Lessons Learned from the November 2008 California Wildfires

The 2009 wildfire season may be just beginning, but what lessons can be learned from California’s November 2008 wildfires?

The Tea Fire, the Sayre Fire, and the Freeway Complex Fire that burned in Southern California last year provided risk modeling firm AIR Worldwide with real-life examples to compare damage patterns with its models functionality in determining loss estimates.

The Tea Fire, which got its name from Montecito’s historic “Tea House” near which it ignited, began on Nov. 13. By the time it was over, some 1,900 acres and more than 200 houses had burned in and around Montecito in Santa Barbara County.

On Nov. 14, a resident of Sayre Street in the Sylmar section of northern Los Angeles reported a fire that came to be known as the Sayre Fire.The fire burned more than 11,000 acres and more than 500 structures.

The Freeway Complex Fire ignited on Nov. 15 along the 91 Freeway in Corona, Calif. On Nov. 16, the smaller Landfill Fire in Brea merged with the Freeway Complex Fire. When it was over, more than 30,000 acres near the nexus of Orange, Riverside, Los Angeles, and San Bernardino Counties had burned, and some 300 homes were damaged or destroyed.

An important source for real-time perimeter information during wildfires is the Geospatial Multi-Agency Coordination Group, or GeoMAC (www.geomac.gov), a consortium of federal and state agencies. Fire management officials use GeoMAC information for resource allocation decisions. Risk modeling firms also rely on GeoMAC perimeters to produce reliable real-time loss estimates.

To estimate the perimeters, the GeoMAC consortium uses GPS data and infrared imagery from fixed wing and satellite platforms. The most common method for collecting GPS data involves flying helicopters along the perimeter of a still burning fire, as low to the ground as possible. As the pilot attempts to trace the perimeter, a GIS specialist with a handheld GPS unit takes readings.

The process creates uncertainty about the reported perimeter. For example, some experienced pilots are able to make 90 degree turns, while others are hesitant to fly into very thick smoke. It can also be difficult to see through the smoke to determine how far the fire itself has actually progressed. The helicopter data is supplemented by MODIS satellite hotspot data and reports from fire crews on the ground to create the final reported perimeter.

As a result, the accuracy with which published perimeters matched actual perimeters on the ground can vary considerably from location to location. For instance, the perimeter survey of the Tea Fire as reported by GeoMAC proved to be reliable. Yet, the reported perimeters for the Sayre and the Freeway Complex Fires were less reliable. At least in some neighborhoods, the fire’s progress was stopped successfully while still in the wildlands (or reached one or two houses into the end of a dead end street), while the reported perimeter indicated penetration into housing developments tens or even hundreds of meters deep. Clearly, such discrepancies can impact post-mortem loss estimates.

A loss model approximates the probability of damage based on event wind speed, topography and the nature of fuels present, for any structure contained within the perimeter. By including houses that should not be within the reported perimeter, loss estimates are inflated.

Conversely, losses may be underestimated if the reported perimeter fails to include some damaged exposure. It is near the perimeter that wildlands typically meet the urban environment, and therefore losses during a wildfire tend to cluster there. Inaccuracies in how the perimeter is reported can lead to uncertainties in estimated losses.

If losses resulting from smoke damage are excluded — a type of damage that is challenging to assess without access to he interiors of buildings — that also could lead to loss estimate inaccuracies.

Perimeter reports affect the ability to estimate how much damage is likely to occur given that a building sustains some damage — also referred to as the conditional damage ratio within a fire perimeter. In the November fires, if a house was burned at all, then in the majority of cases (roughly 90 percent) that house was a total loss, while in the few remaining cases (roughly 10 percent) the damage was typically minor, (e.g., a window or a garage door destroyed).

Estimating what fraction of buildings within a perimeter sustain some level of damage — that is, the unconditional damage ratio — can be more challenging. This is closely related to the problem of uncertainties in fire perimeters as reported by official agencies. In the November fires, many neighborhoods within the reported perimeters, as mentioned earlier, had no visible damage whatsoever. In areas that clearly had some damage, there were streets with widely varying damage ratios.

For example, on Aviemore Drive in Yorba Linda, six homes in a row were destroyed. On the other hand, on nearby Rolling Hills Lane, there was only one total loss out of nearly three dozen homes. At the UCLA Olive View Medical Center, while surveying the Sayre Fire, there was little visible damage to the buildings, yet most of the buildings suffered substantial smoke damage. Numerous other buildings had no visible damage, yet their contents were in piles outside.

When estimating losses from fires, it’s important to remember that officially reported perimeter coordinates vary in quality from fire to fire, due to the challenges of constructing such perimeters. This speaks to the always-present uncertainty involved in real-time loss estimation. Modeling firms are in contact with several governmental and private sector organizations, pursuing cooperative arrangements to help acquire more accurate real-time and post-mortem fire perimeters in the future.