The pioneers of fire research, including Gisborne himself, had proceeded on the assumption that the science of fire spread could be discovered by repeated trial burns using fuel samples gathered from all over the country. Two big flaws appeared in this method that kept it from producing accurate science or useful information. As discussed before, it did not produce “controlled experiments” with built-in assurances that what was meant to be tested was in fact being tested and hence could be repeated and used as data for statistical inferences. Another difficulty was that, unless you have some good idea of what you are looking for and how to find it, you can approach infinity with nothing more than a mishmash of little things you know about a lot of little things—certainly with nothing that could have been put in a hand calculator and dropped with Wag Dodge into Mann Gulch to avert his tragedy.
It is at this point that “fuel models” become necessary and make their appearance. They are fashioned roughly like ready-made suits of clothes. It’s a case of picking the fuel model you think will come closest to fitting, having a few adjustments made in the sleeves and shoulders and always in the legs, putting the model and the adjustments in the calculator, and watching the predictions come out almost instantly, usually with a margin for error. But mathematical tailors improve with practice, this experiment the four of us were conducting being something of a test as to how much.
There were twenty-six years between the first scientific concept of the spread of wildfire and its general availability as a field of thought with practical applications (the result in 1972 of Rothermel’s Mathematical Model for Predicting Spread and Intensity of Fire in Wildland Fuels). Seven years later Robert Burgan of the Fire Lab developed a hand-held calculator program that could predict fire spread in the field. When thought has moved from general concept and definition to practical application in the field, thought has taken shape and become a field of thought.
At the time of our talks around the conference table in the lobby, thirteen models of different forest fuels were ready for use in the field: three for grasses, four for shrub, three for timber, and three for slash. As for environmental factors influencing burning fuels, coefficients have been developed to measure their effects on fire behavior. A coefficient shows the relationship between two factors. Here, for example, is one that Rothermel had already established and that will certainly be transferred back to Mann Gulch for a scientific as well as a simple human understanding of its tragedy: “The percentage increase in the spread rate [one factor of a fire] varies in proportion to the square of the percent slope [an environmental factor influencing the fire].” This is a tragic statement; it was very steep where they died.
NEAR HERE A CHANGE STARTED occurring in our procedure around the long conference table. The grammatical structure of our sentences changed places across the table. It had started with flurries of interrogative sentences ending with rising inflections from Laird’s and my side, which were followed in a much lower key by a patient monotone of declarative sentences from the mathematicians, who knew they would have to repeat and explain their answers. The first movement of this grammatical structure was expressive of Laird’s and my attempts to get into our heads, however fleetingly, what had been solidly installed in the computer, a “philosophy of wildfire.” But as the grammatical structure veered around, Rothermel and Albini were asking the questions, still patiently, and Laird and I were giving the answers, still flutteringly with a hurried exchange of nondescript prose before one of us uttered the agreed-upon answer. Much of this would be put back in the computer to be combed over by what the computer knew about fire spread. I believe the trade name for this is “input,” a name I suppose one has to accept for an aftereffect of fire. But it also is to be taken as a sign that we were nearing the final step back to Mann Gulch—getting its exact “facts” in order to pick the right fuel models and sets of environmental factors and then make the adjustments needed for a “close fit.”
Many times just the practical problems of gathering accurate data on local fire conditions can be as difficult and complex as the mathematics that follows. In this quest to seek for scientific as well as human causes of a tragedy, the speed of the wind was very important, so it is not a diversion to consider here the kind of information we tried to dig up and relate there at the conference table in order to determine the probable wind velocity (or velocities) that powered the fire on its tragic course. The data we needed came from three different sources: official weather reports, testimony of those present at the fire or closely connected with it, and our general knowledge about the behavior of winds on large fires, especially on fires that are blowups. We knew from the dispatcher at the Missoula base and from survivors that air conditions at the time the plane left Missoula were turbulent and that the plane ride to the gulch was so rough that one jumper became too sick to jump and the others felt ill; that the wind velocity at the gulch was so great that the plane dropped its cargo from an unusually high altitude to avoid entering the narrow, windswept canyon; and that the cargo when dropped was scattered over an unusually large area and took a fatally long time to collect. We knew the official wind velocity at Helena at the approximate time of the tragedy (at six o’clock in the evening it was nineteen miles per hour), but Helena is more than twenty-five miles away, on the side of a wide valley, whereas Mann Gulch is a narrow notch that connects with the twisted canyon of cliffs of the Missouri where winds are compressed and are often substantially higher. From our common experience with forest fires, we knew long before Mann Gulch that big fires add to their own wind velocity by the whirling motion set up when cooler air rushes into the lower part of the fire to replace the hotter, lighter air that has risen and escaped. We knew from several sources that this particular fire was a blowup, with fire whirls throwing burning cones and branches across the gulch and starting spot fires that soon were racing upgulch toward the crew.
