You are here

First-Person Science: Colleen Iversen on Belowground Ecology

In the First-Person Science series, scientists describe how they made significant discoveries over years of research. Colleen Iversen is a senior staff scientist at the Department of Energy’s Oak Ridge National Laboratory.

 

As a kid, I spent a lot of my family vacation pulled over on the side of the road looking at rock strata – my parents were both geologists. When I became a scientist, I decided to study the living things in the soil rather than the rocks themselves.

After I spent my childhood poking under rocks and in streams, I came to the University of Tennessee to do my PhD work. There, I got involved in a climate change experiment at the Department of Energy’s Oak Ridge National Laboratory. We looked at how trees respond when you give them extra carbon dioxide. Although we had hypothesized they would grow more mass above ground, we found that trees put all the extra growth into their roots. By investing in roots, the plants could get extra water and nutrients in the long term.

That’s when I got really fascinated with the world beneath our feet. I get to see things that nobody else sees and I get to understand them in a way other people haven’t thought about. Using special tubes and cameras that can see underground, I can see beautiful white fungal strands growing on a root and know that they’re working together to keep the plant alive.

So I think of myself as a cheerleader for the belowground world.

Peering Below the Surface

Because researching roots is hard, they don’t get as much attention as leaves. I tease people on our team who study photosynthesis that if they wanted to sample leaves in the same way that we sample roots, they should close their eyes and grasp around blindly to find a leaf.

Studying fine roots is even harder. Fine roots are less than 2 mm in diameter - that’s narrower than the cord that connects my phone with my earbuds. These little roots are tiny but mighty; they’re the ones that do all of the resource gathering for the plant.

As there’s no one best way to gather information on these fine roots, we use a lot of different techniques depending on the questions we are trying to answer.

People have been looking at roots forever because people have been growing food forever. The original version of a tool we use called a minirhizotron was first developed using a mirror and clear tube. These days, we still use clear tubes, but they instead have automated cameras. These cameras take pictures of what the roots are doing, when they’re born, how much they’re growing, and when they die. Some of them even have mini-robots that carry the cameras around the tubes. When a spider hitched a ride, we ended up with thousands of spider pictures. That was pretty funny.

But minirhizotrons only capture photos – I also want to measure the root’s chemistry. For that, we use traps called in-growth cores. Roots grow into a mesh cylinder filled with soil, allowing us to measure the root’s mass and chemistry. Taking soil samples helps us see how many roots there are at any given time and where they are growing.  

Once we have these roots, we have to separate them from the soil. So we spend hours in the laboratory wearing these really goofy-looking jeweler’s glasses. Some of the roots we work with are as fine as a human hair. Getting miles and miles of them out of the soil can take a while.

Partnering to Create Virtual Worlds

Picking tiny roots out of the soil seems far away from the complex Earth system computer models my co-workers develop. But we both depend deeply on each other. As a graduate student, I hadn’t really thought about models. I would publish my papers and throw them over the fence assuming modelers would totally pick my data up and use it. But that’s not how we make advances.

Now, I know more about models than I ever thought I would. In Oak Ridge’s Climate Change Science Institute, where modelers and field researchers work side-by-side, we joke that we regularly “take a modeler to lunch.” But the informal conversations have led to interesting and useful discussions. Although the models can be kind of opaque to someone who’s not elbow-deep in programming languages, having modelers paint pictures of the virtual worlds they create helps me understand them better.

With this perspective, I started thinking about how what we know about ecosystems underground is represented in these computer models. Because models can’t be too complicated – they’d overwhelm computer systems – they necessarily have to simplify the world. But we need to simplify it in the most representative way possible.

While most models include roots in some way, they don’t really capture what’s happening in all of the corners of the world. They suggest without explicitly representing a lot of the aspects of underground systems. But if you’re not being explicit about those relationships and their mechanisms, then you can’t be explicit in your predictions of how they might change in the future.

For example, most models spread out nitrogen fixation – an essential process by which some plants pull nitrogen from the air – across all plants. But that doesn’t give you the right spatial distribution. Instead, we need to have a specific shrub represented in the model that plays that role.

Instead of complaining mightily about it, we have to put our money where our mouths are and provide data that can improve the models. We’ve been saying, “We don’t think this representation is right, but how do we think it should be? Where are the data that a modeler could use to improve their virtual world?”

