984, and it’s been moving into inland lakes ever since. It likes lakes that are cold, deep and generally clear, so early models of invasions suggested it might skip Lake Mendota.

Midwest lake managers soon began sounding vigorous alarm bells because the spiny is a giant among zooplankton (tiny lake creatures) with a voracious appetite that can profoundly change a lake. It gobbles up the zooplankton that feed on algae, which leads to algal growth and murkier water. It also starves out the fish that eat zooplankton.

To understand Lake Mendota’s pain, you need to meet Daphnia pulicaria. This tiny grazing zooplankton helped Carpenter arrive at some of his more important ecological insights.

You probably recognize a food web from a grade school textbook or a poster at a nature center. Typically, there’s a pond with fish, zooplankton, some plants and insects, some sunlight, and a few microbes stuck in the mud at the bottom.

What matters to ecologists is the arrows, the relationships between the different pieces. They figure out the complex flow of energy or nutrients by fitting numbers to these arrows. Carpenter’s early research targeted how these relationships can transform an ecological system.

He also suspected that you could tweak the Lake Mendota food web to improve water quality. The idea was simple enough: More fish-eating fish (walleye, in this case) would eat more of the fish that eat daphnia. That would lead to more daphnia, who would eat more algae, improving water clarity.

In the early ‘80s, Carpenter and fellow UW limnologist Jim Kitchell used models to imagine just such an experiment in Lake Mendota. It would take three years to implement, and most of the walleye that state hatcheries could produce, costing about $1 million.

They took the model to a series of experimental lakes in northern Wisconsin. The fish in Peter Lake ate fish, while Tuesday Lake was filled with plankton eaters. For two weeks they fished both lakes intensively, working with skilled anglers and electrofishing rigs. They stored the fish in giant seine pens, then swapped the fish in the lakes. The proof of concept worked, and in 1986 and 1987 they refined the models, while some of Carpenter’s students also built a separate model of the biomanipulation from scratch.

The experiment worked. “Model fingerprints are all over that,” says Carpenter. In fact the biomanipulation begun in 1987 is the single most effective management action ever used in Lake Mendota. It provided an extra meter or more of clear water for more than 20 years. “Nothing else that has been done in the management of the lake has brought that much improvement in water clarity,” says Carpenter.

The real tragedy of the spiny invasion is that daphnia populations have crashed, crippling one of our most powerful natural allies for water clarity. “It had exactly the adverse effects that we feared it would have,” says Carpenter.

Models have come a long way since the simple ecological ones discussed at the 1967 eutrophication symposium. Technology has spurred the growth of more and better monitoring data, while computing technology has expanded our capacity for complex calculations. In addition to the many different models of the Mendota ecosystem, models of soil, crop and groundwater have grown increasingly sophisticated. These natural systems all interact, so connectivity between models is a critical and necessary next step.

In the 1980s, hydrogeologists at the U.S. Geological Survey began to create MODFLOW, now a standard for groundwater simulation. After the 2008 floods, researchers were surprised to detect viral contamination in Madison’s deep aquifer. We thought our groundwater recharge took centuries, a process that destroyed viruses. Eventually, they deduced that in addition to the slow aquifer recharge, large storm events were causing a rapid recharge through larger fissures.

While working this out, hydrogeologists also realized that pumping pressure had reached the point where groundwater could no longer recharge Mendota (or Monona) around the urban core. “It’s a tremendous message about our groundwater use and our connections to the groundwater,” says Carpenter. “The lake is no longer receiving cold water recharge.” That jeopardizes some aquatic plants and animals native to Lake Mendota that depend on those cold spots to survive.

This kind of cross-cutting insight on the lakes may become more common as UW researchers work to splice different models together to create a picture of the whole watershed. UW agronomy professor Chris Kucharik is working with Carpenter to build this supermodel, and readily admits that for a long time he stayed in his silo and ignored the groundwater connection. “I knew it was there, but there wasn’t a lot in the literature modeling a 3D connection between the groundwater and what’s happening in the soil/plant/atmosphere system,” says Kucharik. “We were both living in our separate worlds.”

His workhorse model is called Agro-IBIS and is tailored to the corn belt. It’s actually a series of interacting models that must capture how a crop like corn interacts with nutrients, water, climate and soil. Sometimes he operates at a massive scale, modeling the Mississippi catchment at 64 square kilometers per pixel. But in Dane County, he’s working in incredible detail, pixels just 220 meters square.

When you’re working at large scales, there is a certain amount of hand-waving, where little things get swept under the much bigger picture. “You can’t get away with that when you’re trying to model at this scale,” says Kucharik. It’s a digital landscape connecting Lake Mendota to every well head and cornfield, informed by incredibly precise knowledge about how phosphorus moves from field to lake, and how the lake responds. “We’re not quite able to do wholesale watershed simulations yet, but we’re working toward that.”

These are cutting-edge models, says Carpenter, precision tools with the potential to influence environmental practice far beyond Lake Mendota. You can imagine practically any future, as long as you can book computation time: Running a full scenario on the fastest computers on campus takes two days.

Steve Carpenter knows that people are uncomfortable with modeling. “And they should be,” he says. Kucharik likes to quote the British statistician George Box, who founded the statistics department at UW-Madison and who famously quipped, “All models are wrong but some are useful.” Kucharik argues that we probably benefit more from discussing the shortcomings and limitations of modeling than celebrating its successes.