All the survivors testify to the great heat of the fire and its high wind, but none left precise estimates except Jansson. “The final settling of the wind…was to a strong twenty to forty mile per hour wind,” and we should know enough about Jansson and the language of the woods to know that Jansson is not admitting he could not tell a twenty mile per hour wind from one blowing forty miles per hour. He is saying as precisely as a trained and involved observer could say that air conditions in Mann Gulch late in the afternoon of August 5, 1949, were highly turbulent and that the wind velocity fluctuated from twenty to forty miles per hour. In this story, these figures will be taken as the most precise we could find in historical documents, but when a numerical velocity of the wind at the time of the tragic race has to be made a part of the arithmetic, the figure of thirty miles per hour will be used with the understanding that it surely varied, as Jansson said it did.
With their questions, the mathematicians were gathering data about fire conditions in Mann Gulch so detailed that they could construct fire-condition maps of the critical stretches of ground in the race between fire and crew. Each map showed the critical changes in fire factors that the crew had to face as they raced from one piece of ground to another. As it happens, these stretches of ground are much the same divisions or scenes that a storyteller would mark off to show the progress of his tragic plot.
The end of the summer was coming and so was the end of the fire season, which put a temporary end to our conferences in the laboratory. But there was still time after the other three returned from the fires to get in some work before I left Montana for the year; although we were separated, the work went on and was completed in first form that winter. Laird was in Missoula and continued the conferences, and I could supply my part of the answers by letter or telephone. I couldn’t pretend to keep up with Laird when we were in the field, so we had tended to specialize—he swarmed over Mann Gulch until he could remember where even the individual trees lay rotting; I tried to compensate by carrying a packsack of historical records and by remembering most of the rest of them. I don’t think we made many mistakes,
although it is hard for me, even in old age, maybe especially in old age, to admit how much of the truth can escape.
14
I HAD FIRST HEARD OF THIS BUSINESS of fighting fire with mathematics instead of Pulaskis from an old-time firefighter. He didn’t think, much of this as business and certainly didn’t know much about it, but he said it was coming and was coming from a big building just next door to the Smokejumper base in Missoula, and he said computers did most of the work, which was counting. He said they counted burning sticks and some of the sticks were homemade in the lab. He obviously saw no great future for this business. But oddly he thought I should go out and take a look at it, and I did, because he was one of the best old-time firefighters in the Forest Service and nearly always what he said should be done I did if it had anything to do with forest fires.
Even before I went out to the Fire Lab I had an idea or what eventually turned out to be an idea. It was in the form of a somewhat smoke-obscured question: If mathematics can be used to predict the intensity and rate of spread of wildfires of the future (either hypothetical fires or fires actually burning but whose outcome is not yet known), why can’t the direction of the analysis be reversed in order to reconstruct the characteristics of important fires of the past? Or why can’t the direction be reversed from prophecy to history? The one great tragedy suffered by the Smokejumpers was fading out of memory before its outline had been cleared of the smoke of controversy, before the missing parts, perhaps some self-cultivated, had been recovered, before its deferred trial had taken place in public court, and before its suffering had finally been placed within the reach of the public that would like to remember and honor it with sympathetic understanding. I can’t say that the idea caught on like a crown fire. More like a spot fire, it started something that smoldered and kept growing, not of course without changing direction several times.
It is fundamental for an analysis of a past fire that it should be attacked with a method that combines two methods, the predictive and the reconstructive. Prediction by its nature depends largely upon the scientific method, which, quoting Rothermel, “usually conjures an expectation for an answer to a problem that has somehow been arrived at by logical deduction. In the case of fire behavior, the logic is supplied by mathematical models.” When you change from prediction to reconstruction, you combine the thirteen fuel models with every bit of information taken from observation, measurement, and the historical record. I don’t suppose we were the first to make a serious effort to use this two-in-one method, the predictive and the reconstructive combined, in trying to state as accurately as possible what happened in the critical stages of a complicated and tragic fire, but as a team we were fortunate in the diversity of our specializations, and no combination of investigators would again have access to the only two survivors of the fire, whom Laird and I were able to persuade to return from far away to spend a day with us in Mann Gulch. Now there is only Sallee, and I doubt that he will ever go back again. Even so, it took us some years to analyze the tragedy of the Mann Gulch fire by both predicting and reconstructing, because only recently has the science of fire behavior developed to a level where it is possible to analyze a legendary fire with any accuracy. There must have been a good many reasons drawing me back to the job, and certainly one of them was that, the more accurately the race between fire and crew was analyzed, the more it took on the form of the plot of a tragedy emerging from concealed to complete inevitability.
The tragic convergence of fire and men in Mann Gulch offers itself as a tragic model for a graph, the modern scientist’s favorite means of depicting what he wishes to present as clearly as possible. Drawn along axes of time and distance is one line depicting the course of the fire and one depicting the course of the men, and where there is a convergence of the two, graphically speaking, is the tragic conclusion of the Mann Gulch story; the two lines converging to this conclusion constitute the plot. Along each line are numbers which are turning points in the race between men and fire, and if the lines are viewed as a race the numbers mark off legs of the race, if they also have religious significance they are stations of the cross, and if they have literary significance they mark off acts of drama. If it is drama it has the same old five acts as traditional drama, but the acts are much shorter, possibly because modern wildfire allows no time for soliloquies.
