How PG&E Is Reducing Wildfire Risks Using Satellite Imagery

Wildfires have had a devastating impact on California and on the state’s largest utility company, Pacific Gas and Electric (PG&E). PG&E’s equipment has been linked to several major wildfires in the past including the 2018 Camp Fire (the deadliest wildfire in California history, killing 85 people, according to CAL FIRE, the state agency responsible for, among other things, protecting natural resources from fire) and the 2021 Dixie Fire (the second-largest wildfire in state history, burning more than 963,000 acres).

Potential wildfire liabilities exceeding $30 billion led PG&E to file for bankruptcy in January 2019. The company emerged from bankruptcy on July 1, 2020, with a renewed focus on mitigating wildfires within its 70,000-square-mile service territory in northern and central California.

Identifying Wildfire Risks

“A lot has changed,” Andy Abranches, senior director of Wildfire Preparedness and Operations with PG&E, said as a guest on The POWER Podcast. “We really saw the devastation that could occur from these wildfires, and so, that was the point that PG&E started really making a big pivot to addressing the wildfire risk. The way we address the wildfire risk is really through what we consider our layers of protection. We started initially learning as much as we could from San Diego Gas and Electric [SDG&E], and put in place the public safety power shutoff program.”

High-fire-threat district maps were important in understanding risks. About half of PG&E’s service territory falls in high-fire-threat areas. “We have 25,000 distribution miles that run through the high-fire-threat districts and 5,000 transmission miles,” said Abranches. Vegetation plays a critical role in the risk, and while precisely quantifying the number of trees in and around those risky transmission and distribution lines is difficult, Abranches estimated it’s in the range of eight to 10 million.

With such a large area and so many trees to monitor, PG&E turned to Planet Labs, a San Francisco-based provider of global, daily satellite imagery and geospatial solutions, for help. Planet’s satellite-derived data on vegetation, including canopy height, cover, and proximity to electric-system infrastructure, is used by PG&E to prioritize the mitigation of vegetation-associated risks.

Quantifying Threats and Consequences

Abranches explained PG&E’s risk characterization process by likening it to a bowtie. “The first part of your risk bowtie is: ‘How do you quantify and in a probabilistic way build a risk model to predict ignitions are going to happen?’ ” He noted that the biggest source of ignitions is through contact with vegetation, such as a tree falling on a line or a branch coming into contact with a line on a windy day, but birds and other animals can also cause ignitions.

“The second half of the bowtie is the consequence,” said Abranches. “If an ignition occurs at a particular location, if the vegetation around it is just not there, that ignition will never spread.” The fire triangle requires heat (or a spark), oxygen, and fuel. The fuel is the vegetation bed around the line where the ignition event occurs. If there happens to be a lot of dry fuel, that’s when an ignition becomes a wildfire. Depending on the oxygen, which can be heavily influenced by wind conditions, it could become a catastrophic fire, Abranches explained.

“As we built our risk models, you needed to understand the vegetation dimension on two levels. One level is for probability of ignitions: ‘How do we get better at predicting where we expect vegetation ignitions to occur?’ And the data that we’re able to get from Planet every year helps improve and keeps those models updated,” said Abranches. “The second piece of it is the consequence of the ignition—understanding the fuel layer. That also—data from Planet—helps inform and continually refreshes that information to make sure it’s most current. So, the risk model actually uses the Planet data on both sides of the bowtie, because it’s probability of ignition times the consequence of ignition gives you the risk event.”

Measuring Forest Structure and Health

“The satellite imagery is really great as a measurement tool. It measures spectral reflectance from sunlight that is absorbed and reflected by different surfaces,” Chris Anderson, science lead of the Forest Ecosystems team with Planet, said as a guest on the podcast. Anderson said different insights are gleaned from trees with dark green versus light green leaves. Other factors that could affect understandings include time of year and position of the sun, as well as the calibration of instruments.

“Parsing through these different sources of variability is one of the real value-adds of artificial intelligence,” Anderson said. “We’ve been working on developing finer and finer metrics of forest structure and forest health to be able to quantify some of those patterns, and then provide those metrics of forest health to PG&E’s team to help them better characterize these patterns and help them identify where these potential failure modes may occur.”

The various vegetative species also factor into risks. Some vegetation greens up earlier in the year and perhaps browns more quickly, while other species may put out seeds or flowers that change fire risks. “You can analyze and capture this information with high-frequency satellite imagery to be able to map out those differences,” said Anderson.

The results obtained through all the work PG&E has done are quite remarkable. The company calculated a 72% reduction of ignitions in high-fire-risk areas in 2023 compared to the three-year average. Furthermore, the size of the fires that have occurred from those ignitions has also come down dramatically.

Future Enhancements

Meanwhile, further risk evaluation improvements are coming soon. Abranches mentioned hyperspectral capability as a feature that could be implemented within a couple of years. “That’s something that will add another degree of richness to this information set that we’re very much looking forward to,” he said.

Anderson said the mission of hyperspectral satellites is really oriented around pinpointing, quantifying, and tracking point-source methane and CO2 emissions, but the technology can be useful for many other things. “Hyperspectral satellites can identify the spectral signatures of chemicals and materials and processes around the world, and they can reveal some of these hidden trends that can fill intelligence gaps and mitigate risks,” he explained.

“In our use case, we typically use this higher dimensionality—this really rich dataset of imaging spectroscopy data—to better characterize some of these patterns, like species richness—where they are invasive species versus native species, where you have the effects of post-burn fire severity, where you have increased grassland biomass,” Anderson said. “The additional spectral depth of these sensors gives you much more explanatory power as to what’s happening on the ground level with the vegetation, which can help better characterize some of the upcoming consequence effects of these areas, and also gives us more precision into the makeup of forest health across large areas.”

Abranches suggested the work California’s three public utilities—PG&E, SDG&E, and Southern California Edison—have done is making a difference for customers served by all of the companies. “With weather patterns changing, extreme events are putting upward pressure on the risk. But we’ve made huge strides in California, in partnership with companies like Planet that have given us information to act on, and all three utilities have taken the action to make the community safer,” he said.

To hear the full interview with Abranches and Anderson, which contains more about wildfire risk management and satellite data analysis, listen to The POWER Podcast. Click on the SoundCloud player below to listen in your browser now or use the following links to reach the show page on your favorite podcast platform:

For more power podcasts, visit The POWER Podcast archives.

Aaron Larson is POWER’s executive editor (@AaronL_Power, @POWERmagazine).

SHARE this article