So why do we model? Pull back from the extraordinary detail of Lake Mendota’s current models to consider a very simple prediction of zebra mussel invasion built in the mid-1990s by then UW zoology professor Dianna Padilla.

Zebra mussels appeared in the Great Lakes in 1988, but had already been spreading through Western Europe for about 200 years. With collaborators Charles Ramcharan (now at Laurentian University) and the late UW zoologist Stanley Dodson, Padilla (now at SUNY-Stony Brook) compared lakes in Europe that had zebra mussels to those that didn’t despite being connected to infected lakes. From that, she deduced what they needed to survive.

Comparing this to Wisconsin lakes, her lab added DNR survey information showing how frequently people moved boats from lake to lake. From those simple variables they ranked Wisconsin lakes by their likelihood of invasion. Mendota was a white hot target.

Because it took 20 years for zebra mussel occupation, we’re tempted to question the model.

“This is the hardest thing about models,” says Padilla. The model wasn’t wrong: it didn’t guarantee invasion and it didn’t predict a particular order of invasion.

In fact, we tend to both overestimate what they can do, and underestimate their basis in reality.

“People think that we invent models out of the air, rather than basing models on information that we know and the behavior of systems that we know,” she says. Carpenter’s important contribution is in grounding models in ecological reality, an endeavor made possible by the deep data set chronicling Lake Mendota. It’s invaluable insight at a time of accelerating environmental change. Because we know that normal does not mean constant. Just because the lake bloomed blue in June this year doesn’t mean it’s going to bloom at the same time next year, or that it’s never going to bloom again.

“This is where Steve’s work is really, really important,” emphasizes Padilla. “He has been at the forefront of using long term monitoring data to inform his models about what normal looks like, so you can detect changes from normal. When we know enough about a system and we have data to understand what typical variability looks like, we then have the tools to detect and predict what changes from normal look like and what their likely impacts are going to be.”

As long as we pay attention. Zebra mussels could have been kept out of Lake Mendota if people had taken some simple preventative measures. “This is a completely preventable disaster,” says Carpenter. “And it wasn’t prevented.”

Now zebra mussels are exploding exponentially, an underwater runaway train likely to shock you when the water clears in October. We’re beginning to see the secondary ecological effects: glop washing up on the shoreline, and possibly the toxic algae bloom. “I think we’re just beginning to see the impact of zebra mussels,” he says.

When Carpenter riffs on the lakes, time really does get bendy. He knows the spiny water flea will be gone because they can’t handle warmer water. Because climate change is real, somewhere in his head it’s already happened.

This is one facet of the lens that he calls “the long now.” What an ecosystem looks like depends on its past, just as present conditions will affect its future.

“That’s a view I’ve come to,” he says. “When I was a younger scientist, I was paying more attention to fast processes.” Partly that was about academic pressures, but it was also a young person’s perspective. “Over time you begin to realize that all of this fast stuff you see is very cool, but it’s just really the ripples on top of a big, surging, slow-moving thing,” he says. And that big, surging, slow-moving thing is where the fundamental change happens. “Now I’m 65 years old. I was born in 1952, and it’s sort of natural to flip your time perspective. If I can look backward 65 [years], why not forward 65? It’s only imagination.”

The upshot of all the complexity he balances is that Carpenter has two very simple prescriptions to save our lakes.

One is to clamp down on invasive species. The spiny water flea and the zebra mussel are just mistakes, and we need to eliminate mistakes. We need more public education, more boat ramp cleaning stations. We could also protect a headwater lake — Wingra would make the most sense — by eliminating trailered boats. “The next invasive species will come. It will not be good. I guarantee it will not be good, whatever it is.”

The underlying problem is that, as long as there is a lot of phosphorus, the lake will grow something. Invaders tilt the field toward noxious things.

Dane County has been leading the fight to prevent phosphorus pollution for 40 years, and yet climate change and rapid growth are overwhelming any progress it might have made. “This is a really innovative, hard working county on phosphorus control,” Carpenter acknowledges. “Yet we’re slipping behind. We’ve got to fix that.”

In today’s anti-regulatory climate, Carpenter suggests data transparency. Some important soil phosphorus information is not released to the public, data that would reveal both good stewards and bad actors. Breaking that privacy seal would improve the models, which are only as good as the information that goes into them.

Transparency is also an incentive. “It’s just the psychology of people wanting to be good neighbors,” he says. Publish the data, maybe give an award for the most improved soil every year, and perhaps allow some public sanctioning or shaming of those causing the most problems. “There is a lot evidence that open, transparent communication can lead to adoption of better practices,” he says. The rise of recycling and the decline of smoking both prove this. “Norms can change.”

Share this trivia on Facebook: 48 percent of the phosphorus flowing to Lake Mendota is unadulterated cow poop. “If manure at that flow rate was coming out of a pipe, into the lake, the public would be absolutely outraged,” Carpenter says.

In the 1950s, the whole Yahara community began to realize that dumping even treated human sewage into the lakes was uncivilized. By 1971, the diversion was done.

What happened was a decision to value these lakes, to treat them as a community asset and not as a cesspool. “I think we need a similar moral conversion here,” Carpenter concludes. “A decent community would not be taking the waste from however many cows and essentially dumping it into the lake.”

Can our model community rise to the protection of our model lake?

The models don’t answer that question, but Carpenter still has hope: “Maybe someday that moral conversion will take place.”