In November 2018, The New York Times Magazine published an article called Insect Apocalypse. It was largely based on studies about insect abundance in Central Europe. The article got a lot of attention, including that of Dr. Nico Franz, a professor at Arizona State University’s School of Life Sciences. “I cannot have been the only entomologist — insect person — who was wondering, ‘wow, do we even have data for that in the U.S.?’” he said. There really hasn’t been in the past, but now there’s a huge project that’s trying to gather that sort of information for the future.
The National Ecological Observatory Network (NEON) is a group of sites across the U.S. where ecologists and tools are collecting scads of data to try and get a big-picture view of climate change, biodiversity, freshwater resources, and other complex ecological issues. It’s only been fully operational since last year — and then the pandemic happened. But that’s far from the first challenge the ambitious project will face, and it certainly won’t be its last. NEON is supposed to go on gathering data for several more decades.
Over two decades in the making, 30 years into the future
NEON’s history goes back to the late 1990s and early 2000s, with ecologists and biologists pushing for the creation of a cross-continental network, something akin to an earthquake sensor network, but for ecology instead of earthquakes. They approached the National Science Foundation (NSF) with the idea and the cost — almost $400 million to build, plus millions of dollars a year to run, for three decades. “It’s a really huge investment from the National Science Foundation, one that really hasn’t been seen before on this scale within ecology,” said Dr. Chelsea Nagy, a terrestrial ecologist at the University of Colorado Boulder.
The final site was up and running last year, and the interim between the first idea and now was beset with all sorts of headaches. There were funding issues with Congress, it went over budget (it ended up costing $460 million), and project oversight changed hands. NSF then chose Battelle, which runs several national laboratories, to oversee NEON. Just last year, the organization fired two high-level staff members, causing others to resign. But then, a few months later, in May 2019, NEON became fully operational.
However, it wasn’t like the lights — and sensors — came on all at once. Many of NEON’s 81 sites have been gathering data for years already. There are 81 sites in 20 domains, including tundra and taiga in Alaska, the southern plains covering a good chunk of Texas, and the great lakes, with a couple of sites in Wisconsin. These ecoclimatic zones are meant to represent the diverse terrains, ecosystems, and vegetation of the U.S.
At each site, an array of technology, as well as field staff, are supposed to gather observations about the plants, animals, weather, water, soil, and more. With standardized tools and measurements, the data should be easier to compare.
“While Neon is not an experiment — we’re not manipulating forests or systems by making measurements over the long term — you start to see ecosystems as a complex system.”
It may sound simple enough, but ecology doesn’t always work that way. “One of the things about ecology, historically, is that we’ve tended to ask a particular question in a particular location using a particular set of methods, and then no one ever does exactly that same thing again,” said Dr. Ethan White, a professor in the Department of Wildlife Ecology and Conservation at the University of Florida. “And so when the conclusions that one set of papers reached differ from another set, we’re not really sure exactly why those conclusions are different.”
“I’m doing a meta-analysis right now and collecting individual studies,” said University of Colorado’s Nagy. “And so for that, it’s really hard to analyze the data that’s collected by many different people in many different ways, because sometimes they’re measuring different things, and sometimes they’re measuring the same thing, but they use a different method.” With NEON, she said, everything collected across the sites is done using the same methodology, making it easier to compare.
The data is also spatially integrated — meaning, it’s a bunch of measurements collected from the same site. “What it affords a researcher, for example, even at that one site, is a way to just integrate all these data,” said Dr. Paula Mabee, NEON’s chief scientist and observatory director. Despite its more than 8,000 sensors, NEON isn’t designed to monitor everything.
“A lot of the NEON sites are in places where invasive species are actively removed. So they’re certainly not going to help us with that,” said Dr. James Clark, an ecologist at Duke University. Nagy is interested in how wildfires impact invasive species, but as NEON’s sites haven’t burned, she can’t include them in her research.
