Skip to main content
Close Search

Doctoral student receives national recognition for characterizing peatland fires

Burned area in peat swamp forest
Satellite data must be used to characterize these burned areas in the Indonesian peat swamp forest because no burn maps exist.

South Dakota State University doctoral student Yenni Vetrita received third place in the student poster competition at the 20th William T. Pecora Memorial Remote Sensing Symposium for her research aimed at identifying peatland fires in Indonesia to track carbon emissions. The national conference, held Nov. 13-16 in Sioux Falls, brings together scientists who evaluate land and water resources using Earth-observation satellite data.

In 2015, scientists estimated that Indonesian peatland fires contributed 5 percent of global carbon emissions. However, Vetrita, who worked in remote sensing at LAPAN—Indonesia’s National Institute of Aeronautics and Space, explained, “There is uncertainty here because we don’t have a historical burn map, so we rely on satellite-based mapping.”

Her dissertation research looks at how land use/cover and water stress affect recurrent peatland fires in Indonesia. Because no historical maps exist, Vetrita must first figure out which existing Earth-imaging data can be used to characterize the burned areas.

Vetrita said, “It is an honor to have my work recognized. This is ongoing research that will be useful for my country—and something that I will continue back home when I complete my doctorate.”

She met her dissertation adviser, former Geospatial Sciences Center of Excellence senior scientist Mark Cochrane, now at the University of Maryland Center for Environmental Science, at a workshop in Indonesia. “LAPAN was one of the stakeholders for peatland fire monitoring,” she explained. Cochrane’s NASA research on peatland fires drew her to SDSU in Aug. 2015. Her advisory committee chair is associate professor Xiaoyang Zhang.

How peatland fires began

Deforestation from an ill-fated 1996 plan to convert nearly 2.5 million acres of the Kalimantan peat swamp forest on Borneo into rice paddies altered, not only the landscape, but also the lifestyles of the inhabitants, who once relied on hunting and fishing. Villagers set fire to the peat to clear the land for subsistence level farming and burn timber left over from logging to make charcoal, which they can sell.

Yenni Vetrita
Yenni Vetrita

An international program, known as REDD, offers financial incentives to countries that reduce their carbon emissions; however, that involves tracking progress. That’s where Vetrita’s work comes in.

"We need to know where the fires are so we can estimate carbon emissions,” she said. That means documenting the history of fires. For instance, areas that experience more frequent fires have fewer emissions because the trees have already burned so what remains is herbaceous vegetation, such as shrubs and bushes.

In addition, Vetrita pointed out, “the emission rate is different for fresh peat versus peat that has burned before and thus contains more char and ash.” That also affects the temperature of the peat during burning and the smoldering, which creates the smoke that affects southeast Asia.


Identifying burned areas

Vetrita compared three available datasets from moderate resolution imaging spectroradiometers aboard Terra and Aqua satellites, known as MODIS, to baseline data obtained through a French satellite called SPOT 5 to map peat swamp fires in 2014. Because of its finer resolution, SPOT data detects small fires that burn less than 300 acres.

Each MODIS dataset uses a different algorithm, explained Vetrita, who validated the model using a 2014 map focusing on a reference site in which fire activity was well defined from baseline studies.

She evaluated 3 MODIS products. MCD45A1 Collection 5.1 missed 99 percent of the fires. “It was the least reliable product for this application,” she said. The two versions of MCD64A1 fared better, but still underestimated the total number of fires by approximately 60 percent.

Though small fires accounted for approximately 70 percent of the total 2014 peatland fires, they were also a main source of error in the MODIS data. By excluding small fires, the accuracy rate for MCD64A1 version 5.1 increased from 46 to 58 percent and version 6 from 37 to 47 percent.

However, when she tested the two MCD64A1 products during the 2015 fire season, no burning was detected in parts of the reference site that were known to have burned. Yet 2015 was the biggest fire year since 1997. That is, in part, because the heavy smoke interfered with fire detection over such a huge area, Vetrita explained.

Using data from Landsat, which takes images of the peatland forest every eight days, and Sentinel-2A, another European satellite which captures data every 10 days, Vetrita was able to characterize fire ignition and development in 2015 which could not be detected by MCD64A1 products.

“Pre-fire is easier to get; post-fire is difficult,” she explained. The fires typically burn from August through October, but the rainy season that puts out the fires starts in November. “From then until March, thick clouds make it difficult to produce an accurate map,” she said. However, the good news is that “the spectral signal stays longer for an area that is initially forest,” Vetrita noted. “For more than three months, I can see the area of burn, but in herbaceous area the regrowth is really quick.”

Despite the challenges, Vetrita is determined. “By providing a better map, we can reveal the truth and thus come up with a solution. I have a responsibility to do this as an Indonesian citizen and a government employee.”