by National Ecological Observatory Network
When humans can’t get to the field for ecological research, could the robots take over?
When humans can’t get to the field for ecological research, could the robots take over?
COVID-19 has limited the ability of ecologists to get to the field to collect samples and check instruments in many parts of the world. While travel and work restrictions may be lifted in some areas over the summer, the pandemic will almost certainly result in some data gaps for long-term ecological programs, including the NEON program.
Robots, drones, and automated instruments may be able to continue many kinds of ecological data collection when it is not safe for humans to travel to field locations. Some of these technologies are already being used to collect data and physical samples in remote or dangerous parts of the world. If COVID-19 continues to restrict access to field sites into late 2020 and 2021, additional automation may allow ecologists to continue to collect much of the data they need for long-term analysis of ecological trends. What is gained, and what is lost, when we turn to technology to gather data traditionally collected by humans?
What’s Missing When Humans Leave the Field?
At the NEON program, work and travel restrictions due to COVID-19 prevented technicians from getting to the field for scheduled sampling bouts in the spring of 2020. That means that the program will be missing a number of data products for this time period, including collection of soil, rainwater, and litterfall samples; insect, small mammal, and bird counts; and plant observation and clip harvests. The flight crews that run the NEON Airborne Observation Platforms (AOPs) for remote sensing have also been grounded since March. The AOPs collect lidar and hyperspectral data that enable analysis of topography, vegetation structure and composition, and canopy chemistry. The NEON program is currently exploring options to resume data collection, including remote sensing and field sampling, in limited capacity for the remainder of the 2020 sampling season. (See the updated observatory status for more information on COVID-related disruptions at the NEON program.)
On the other hand, there are many types of data that have continued to be collected with little or no interruption. These include data from automated instruments on the eddy flux towers (including meteorological data and eddy covariance data) and soil sensor arrays. While regular maintenance of in situ instruments has been suspended, most instruments are continuing to collect and transmit ecological data.
That means, even in the case that human sampling cannot be fully resumed in 2020, NEON will not have lost more than a small fraction of its full operations. The data products collected by the automated instruments allow analysis of many long-term ecological trends—such as changes in climate, carbon cycling, and other key ecological indicators—to continue without significant interruption. Automated instruments may also pick up interesting data during this time, such as changes in particulate levels as a result of reduced human activity during the pandemic.
Still, the automated data products cannot totally make up for the loss of physical samples and human observations. Dr. Mike SanClements, the Terrestrial Instrument System Lead for the NEON Program, explains, “Sensors collect data that show trends and patterns over time, but observational methods and experiments in the field are where we get an understanding of the mechanisms of ecology. It is the coupling of sensors and observational methods where the true power lies, so you can not only see the trends but start to understand the drivers behind those trends.”
When the Robots Take Over
So, can robots and other automated technologies take over to collect the missing data? In some cases, it may be possible to fill in data gaps with additional technology. For example, drones—which are easier to operate by an individual or a small, socially-distanced team—could be used to supplement remote sensing data that is traditionally collected by the AOPs. The NEON program has already done some work to correlate drone-collected data to AOP data so that data can be compared and analyzed together. Drones are not likely to fully replace the sophisticated (and heavy) instrument payloads carried by the AOPs for some time, but they could enable ecologists to collect data when AOPs cannot be flown or to expand remote sensing capabilities with cheaper, more nimble technology.
Other researchers are thinking much bigger when it comes to the use of robots and drones in ecology. Dr. Harris Lewin, a professor of evolution and ecology at University of California – Davis (UCDavis) and the chair of the Earth BioGenome Project, says that he expects robot and drone technologies to play a big role in sample collection for his program over the next decade. The Earth BioGenome Project aims to sequence the genomes for all 1.5 million known species of eukaryotes on earth—a category that includes all plants, animals, and fungi as well as unicellular protists—in ten years. To complete this ambitious project, they will need to collect and catalog genomes of species in some of the world’s more remote and dangerous locations, from deep-sea vents to the peaks of the Himalayas.
Dr. Lewin envisions automated robot or drone technologies that could be deployed to remote locations to collect soil and water samples, vegetation clips, or even small organisms. The robots would need to be able to locate the appropriate sample, collect it, and either fly it back to the lab for analysis or perform some kind of analysis (such as DNA sequencing) on site. Dr. Lewin says, “A lot of the raw technologies already exist, such as drones, sensors, robotic arms and portable DNA sequencers. Somebody just needs to put all of this together. I think we might be three to five years out from having the technology we need.” If he is right, the technology will be available just as the Earth BioGenome Project begins to run out of samples to sequence that can be easily and safely collected by humans. Some of the technology could come from work being done on robots built for deployment on Mars. The X-Prize currently has a $10 million purse for development of automated technologies to study rainforest biodiversity.
Robotic technologies could one day enable safer physical sample collection in many environments. Dr. Lewin says, “Right now, we are dressing people up in space suits with fancy equipment to collect guano from bat caves to study coronaviruses. Imagine how much safer that could be if we could do that collection using robotic devices.”
Still, it could be a long time before robots are ready to perform the kind of sophisticated decision making that humans perform in the field when collecting samples. Dr. SanClements explains, “There is both an art and a science involved in many types of sample collection. For example, think about collecting soil samples. You have to be able to recognize where soil horizons begin and end as well as physical properties like texture and roots, rocks, etc. It’s not just about sticking a shovel in the ground and digging up some soil. These kinds of problems, which require judgment as well as sensing, are still very difficult for robots to solve.”
Finding a Perfect Partnership of Technology and People
For the foreseeable future, humans and technologies will both be needed in the field. Automated sensors and remote sensing technologies have already vastly expanded our perception, both spatially and temporally. Dr. SanClements believes that in the near future, adding artificial intelligence (AI) to these sensors could make them even more powerful. He says, “Imagine automated technologies that can respond to their environment, such as a water sampler that automatically increases sampling rate in real-time when a sensor detects a probable flood event, or a camera that sends an alert and starts recording when it detects a plume of smoke for early wildfire detection. Real-time or near-real-time automated sensing and data collection could enable early detection of anomalous events and improve iterative forecasting models.”
Ecologists are also starting to use robot and drone technologies, such as the DeLeaves canopy sampling drone, to speed up physical sampling or collect samples from inaccessible locations. Currently, most of these technologies still require a nearby human operator to guide sampling efforts.
Even if we do get to the point where robots and drones can recognize and collect certain samples independently in the field, that doesn’t mean there will be less work for humans to do. Automated sample collection could vastly expand the number and type of samples collected as well as the locations they are collected from. All that new data will create many more opportunities to ask and explore ecological questions—work that will always involve humans.
Dr. SanClements says, “The coupling of sensors and observational data and physical samples is what makes the NEON program so powerful. We are always going to need hands-on samples, and we are always going to need humans to ask the right questions, conduct experiments, and interpret results. When we bring the technology and people together, that is where the truly interesting advances in ecology are being made.”