Regularly collecting data from Hawaii to Puerto Rico is meant to answer six “grand challenges in environmental sciences,” such as how humans are impacting biodiversity and how to prevent infectious diseases for a variety of species. Cameras, traps, and thousands of sensors are set up at each site. Researchers can look at biogeochemistry, soil microbes, remote sensing information, and aquatic data to get a sense of how they’re interacting, as well as how changes are manifesting in mammals, minerals, and microbes.
“While Neon is not an experiment — we’re not manipulating forests or systems by making measurements over the long term — you start to see ecosystems as a complex system,” said Dr. Ankur Desai, a professor at the University of Wisconsin-Madison. “All of these different parts — the water cycle, the carbon cycle, the species that are interacting — all change. And that’s the hope, that we get enough information in enough different places to make sense of the world around us.”
The great mouse detection
If you find yourself confronted with a small mouse with brown fur; a white belly, and big, dark eyes, it probably won’t be immediately apparent if you’re looking at a white-footed deer mouse or a woodland deer mouse. The trick is the ear length. Both species are found in the upper Midwest and are crucial to the research of Dr. Michael J. Cramer, assistant director at the University of Notre Dame Environmental Research Center in Wisconsin.
“They serve as a model system to help understand how natural populations react to environmental change, which is why I study them,” he said. Owls, hawks, weasels, and foxes all prey on the mice, which are eating plants and seeds themselves. “These mice are right in the middle of the food web,” said Cramer. Their presence impacts not only the animals that eat them, but the tree species that depend on them to disperse their seeds. Then there’s the competition, like squirrels and chipmunks, which eat some of the same resources.
“If you have a healthy population of mice, that’s generally indicative of a well-operating system, in terms of the forest dynamics,” said Cramer. He catches the mice, tags them, releases them, and recaptures them to get a sense of how many are in the forests that stretch across the top of Wisconsin and into Michigan’s Upper Peninsula.
The area is fairly remote, which is probably quite nice for the mice but not necessarily that helpful for all the questions Cramer hopes to answer. “We’re surrounded on three sides by national forest, so we are not a human-dominated system,” he said. But NEON is also trapping and releasing these mice at another site, called Treehaven, about an hour-and-a-half drive away. “I can look at their data, based on the observations that I’ve made in a relatively pristine environment, and then compare that to NEON’s data that is collected in not-pristine environments, and try to draw some conclusions about how I think the mice are responding at the species level,” he said.
Even though these are forest mice — not the type you typically find in a house, said Cramer — it’s still important to keep track of them for human health. They’re reservoirs for Lyme disease. “They’re not spreading Lyme disease, but they’re maintaining it in that population,” he said. And these mice are found in many states in the U.S. If Cramer wanted to do some comparisons between the Yooper mice and those in other regions, he’d need a huge grant and an army of grad students, he said.
Instead, “I can utilize data from other sites and other habitat types and other biomes and also try to make comparisons about what I think the mice are doing using NEON’s data,” he said. “It allows me to extend my inferences to much larger scales.”
While Cramer’s approach is fairly low-tech — “I don’t have the money to buy that many little, tiny radio collars for each mouse,” he said — the NEON sites do have plenty of tools.
How green is this valley?
At regular intervals, NEON flies an aircraft over its many sites. Onboard the plane are cameras, lidar, and imaging spectrometers. The spectrometers provide hyperspectral images, which are high-resolution and capture light that the human eye can’t see. It’s a bit like putting a color filter on a camera, but for hundreds of colors. Ethan White tries to determine the species of trees based on these hyperspectral images.
Depending on their color, leaves will reflect and absorb different wavelengths of light. If you map out the reflectance against all the colors of light, you get a spectrum. Measuring that spectrum at each pixel of the image reveals different leaf traits, like the amount of nitrogen or phosphorus. Your typical camera will capture light in the visible — or red, green, and blue — spectrum. A hyperspectral image comes from a device that records 426 spectral bands. “It can allow us to see differences between plants, in particular, from above,” said University of Florida’s White. “That would be difficult to see if everything looks kind of green, but we could see big differences between two things that were green by using different information from that hyperspectral imaging.”