A Database Named FRED

So in 2017, we created FRED – the Fine Root Ecology Database. My interns, A. Shafer Powell and Holly Vander Stel, originally named it the Fantastic Root Ecology Database, but we didn’t think that would fly in a publication.

FRED compiles datasets from around the world to help models better represent information about roots. It currently has more than 105,000 observations from 1,200 data sources. They include more than 300 root characteristics or traits, including root order (how close or far away the root is from the main one), diameter, and branching patterns. By documenting these traits, we have a better understanding of roots’ chemistry, relationship to microbes, decomposition, and role in the plant. FRED also includes data on the climate, soil, and growing conditions. Having all of the data in a common framework allows field scientists and modelers alike to compare data sets.

Assembling the data revealed some major gaps in how we think about plant characteristics and our research.

As we organized the data, we took a hard look at the idea of how plant ecologists categorize plants into groups. To make models simple enough to run, scientists sort the nearly 400,000 plant species into 14 categories, called plant functional types. The categories are based on where plants fall on the evolutionary family tree, the biome they appear in, and some above-ground traits. This division assumes that plants with similar traits play similar ecological roles. But we started wondering, “Are those divisions good enough to represent what’s happening below ground?”

From the FRED data, we saw the answer was “No.” In fact, the ratio of carbon to nitrogen in fine roots – a major trait – varied more within an individual plant functional trait category than between them.

In fact, the more we looked, the more we realized traditional plant functional traits seemed to be a poor way to categorize plants in general. Even for above-ground traits, these groupings didn’t always make a lot of sense. Looking at tundra plants, we found that the traditional functional groups explained only 19 percent of variation in six common traits. So these groupings may not help us understand how plant communities in the tundra will react to environmental changes. Instead, we suggested using individual traits in the model that reflect different ecological strategies.

But this approach requires a lot more data than we currently have. At the time, 71 percent of the root traits had fewer than 100 observations. Geographic diversity was lacking too – 90 percent of all plant trait data was from temperate and tropical biomes.

There was very little information on the tundra at all. When we tried to see if above-ground characteristics matched up with belowground ones in the tundra, we didn’t have enough information to draw many conclusions. Out of more than 100 studies that reported data on roots in the tundra, most of them didn’t have information on root traits specific to the Arctic. They also lacked information on important characteristics like root turnover and lifespan, respiration, and timing of growth. As fine roots may play a big role in how much carbon dioxide and methane the tundra releases, understanding their role is essential to creating accurate Earth system models.

The Next Generation of Research

Gathering that data is where people like me come in. To help us inform virtual representations of the world, the DOE Office of Science has been supporting next-generation ecosystem experiments that make observations across geographic space and time.

The Next-Generation Arctic Experiment looks at tundra regions north of the Arctic treeline. Soils in permafrost – which never defrost – hold half of the carbon stored in soil worldwide. These areas have harsh conditions like very low temperatures, wet soil, and high winds that can lead to plant communities and relationships that are like nowhere else. One study estimated that 90 percent of the plant biomass in those areas was belowground rather than above!

But we’re not just gathering data on current ecosystems. We’re also trying to look into the future. For example, we want to see what happens if you both warm up and increase the carbon dioxide in an ecosystem, such as in the SPRUCE experiment in northern Minnesota.

To better inform the models, we collect measurements both on types of data that are currently in models and those that aren’t.

While the model includes some information about root depth, more data could make it more accurate. Depending on where you are on a hill slope, there are different soil depths and different species. Which plants will dip their toes all the way to the rock or the permafrost and which ones won’t?

One of our team’s favorite things to do is “break” the model. In the paper we’re currently writing, we measured five times as many roots below ground than the model predicts.

There’s also things that aren’t well-represented in the model at all. If I only made measurements that informed what’s currently in the model, I’d be out of a job pretty quickly. The models don’t show roots taking up nutrients – plants just get what’s left over after the microbes take them up. With more data, we can better show what the roots are actually doing. That data allows me to say to the modelers, “I think this process is important and here I have data to show how it should be represented.” We can’t just say, “You got it wrong, do it better.”

Unlike some fields, there’s no “aha” moment. Instead, there’s the gradual dawning realization of how all of these tens of thousands of data points fit together. I enjoy thinking about how to develop the big picture in a way that does justice to the nuance in the ecosystem.

 

The Office of Science is the single largest supporter of basic research in the physical sciences in the United States and is working to address some of the most pressing challenges of our time. For more information please visit https://www.energy.gov/science.