Using Biomass to Estimate Wildfire Impact
On August 19, 2020, the California sky turned red. Social media was flooded with surreal images like the one above—no filter needed for this ominous cast. People in Santa Cruz county prepared to evacuate their homes as the smoke thickened. The CZU August Lightning Complex fire had come to town. How much biomass—think tree trunks, branches, and leaves—had to burn to paint the sky this color?
Growing up on the coast of Southern Maine, “wildfire” wasn’t really in my vernacular. Just like earthquakes, tsunamis, tornados, and pretty much any other natural disaster, wildfires were a far away something that I learned about in science class. But, every few years my family would take a long car trip up to Québec, Canada which is when I first remember understanding the phenomenon. Bouncing along in the backseat of the station wagon, I carefully tracked the scenery that whizzed by as a way to fend off my car-sickness. I noticed every change in the landscape, but I was most attuned to the patches of forest with lanky, barren trunks, some still sticking up out of the mush and others keeled over. These were the places, Dad said, that had been burned in forest fires, and if I paid very close attention, I might see a moose.
And so, when my friend and classmate Lauren Low asked me to join her in investigating the wildfires out West using ecological forecasting techniques, the curious kid that I was and still am said, “Yes please.” We then embarked on the most daunting part of the research process: choosing a question with the right scope to fit our crunched timeline. Can we determine which tree species that are most susceptible to and impacted by wildfire? Can we develop a model to predict which trees and ecological communities will be impacted by future wildfires based on the locations of wildfires that have already occurred?
How can we estimate biomass burned in California wildfires?
After doing some preliminary research on available data and project precedents, we submitted an official request for ForestGEO data from the University of California Santa Cruz Forest Ecology Research Site (UCSC-FERP). We chose this site in particular because of it’s central proximity to the prominent wildfires in California (see maps below). Upon approval, we also got an email from Prof. Jim Lutz, a member of the ForestGEO network who kindly compiled a list of adjacent research articles which gave us a strong basis for understanding what we could expect to accomplish.
Quick Win
Suffering from a case of what Prof. Albert Y. Kim calls “analysis paralysis,” we knew we had to just get something down on paper as imperfect as it might be. In other words, we needed to get a quick win to get the work flowing. With the help of the get_biomass() function from the allodb package, we used diameter at breast height (cm) and geographic location to estimate the above-ground biomass (kg) of each tree at UCSC-FERP. The resulting plot (shown below) is undeniably wonky with four distinct groups of points. But, it was a jumping off point, and that’s what matters early on.
Making note of the irregularities in the distribution of the data, we totaled the estimated above-ground biomass at UCSC-FERP to be 4,102,997 kg. With this estimate nailed down for our reference site, it was time to start the search for a comparable, “sister” site that had experienced a wildfire. Given our limited timeline, we kept our criteria broad: The sister site must be similar to our UCSC-FERP reference site in its ecological composition.
Minimally Viable Product
Keeping my expectations low was key throughout the next couple weeks of work, but this is easier said than done — especially as someone who always strives for perfection. I continually reminded Lauren and myself that we were researching in uncharted territory, and so a small amount of visible progress doesn’t negate hours spent searching, skimming, scrapping, and restarting. Keeping in mind the mantra of “simple yet effective,” we started piecing together a minimally viable product .
I continued exploratory data analysis by visualizing the species at UCSC-FERP according to their assigned coordinates, tally, and biomass (shown above). I also sourced supplemental facts relevant to our comparison; the 39.54 acre (~16 football fields) UCSC-FERP plot has 31 total species divided between Evergreen Coastal and Redwood-Dominated forest.
Meanwhile, Lauren made headway in identifying a sister site: the CZU August Lightning Complex fires. This massive wildfire was caused by lighting strikes in mid-August and decimated 86,509 acres of Santa Cruz and San Mateo county. In the thirty-seven days leading up to the full containment of the fire on September 22, 2020, there were nearly 1,500 structures destroyed, one injury, and one fatality.
With this being a first-attempt, we made the assumption that the UCSC-FERP reference site and the CZU wildfire sister site had similar ecological composition because they are both in Santa Cruz county. By solving a basic proportion — the kind of thing you see in seventh grade math—we made our initial estimate of biomass burned x = 8,976,888,403 kg. For context, an Olympic swimming pool is 1 million kilograms in mass, so the estimated biomass burned in the CZU August Lightning Complex fires is 9,000 times that.
This is a simple estimation approach, but is it accurate? How do we even go about testing for accuracy? Knowing that we had the true distribution of tree species for the UCSC-FERP reference site, it was time to search for these proportions in the context of the CZU sister site. Adding this information would help us move from assuming the sites are comparable to knowing that the sites are scientifically similar— an important distinction.
Because a portion of Big Basin Redwoods State Park was wiped out by the CZU August Lightning Complex fires, we started our next set of research there with hopes of finding government-commissioned forest data. A weekend of dead-end searching later, we were nearly out of luck. There were species lists, yes, but no record of a percentage breakdown of species. However, we did recover something of substance; research by Ecologist Roy W. Martin confirmed that Big Basin shares plant communities with the UCSC-FERP site: Evergreen Forest and Redwood Forest. Based on this commonality, we came closer to confirming that UCSC-FERP and CZU could be considered sister sites.
Next steps?
Tree species have a relationship with elevation. I’m no ecologist but even when I’m low on oxygen and hiking from one cathedral to the next, I notice changes in which species tend to grow and survive at certain altitudes. When I rise above the tree line, there are shrubs and lichen growing amongst the rocky face that don’t live any lower. Lauren also confirmed this pattern by doing a mock fly-over of Santa Cruz county on Google Maps.
Using this tree species versus elevation logic could be helpful in estimating percentage breakdown of trees at CZU August Lighting Wildfire site. For a simple example: Say we know Tree X grows at high elevations, Tree Y grows at low elevations, and the site is ~10% high elevation and ~90% low elevation. Then, we could determine that the ~90% of the site is populated with Tree Y.
Making small discoveries like this and stitching them together will lead us to a robust wildfire biomass estimation approach that balances goodness-of-fit and complexity. As climate change worsens and wildfires grow more uncontrollable, having the ability to quantify the ecological impact of wildfires could be an essential next step in helping us to better predict, prepare, and protect our communities.