Being able to see detailed color differences between the leaves of a turkey oak and a sand live oak requires huge image files. “We’re working with terabytes and terabytes of imagery,” said White. “We’re doing large amounts of very intensive computing on the University of Florida’s high-performance computing center, which they call the HiPerGator, because we really like gators down here.”
Though White has been teaching an introductory computer science class for biologists for about a decade, he said at a wider level, his field still needs to catch up. “Engaging with this with data at the scale that’s now being produced by NEON is something that’s very unfamiliar to most ecologists,” he said. The kinds of questions White is looking into require that amount of data, though.
“One of the things that we’re really interested in is that the processes that govern ecological systems change depending on the scale at which you’re looking,” he said. What affects an individual tree — the amount of sunlight it’s getting based on the height of its neighbors, for example — become less important as you zoom farther and farther out, too look at an acre, forest, or ecosystem of trees. Larger patterns of rainfall and temperature fluctuations become more important when comparing trees in Florida versus those in New York. “We understand that these processes change with scale, but we haven’t really had datasets that allow us to think about that in a sort of fairly continuous way,” he said.
Lichen lending library
In an industrial area of Tempe, Arizona, there’s a low, off-white building. Despite its unassuming exterior, it actually holds thousands of biological samples, both from the Arizona State University natural history collection and NEON’s biorepository. Microbes, beetles, mosquitoes, and soil samples will all be stored in the building for the 30 years NEON will run. Those that need to be kept cold are stored in a cryogenic freezer, and ASU is adding a liquid nitrogen storage facility as well. ASU’s collection is also on display, with cases of butterflies, jars of vertebrates, and a bear skin hanging on the wall. Arizona State University’s Franz is the director of the collection.
“We are absorbing roughly 110,00 samples annually, per the statement of work that we currently have with NEON,” he said, “and about 70 percent of those samples need to be kept cool.” For several months, the facility was getting daily deliveries of samples, sometimes multiple times a day. Some of NEON’s sites have been collecting samples as far back as 2013. While properly storing these samples is an important component of what Franz’s team does, they also need to catalog and track them. “We’re not actually supposed to be, for every sample, the final stop ever,” said Franz. “To the contrary, right. We’re supposed to be sort of like a passthrough for these samples, so that additional research can be done.”
In order to make NEON’s main data portal useful to researchers, the biorepository staff have to enter a lot of information, based on the Darwin Core Standard. “This is relatively close to something like a lingua franca, like a broadly, globally utilized standard to transfer and annotate data, relative to natural history collection specimens,” said Franz. It’s meant to provide researchers with everything they need to know about a sample they didn’t collect themselves: Where, when, and how it was collected; its taxonomic name and how it was identified. “We can also record relationships among samples,” said Franz. “So, for example, if ‘A’ is a known parasite of ‘B,’ we have samples from ticks that were taken from mammals.”
Some of the requests Franz has gotten could help with a U.S. version of an insect apocalypse study. The University of Oklahoma is looking at changes in the volume of invertebrates across about 50 sites, he said. They’re using what’s called pitfall trap samples. Plastic cups are left in the ground for days or weeks, collecting spiders, beetles, earthworms, and other invertebrates. From these samples, researchers can extrapolate about the number of such animals at each site, as well as their body size. The researchers are already seeing differences across various regions. “There seem to be subregional trends to this insect apocalypse that are pretty interesting,” said Franz.
Pausing for the pandemic
Maybe you picture an ecologist as someone holding a clipboard and wearing rubber boots in the middle of a forest. Perhaps your image of a NEON site is of sensors humming along, no humans in sight. NEON has 181 data products — collections of measurements like barometric pressure or infrared temperature or CO2 concentrations — that it puts out; 108 of them require human observers. “We rely a huge amount on observational sampling,” said Zoe Gentes, a senior communications specialist at Battelle, which manages NEON. “But in addition to that, our automated systems need maintenance.”
