When it comes to counting trees, are sensors in the sky as accurate as people on the ground?
Researchers with Spatial Informatics Group (SIG) will help address that question, thanks to a major grant from NASA’s Research Opportunities in Space and Earth Science (ROSES) program.
Recent years have witnessed big advances in the field of remote sensing, opening new opportunities for the use of this technology in carbon accounting. Satellite optical imagery, radar, and lidar can be used to estimate the amount of carbon in a forest in ways that are less costly than traditional field inventories.
Carbon markets, however, have been slow to embrace these new tools, largely out of concern for their accuracy.
Carbon markets depend on trust: buyers must be confident that carbon credits represent actual carbon stored in trees. That’s why carbon registries—which are responsible for creating accounting protocols and issuing credits—require the clear quantification of uncertainty in estimates of forest biomass.
The NASA ROSES grant will fund research into methods for quantifying uncertainty in estimates of forest carbon using remote sensing—also known as Earth observation platforms—so that this technology can be used with confidence in the carbon offset market.
Robert Kennedy of Oregon State University will lead the project, working alongside co-investigators from SIG, the US Forest Service, the University of Minnesota, and the University of Washington.
“The grant is really about making remote sensing more accessible,” says Tim Kramer, Director of Operations at SIG Carbon. “The work will help the registries integrate this technology into their protocols by providing an independent and transparent calculation of uncertainty for biomass measures at any spatial scale.”
Forest Inventories and Carbon Markets
Registries such as Climate Action Reserve and American Carbon Registry issue carbon credits to landowners who commit to following sustainable forestry practices, thereby sequestering more carbon than would be expected under standard forest management.
The credits are bought by companies who either must satisfy a government mandate (known as the compliance market) or have voluntarily committed to reducing their carbon footprint (the voluntary market).
These registries require that carbon projects follow strict protocols for estimating biomass, both at the start and throughout the life of the project. Traditionally, carbon has been estimated using boots-on-the-ground field sampling, which is expensive. Remote sensing offers an affordable alternative—but the uncertainty in this approach has proved to be a sticking point.
All uses of remote sensing involve uncertainty—imperfect or unknown information.
Remote sensing datasets contain errors of measurement in space, where collected data is misaligned with its true position on Earth’s surface. Moreover, when we simplify natural systems such as forests in order to represent them in models, there is uncertainty regarding the degree to which the model accurately represents the system. A model that estimates the number of trees in a certain patch of forest or the amount of carbon in a particular tree can never be 100% accurate.
Different technologies have differing levels of uncertainty. Lidar and synthetic aperture radar (SAR), for example, have lower uncertainty than optical imagery because they can penetrate the top of the tree canopy and provide information about the height of trees, allowing more accurate carbon estimates. With SAR data, however, uncertainty increases as terrain becomes more varied and the density of forest increases.
It’s not enough to know that uncertainty is present; the uncertainty must also be quantified at a project scale. Quantifying uncertainty is necessary in order to describe our confidence that the data and analysis are accurate. This helps data producers identify areas for improvement, and it helps end-users understand the limitations of the data.
In forest carbon projects, increasing uncertainty can lead to lower financial returns. If uncertainty in total biomass for a project exceeds certain confidence intervals (90% for the voluntary market, 95% for compliance), carbon registries will reduce the number of credits awarded.
In other words, uncertainty can be costly: if remote sensing introduces too much uncertainty, the loss of carbon credits could erase any financial advantage offered by avoiding expensive field inventories.
Mapping with remote sensing data involves model-based estimates of biomass. First, the relationship between remote sensing data and carbon is modeled at a select number of points. Then, the model is used with remote sensing data to estimate carbon for the rest of the property.
Unfortunately, techniques for quantifying the uncertainty of these methods are not yet well established.
The new project—titled “Developing a Framework to Quantify Uncertainty and Harmonize Diverse Earth Observation Estimates of Forest Carbon”—seeks to fill that void. It is funded through NASA’s Carbon Monitoring System, an initiative that supports research into global carbon stocks.
The research team will build a framework for quantifying uncertainty based on a mathematical approach known as hierarchical Bayesian modeling.
The team will identify several test projects in California, Oregon, and Washington where traditional field estimates of forest carbon have already been completed. They will then use the Bayesian framework to estimate biomass and quantify the uncertainty of those estimates for several different remote sensing technologies. Finally, they will evaluate how the uncertainties estimated using remote sensing compare to those derived using traditional methods.
The framework is designed to allow for continuous improvement. As new technology and data products become available, the framework provides a means to evaluate how these data products and maps affect estimates of uncertainty.
The team will also create a web-based tool that will allow the public to estimate carbon stocks for any geographical area.
Putting Research to Work
“We've been using remote sensing to estimate carbon for decades,” says David Saah, Principal Investigator with SIG. “The challenge has always been making it consistent with inventory measurement in the field. How do you integrate uncertainty between the field-based measures and Earth observables?”
This NASA-funded project will help answer those questions.
Though the research will be applicable to a broad variety of applications that rely on biomass estimates, forest carbon development projects should find it particularly useful.
“The NASA ROSES grant is evaluating this new technology and helping to bridge it into the carbon registries and protocols,” Kramer says. “Identifying uncertainty at different spatial scales and validating that uncertainty at a project level is really important.”
The ability to identify uncertainty at different spatial scales will allow carbon markets to operate more efficiently.
With increased transparency, buyers can be confident that their carbon credits accurately represent real carbon stocks on the landscape. These new approaches will also lower the cost of entry into carbon markets, allowing owners of small- and medium-sized properties to participate—thereby providing financial incentives for managing forests in ways that both fight climate change and improve biodiversity and wildlife habitat.
“Government agencies, private companies, and nonprofit conservation groups all have an interest in increased use of Earth observation tools in forest carbon mapping,” Kramer says. “This research will help them use those tools with more confidence.”