Predicting When and Where the Next Big Fire Will Burn in California

 As anyone living in the Golden State knows, California has been suffering through an unprecedented increase in fires, many of them burning through forests in the Sierra Nevada. Incredibly, as noted in a recent paper by Hiraga and Kavvas in the journal Fire, more than 7% of the land surface in California burned in 2017 - 2020. These fires have resulted in hundreds of deaths, the destruction of tens of thousands of structures, and severe financial losses.

Sparked by lightning, the North Complex Fire in the northern Sierra Nevada destroyed 2500 buildings and led to 16 fatalities. At 320,000 acres, it was the 6th largest fire in California history and the deadliest fire in 2020. Photo credit: Brian Bahouth, https://www.sierranevadaally.org/

Understanding the precise meteorological and hydrological conditions that lead to wildfires in California is important for a variety of reasons. First, this information can help in mitigating ignition sources. Some of the worst wildfires have been triggered by electric lines coming into contact with vegetation. If the conditions in a certain area were deemed to pose a serious fire hazard, a Public Safety Power Shutoff (PSPS) could be initiated to de-energize the lines. Second, fire-fighting resources could be positioned in areas with dangerous meteorological and hydrological conditions in anticipation of potential fires. Third, knowing the conditions that lead to fires will help anticipate future fire risk and fire behavior as the climate continues to change.

To investigate the conditions that led to recent wildfires in Northern California, Hiraga and Kavvas studied 24 large wildfires (> 10,000 acres) that burned in the Sierra Nevada and the Coast Ranges in the years 2017-2020. For each fire, the authors obtained the following data: vapor pressure deficit (a measure of how 'thirsty' the air is) 2 meters above the surface, horizontal wind speed 10 meters above the surface, wind gust speed at the surface, and soil moisture. From these four variables, the authors developed three indices. The first, MDI, is calculated as the vapor pressure deficit divided by the soil moisture; this is, essentially, a measure of the aridity of both the atmosphere and the soil. The second index, MDIWIND, is calculated by multiplying MDI by the horizontal wind speed. The third, MDIGUST, is determined by multiplying MDI by the wind gust speed. MDIWIND and MDIGUST, then, express how hot, dry, and windy the conditions are. These three indices were calculated for the ignition location of each fire, as well as for the surrounding area. To provide a comparison, the authors calculated these indices for these same locations but for the years before the fire.

The results from these analyses are illuminating. For 23 out of the 24 fires, the MDI was higher at the ignition point when the fire began than in previous years, whereas the vapor pressure deficit was higher at only 21 of the fires. This finding highlights the importance of the aridity of the soil, in addition to the dryness of the air in creating conditions conducive to fire growth. Both indices that account for wind, MDIWIND and MDIGUST, were also higher during the time of the fires; interestingly, both of these indices differed according to fire size, demonstrating that wind speed and aridity contribute to fire growth.

To provide further insight into the relationship between these three indices and fire ignition, the authors presented plots of their data through time for four of the fires. In each of these cases, the fire began when at least two of the indices had reached a maximum. Peaks in MDIGUST, in particular, appear to have the greatest predictive power with respect to the ignition and spread of large wildfires.

Shown below is a plot of the three indices through time for the 2020 North Complex Fire (labeled here as the Claremont-Bear Fire) in the Sierra Nevada. The yellow bar indicates the ignition date. The fire clearly began during a period of time when all 3 of the indices had reached a peak. This is from Figure 2 of Hiraga and Kavvas (2021).

In addition to analyzing how these indices change through time, the authors also looked at their geographic distribution at the time of the fires and found a good correspondence between where the fire started and high values of MDIGUST, the index that accounts for the speed of wind gusts and the dryness of the soil and atmosphere.

In the map below, values of MDIGUST are shown along a blue-red color gradient, with red indicating the highest values. To get oriented, note that the lower red line intersects the San Francisco Bay and that the border between California and Nevada is shown as a black line near the right edge of the map. The dashed lines outline areas with particularly high values of MDIGUST. The ignition location, shown as a black circle just to the right of the center of the map, overlaps with an area of high MDIGUST values. This is from Figure 3 of Hiraga and Kavvas (2021).

Considering that this study only considered four variables, the results are quite good and should encourage others to adopt this framework for predicting when and where these fires might start. The authors note that this approach could be improved by incorporating two other variables. First, including data about fuel load (eg., fuel availability, fuel moisture) would be an important next step. Second, since many fires are ignited by human activity, incorporating information on accessibility and land ownership would be also be helpful for identifying possible ignition sites.