IE11 Not Supported

For optimal browsing, we recommend Chrome, Firefox or Safari browsers.

Can Machine Learning and Collaboration Protect Forests?

A project using multiple data sets and lidar technology can predict and model how fires will affect specific forested areas and help balance the interests of sometimes-competing stakeholders.

Along with a warming climate, one of the factors contributing to the trend of more intense forest fires is the way the forests have been managed.

The longstanding strategy has been suppression: Put out the fire as soon as possible. But that course of action has left the forests with far too much fuel in the form of dense underbrush, a bunch of younger, smaller trees and too few larger, bigger trees.

The combination yields what we have today: more intense, hotter, faster-burning fires that prove devastating. But the solutions to the problem vary depending on whom you ask. For instance, controlled burning may serve one stakeholder — the one that wants to clear underbrush — but cause harm to another, like the one who is trying to protect an endangered species.

But a new machine learning solution allows the different stakeholders to manage risk and understand and share strategies to combat the growing fire issue. The solution uses multiple data sets combined with lidar (light detection and ranging) images to create 3D maps in a localized area. The program then shows what a controlled burn would do and how far it would go, or what effects a firebreak or selective thinning would have.

The result is that you end up with the best solutions for different geographical areas managed by different stakeholders with different interests.

The Truckee Fire Protection District in California’s Sierra Nevada historically had little access to high-quality data or advanced planning tools to develop wildfire plans and assess needs. The district sought to protect its alpine community from constant wildfire risk while collaborating with other jurisdictions on fuels treatment.

The district deployed the machine learning process created by Vibrant Planet to help develop plans and reach consensus among more than 15 stakeholders in the area, including the U.S. Forest Service and private landowners, by assigning priorities based on specific land use. Those include balancing the sometimes competing interests of protecting physical assets and protecting biodiversity and creating treatment for each without laborious planning sessions.

“We’re creating a prioritized mitigation road map for land treatments going forward using Vibrant Planet, with the goal of enhancing community wildfire protections,” said Dillon Sheedy, assistant wildfire prevention manager with the Truckee Fire Prevention District. “We want our stakeholders to use Vibrant Planet to prioritize the things they care about, and then we can work together on those scenarios instead of competing with each other.”

Vibrant Planet CTO Guy Bayes describes his clients as being mostly in “two buckets”: the U.S. Forest Service, and national forests that they manage. Within those are nonprofits and other entities that manage certain pieces of land and have mostly different priorities about what they want to protect and how to go about it.

“It’s a big treatment optimization engine where you can say, ‘OK, this is my area that I care about, this is the number of dollars, this is the time frame, I have these priorities,’ you hit a button and it will generate the optimal treatment to maximize your money and get you data on how much you reduced your risks,” Bayes said.

But it has to be done collaboratively by the multitude of stakeholders.

“One of the big barriers to getting stuff done is there are a lot of stakeholders that aren’t too crazy about setting fires in their backyards,” Bayes said of controlled burns, one of the top forest management solutions. “Water districts, towns, the Sierra Club, utilities — they all have different priorities about what they want to protect and what treatments they will tolerate. What the government has found is that if they’re not careful about managing these stakeholders, they end up in court.”

The Vibrant Planet project aims to find the best solutions for each group and allows them to collaborate and decide on the different treatments.

“We do optimized scenarios for stakeholders based on what they care about,” Bayes said. “I can do one for the Sierra Club that’s all about endangered species. I can do one for the water district for preserving reservoirs, and I can do one for protecting local towns from burning down.”

He added: “They all generate different plans, but there’s a commonality in the plans at the end of the day. Nobody wants the forest to burn down.”

This article first appeared in Emergency Management, sister publication to Industry Insider — California. Both are part of e.Republic.
Jim McKay is editor of Emergency Management.