The legs into which the race seems always to be divided seem always to be the same, as if nature had left natural markers there. The starting legs for the fire and the crew are different, at different ends of the canyon and going in opposite directions. The remaining three legs, however, are almost an overlap of each other, the fire now being behind the men and getting closer.
The coming account of the tragic race from start to finish is based on an unpublished paper by Richard Rothermel, which in turn is based on thirty-two pages of mathematical worksheets. By now you should know enough about how these mathematical woodsmen do business to be able to predict how these sheets are divided. The first big division is made up of four sections, a page or so on each of the most significant fire factors determining the spread and intensity of the Mann Gulch fire: “Fuels,” “Wind,” “Slope,” and “Fuel Moisture.” “Fuels” because fuel is the centerpiece of any fire and surely was of this one; “Wind” because a roaring wind virtually assures spot fires and a blowup; “Fuel Moisture” because the fire’s tragic intensity was made possible by the superdry fuels and air that came with a long dry period and record heat; “Slope” because the abrupt increase in the grade of the slope near the top of the ridge was one of the reasons why the crew did not reach it.
The second big division of the thirty-two sheets is entitled “Sequence of Main Fire,” and it is here near the end that a scientific analysis of fire and a literary analysis of it as tragedy come closest to being one. In this section, each of the five legs has at least a half-page commentary pointing out the major change or changes occurring in that part of the race, each commentary being followed by a full-page “Fire Behavior Worksheet” with twenty-three entries available to characterize the changing conditions on that leg.
It is not necessary, fortunately, to decipher all twenty-three entries to realize that, scene by scene, tragedy is approaching. Instead, let us take the word of our experts, who say fuels are the center of the tragedy; if they are, we should be able to perceive a tragic continuity joining scene to scene and increasing in intensity as we follow the changes in fuels that the crew encountered as they attempted to escape.
1. The start of the race for the fire (from point 13 to point 6). The fire had jumped the gulch near the bottom, which was narrow and “overgrown by shrubs and conifer reproduction.” The two fuel models that most closely fit the conditions of the bottom of the gulch are model 10 (timber litter and understory, etc.) and model 12 (mixed conifers, slash somewhat compact). These heavy fuels under the dry, windy conditions of Mann Gulch would have created a fire of very high intensity “on the order of four hundred to a thousand Btu’s per foot per second” with flames in the surface fuels from seven to ten feet high. Such conditions would generally have produced a very intense, although not fast-spreading, fire, but the strong surface wind at the time “would certainly have caused crowning and spotting, and crown fires with strong winds usually spread faster than surface fires in timber litter.” The main direction of the fire would have also threatened the crew coming downgulch. Not only would the wind take it upgulch but the slope would take it uphill, so on this diagonal sidehill angle it was on its way to head off the crew, which had not seen it yet. What the crew actually saw as it approached point 6 was black smoke boiling over a lateral ridge below and ahead of them. Black, boiling smoke is often what you see before you see the crown fire that is making it.
The distance from point 13 on the map, where the fire jumped the gulch, to point 6, the turnaround, can only be estimated; we have already once rounded it off to 400 yards. The location of point 13 involves some guesswork if for no other reason than that no one eve
r knew how many spot fires jumped the gulch and took off. But the spot fires’ location is not as important at this point as Dodge’s estimate that the fire was still 150 to 200 yards away from the crew when he ordered his men to reverse course. Using fuel models and the fact that the fire had crowned, Rothermel estimates the fire was traveling at 120 feet per minute toward point 6. If we assume the fire was traveling 120 feet per minute and was still 150 to 200 yards behind the crew when it passed point 6, the fire would have reached point 6 in four to five minutes. That can be thought of as what the crew had in the way of a head start.
2. The start of the race for the crew (from point Y to point 6). The location of point Y can be only guessed at; 400 yards upgulch from point 6 is close and keeps the arithmetic fairly easy. The foreman was somewhat concerned, as he should have been, but it was open hillside and the men were moving gently on contour so they could keep an eye on the flank of the fire on the other side of the gulch. It would be fair to assume that they averaged about three miles per hour between points Y and 6, where they first saw that the fire was on both sides of the gulch. From point 6 on, time became a matter of minutes; 5:45 would be a good estimate of the time they reached the turnaround at point 6. There is hard evidence that at least some of the men were dying at 5:56, and the closer the race became the more accurately both time and distance can be estimated. In something close to eleven minutes the race was over.
3. The turnaround (from point 6 to point 7). This leg of the tragic race, with the fire closing in from behind, was the slowest. There is always time lost just in the mechanics of turning a crew around and getting it started in another direction. But the greatest loss was the loss that came in morale and organization in turning a crew around and retreating from the fire. The training schedule of Smokejumpers includes no class on how to run from a fire as fast as possible.