Applied Science
Initially, the NIMO team provided decision support, but within a few weeks they assumed command of the fire. First, the team, along with personnel from the Los Padres National Forest began the Wildland Fire Situation Analysis process, outlining suppression objectives to the incoming IMTs. As they began gathering information on predicted weather, available resources, values at risk, and the actual fire perimeter, they were contacted by Bernie Bahro, the U.S. Forest Service Pacific Southwest Region liaison for a new Rapid Response Decision Support Team, regarding use of the new Wildland Fire Decision Support System FSPro and RAVAR models in support of the WFSA. (See WFDSS Features sidebar.)
The Boise NIMO Incident Commander, Aaron Gelobter, decided to take a hard look at the new decision support tools and attempt to use them for long-term strategic planning. First, working with personnel from the Los Padres National Forest, he used the FSPro runs to establish WFSA lines and RAVAR to estimate values threatened — more than $66 million.
However, fires often have a habit of disregarding well laid plans. The Zaca exceeded the WFSA two days after the first WFSA box was established, and it surpassed a second WFSA soon after. Gelobter went back to Bahro and asked him to adjust the model, calibrating it to match the actual rate of spread, and using more accurate estimates of fuel moisture. The new outputs were used to draw a much larger WFSA boundary that used the fire spread probabilities from the model to better understand the potential size of the fire, guiding their AMR approach to management of the fire.
"The WFDSS tools were useful initially, but not in the way they were intended. The FSPro, RAVAR and Stratified Cost Index analysis were supportive of projected suppression expenditures. Accordingly, a $70 million to $80 million fire would be cost effective [This quote came from an interview early on during the fire. The fire eventually reached $120 million], and the FSPro runs sustained drawing a large WFSA box, creating more flexibility to do more objective-driven AMR. So, the tools were useful in that sense, but they just weren't reflecting what was happening on the ground. So, we asked Dr. Mark Finney to calibrate the models in order to more realistically represent current burning conditions and provide more value to the manner in which line officers using these tools would make their decisions," says Gelobter.
As it became clear that Gelobter and the NIMO team were serious about using the tools to do long-term planning on the fire, Mark Finney, the developer of FSPro, reported to the fire to interact more closely with the team, and see if the model could be adjusted to better fit their needs.
At first glance the maps, produced by FSPro look like other rate of spread maps produced by programs such as FARSITE (also developed by Finney) — colored, concentric circles radiating across a topographic map. However, a closer inspection reveals an important difference that is essential for interpretation. The contour lines do not represent rate of spread projections or fire perimeters — they correspond to probabilities of fire spread. Where programs such as FARSITE use existing weather forecasts to model fire spread, FSPro runs hundreds of different weather scenarios, reconstructed from historical data collected over the last 10-20 years by RAWS (such as the Figueroa RAWS described in the introduction) and then uses those runs to develop a map showing the probability that the fire will reach a certain spot over a certain time period - usually 7 to 14 days.
Finney describes the difference this way:
"A FARSITE map is essentially a fire spread projection using one weather scenario. And, an FSPro map is the result of hundreds of FARSITE runs based on historical weather scenarios. You can take any spot on the map and see how many times the fire reaches that spot within the given time period under all of the different scenarios. Once the model finishes running all of the different scenarios, it produces an output map with probability contours linking all of those spots, showing the likelihood that a fire will reach a certain area within a given time period. So, if the model runs 1000 different weather scenarios and the fire reaches a point in 800 of those runs, the point will be located on the 80% contour line."
Acceptable Use Policy blog comments powered by Disqus
