Discussion on Using the Fire Signature Method to Predict Risk

from theysaid 1/28/04

Doug Campbell,

You have done a lot of "Finding a Solution" for the different historical fires that have killed people.
Not to be in your face, but I could argue that what you're doing is a little like monday-morning-quarterbacking. In building predictive models in science, the first step is to use data to create the best model to fit the what is observed. The second step is that the model is validated or used a priori (ahead of time) to actually predict the outcome. As I understand it, your first choice for making a prediction for fire behavior is to find another place that has burned that is similar to the new location where you want to predict the fire behavior. This becomes the "Fire Signature" for the new location. If no such example is available, you say you have to guess (use intuition, experience, I don't remember your exact words). That's where slope, aspect, time of day (solar pre-drying conditions that you used to call pre-heat), alignment of forces, time-tagging, etc come in.

Can you describe when you have used the Fire Signature Prediction Method to make decisions that resulted in actions that succeeded or failed. Have you ever actually used it in that way? What was success? Did your prediction ever fail? In science the failures go in the bottom desk drawer and rarely see the light of day. We call it the "Bottom Drawer Phenomenon". Do you have any failed predictions in your bottom drawer?

The success of a model of a complex event is that it provides some order for looking at the multiple predictive variables and that the model better predicts the outcome than the unweighted variables. Good models simplify. Similarly our brains want a way to simplify complex events -- put them in categories that help us survive and function better. (This is why Bayesian probabilities/ analysis has gained popularity in statistics. They're about conditional probability, for example, what is the probability of this outcome occurring given that this other event preceded it? Preceding events are our experiences, knowledge structures, assumptions. Still gotta read that NY Times article...)

The benefit I see in your model, is that it is a way to begin to logically organize and account for complex variables, looking first at what experienced firefighters consider the most important... but keeping in mind the other variables that might be important at some time. So Doug, what other fire behavior factors might need to be considered beyond those most important variables in your model - slope, aspect, time of day? ...Like a cold front coming through, wind-driven fire behavior. What else? How do you account for them?



1/29/04 Reply

Thanks for asking me those hard questions. Monday-morning-quarterbacking can be good or bad I guess. Accident investigations have parts where there is proposed mitigation designed to prevent a similar situation causing a similar accident. Isnt this MMQing? Isnt this the way agencies have always done it? The MMQing is often thought to be something that should be criticized but I do not agree with that line of thought. Accident reports consist of a gathering of facts and witness and victim statements then mitigation measures. Isnt that MMQing? What I try to accomplish is to offer mitigation measures for an accident with the focus on the idea that if the fire behavior was accurately predicted and the crew moved to safety sooner no one would ever be burned over.

What one usually reads in the official accident reports is this core concept. As often as there is an accident investigation, there is a statement telling of how many violations there were of the 10s and 18s and if they had not been compromised, the accident would have been avoided. My preference is to teach firefighters how to recognize fire behavior potential and stay out of the way of a fire run. Is it better to focus on the rules of engagement to be able to avoid an accident or is it better to see the fire potential and get out of the way? One without the other is not enough. Training that teaches when and where fire behavior change potential exists on the fire-ground in the path of the fire is another concept to explore. A reliance on obeying the rules of engagement without knowing wildland fire behavior has not been an effective guarantee of accident avoidance. Firefighters need to know the rules and focus on LCES but without knowing the fires potential to change behavior all the rules cannot be assured to save you.

I like to test them and look at the record of accomplishment to determine whether they were in fact sufficient mitigation. Consider a typical mitigation measure the agencies designed. Take the Loop fire of 1966 where the mitigation for the 13 fatalities the El Cariso Hot Shot Crew sustained were the downhill line construction standards that were added to the requirements of line construction. Consider the South Canyon fire of 1994 where 14 fatalities occurred while attempting a downhill line construction. Did the downhill rule work? Why not? What if the crew could see they were in an area that was in full alignment and was an extremely hazardous area? If they knew that and could talk about it, dont you think they would have moved out of the in alignment path of the fire? Could anyone explain the hazard well enough to stop a fatal action? Apparently not. Why not? Maybe just maybe we havent trained firefighters to be able to describe the potential and the tactical solution to the situation. Were the mitigation measures put in place after the 30 mile fire effective on the Cramer Fire?

Striking a balance between MMQing and successes on actual fires.
The CPS training program presents four fire problems in the workbook that students are tasked to solve. Two (2) problems are fires where accidents happened on and two (2) are fires that by using CPS, similar accidents were averted. I have used CPS on wildland fires for more than 20 years and have participated in many situations that resulted in pre determining the potential and time tagging or trigger pointing the tactical change prior to the fire endangering the position of the crews. I have evidence in various forms in my personal files. I have been doing the escape analysis for Rx burns for over 10 years and have made predictions of the weak portions of the perimeter and have suggested mitigation measures that worked. I have predicted the probability of escapes and had the experience of seeing them do what was predicted.

Mellie you asked if I experienced any failures. Predictions of any kind are not usually perfect representations of reality. No one should expect that. Are weather forecasts accurate down to the time and place of wind speed and direction? If the weather forecast is not exact, then anything that depends on that forecast is subject to error. What is important is that the prediction/forecast is sufficient to avoid injury. I remember an FBA forecast that was posted on one of the Yellowstone fires that used BEHAVE and missed the ROS and flame lengths by 2,4,& 6 times. The next forecast recommended using these factors to tweak the outputs of the model thereafter. No one got hurt but was the forecast accurate, no. The fire modeling program often is said to have been in error.

CPS teaches that a prediction should be, 

During my time using CPS, I never had my prediction leave anyone in a tight spot nor did they result in causing endangerment. On the contrary, there are many occasions that are cited in our class of situations on the fires of 1987 and 1988, the Dillon fire, the Pony Peak fires of 1994, the Marre Fire, Green Meadow fires of 1993 and as recently the 2003 Piru fire. CPS prediction methods aided in determining the tactics and avoided potentially serious accidents.

Students in CPS classes offer many insights into their successes and so an exchange of learning happens.
There are many firefighters that have enviable track records. These people know wildland fire. I did not invent these things but learned them from some of the best. I simply tried to utilize the body of expert knowledge and to do something with it that could lead to better mitigation measures than were offered me by the management. Until I learned what my mentors knew I was at risk and so were those who were with me. I was lucky to survive until I learned enough to finish up my days without becoming a victim myself.

Other factors that are taught and considered in making predictions using CPS are in part: 

I sure hope that I have answered your questions. This response is so long that maybe you should just give me a grade and tell the folks if I passed or not. This paper is not a good substitute for a class in CPS.

Thanks again Mellie
Best regards to all wildland firefighters.
Doug Campbell