“That is something that hit me really within the first few weeks of COVID,” said Mabee, NEON’s chief scientist. “I just had no idea how much maintenance sensors took to be refreshed and have their batteries replaced and stuff.” Gentes said many of the sensors have held up very well, except in cases of power outages or flooding. NEON’s Airborne Observation Platform is now constrained to certain parts of the U.S. that don’t require commercial flights, because NEON isn’t allowing its staff to fly, said Mabee. “This is not an essential job,” she said. “It’s an important one to us. It is our job. But it’s not essential work.”
While NEON has been looking at ways to make field ecology safe to perform during the pandemic, Mabee thinks its practitioners might take a second look at the project now. “Certainly in COVID times, I think people have sat down at their computers and not gotten out into the field too much,” she said. “And so it’s really turned a spotlight on these types of data.”
In the early days of NEON, there was a fear among some scientists that it was trying to fully replace field work. “I think they kind of misbranded NEON initially because they’re saying like, ‘oh, now you can do ecology from your desktop,’” said Cramer, who’s been studying mice in Wisconsin. “And there are a large portion of ecologists that were like, ‘Yeah that’s not why we got into ecology. We want to go outside.’”
Other criticisms of NEON were that it might lead to doing science “backwards,” creating data, then searching for a hypothesis. There were concerns that the NSF’s Long Term Ecological Research Network would lose funding in favor of NEON. “It doesn’t help that NEON has come along at a time when there’s been an awful lot of angst about declining funding for science,” said Clark, of Duke University.
It took so long for the project to fully get off the ground, that in the meantime a whole new generation of ecologists were coming into the discipline. At last year’s Ecological Society of America’s annual meeting, Dr. Kyla Dahlin tweeted: “Tired of hearing grumpy old men complain about @NEON_sci ? Come hear from actual NEON users…”
Terrestrial ecologist Nagy has been involved in some training to make sure no ecologist gets left behind. “I think that’s one of the first challenges, is just making sure that ecologists in general have the skills necessary to use data like this,” she said. “It’s a really cool resource because it’s freely, openly available to anyone, but you do have to have this knowledge about how to use the data in order to make it useful.”
An ecosystem of ecologists
“I think one of the most, if not the most, exciting things about NEON is the potential for fostering community and broader collaboration within ecology more generally,” said White. His group is working with researchers from the school’s computer science and forestry departments on his tree projects. He’s working on combining NEON data with some from the U.S. Forest Service’s Forest Inventory and Analysis program. Just like integrating two datasets, creating a team from different scientific backgrounds requires creating a common vocabulary.
“NEON has helped bring much-needed standardization to tools and measurements that are crucial for answering some of ecology’s big questions.”
Even outside of NEON, ecology has been expanding to fit bigger data and larger questions. White noticed it a decade ago when he started teaching computer science classes. “There are many, many labs now that don’t collect any field data themselves, but simply work on data that can be obtained from remote0sensing products or other network products,” said Clark. Desai echoes that. “We may have students working on field sites that they’ve never seen,” he said. “And that’s maybe a little different — uncomfortable — for some people in ecology.”
Desai has watched NEON’s development from the inside and outside. Previously, he was a member of its Science, Technology & Education Advisory Committee. His feelings about the project are mixed. He praises the scientists that work there but says its management has had its ups and downs. NEON has helped bring much-needed standardization to tools and measurements that are crucial for answering some of ecology’s big questions. Many ecologists think NEON’s data will help start to answer some of those questions.
“If you talk to National Science Foundation, they always say, ‘NEON is not the thing. Macrosystems biology is the thing,’” Desai said. “Neon is the tool to make that science happen. And if it turns out we need a different tool 10 years from now, then so be it.”
