Air Attack Against Wildfires

Q: Why did RAND conduct this study?

A: In 2009, the U.S. Forest Service hired the RAND Corporation to study the composition of a mix of air tankers, scoopers and helicopters that minimized total "social costs" caused by wildfires, which includes the destruction caused by wildfires and the cost of buying and operating aircraft.

The Forest Service has tried to determine the best composition of aircraft to replace its aging air fleet for some time. Its fleet includes leased former military air tankers that date back to the 1950s. These older aircraft have been failing, with two fatal crashes in 2002 and two accidents early in June 2012, one of which was fatal.

Q: Why did the study take so long to complete?

A: The RAND report was delayed twice. The first delay was prompted when the Forest Service and RAND agreed to develop a second analysis drawing on a Forest Service analytical tool called the Fire Program Analysis system. As a result, the RAND study developed two separate, but complementary, simulation models to evaluate the effectiveness of fleet mix options. One is simpler and allows for better evaluation of the influence of some model assumptions, while the other rests on Forest Service assumptions built into the Fire Program Analysis system.

The second delay resulted when the Forest Service proposed further extending the analysis to include a more-detailed examination of the effect of direct versus indirect firefighting approaches on the fleet composition. Ultimately, the Forest Service did not ask RAND to extend this analysis so RAND is now publishing its completed work.

Q: Why did the Forest Service direct RAND to only look at certain kinds of aircraft and not very large air tankers, like 747s or DC-10s?

A: The Forest Service explained to RAND that its own prior analysis showed that the core of the large aircraft fleet would necessarily remain large aircraft capable of carrying between 1,000-3,000 gallons of retardant, not the so-called very large air tankers. As such, they asked RAND to restrict the focus of these analyses to the optimal mix of large aircraft.

Q: Do RAND's models differentiate direct firefighting attacks from indirect firefighting attacks?

A: For aircraft, a direct attack involves dropping water or retardant directly on the burning edge of a fire, while an indirect approach involves dropping retardant some distance in front of a fire's path. Like other wild land firefighting models of which RAND is aware—including the Forest Service's own Fire Program Analysis system that RAND was asked to use—the RAND models assume that water or retardant drops from aircraft, be they in direct attack or indirect attack, support construction of fire control line.

If enough fire control line is created, a fire is said to be contained. Because water can only be used for direct attack, RAND assumes that water will be less efficient in supporting fire control line construction than retardant.

To explore how this relative inefficiency might affect the optimal mix of aircraft, researchers studied the RAND National Model under a wide range of assumptions about the average effectiveness of water versus retardant. Even when water was assumed to be, on average, just a fifth as effective as retardant at supporting fire control line construction, scoopers remain the core of the fleet.

The Forest Service's Fire Program Analysis system also makes the assumption that direct and indirect attacks are comparable in this way, but it does not allow the user to modify the presumed effectiveness of water relative to retardant.

Q: Does RAND do work for scooper manufacturers or have any other conflicts of interest that could bias its results in favor of scoopers?

A: RAND is a nonprofit, nonpartisan research organization. RAND does no business with any of the aircraft manufacturers or owners, and received no compensation from any party other than the Forest Service related to this work. As with all of RAND's work, the study's conclusions were reached independently, based on the best available science. The work was thoroughly peer-reviewed by experts outside as well as inside of RAND.

RAND adhered to Forest Service instructions not to contact manufacturers or owners of any of the aircraft under evaluation. Because the Forest Service had limited experience with scoopers, however, Forest Service experts were not able to provide reliable estimates of key scooper operational capabilities needed for the modeling effort. Using these initial and incorrect parameter values provided by the Forest Service, RAND's first analyses of the optimal fleet mix favored air tankers.

In the course of vetting those preliminary results, the Forest Service permitted RAND to contact one of the scooper manufacturers to request contact information for scooper users whom researchers could interview. After contacting Bombardier Inc., RAND consulted with more than a half-dozen of its customers across Canada and the United States. Those customers helped refine the estimates of scooper cycle times RAND used in its analysis.

Q: Why do you include the cost of fire retardant in the cost estimates for the various aircraft? Doesn't that tilt the results in favor of scoopers?

A: The study's cost estimates include not just acquisition costs, but also operations, maintenance, support and fuel costs. Retardant is a non-trivial cost associated with the operations of three of the aircraft evaluated in this study, so it is included as a cost of using these aircraft to fight fires.

Q: If the Forest Service continues to pursue an air tanker-centric, rather than a scooper-centric fleet, what would some of the possible outcomes be?

A: The RAND analyses suggest that the Forest Service can achieve the same levels of fire suppression with a tanker-centric fleet as with a scooper-centric fleet, but a tanker-centric fleet will cost substantially more.

Q: Doesn't your study exaggerate the availability of scoopable water? Your study doesn't consider the depths of the water bodies, who has rights to that water or seasonal variabilities?

A: A complete survey of water availability for scoopers was beyond the scope of the project. It may be that some bodies of water the study classed as scoopable, are in fact not scoopable. In such cases, the study has underestimated the distance scoopers must travel.

Nevertheless, even if this analytic shortcoming affects every body of water selected as most proximate to a fire, and scoopers actually have to travel twice as far, the RAND results demonstrate that scoopers remain the core of the fleet. In other words, if scoopers have to travel twice as far, they drop half as much water in a given period of time, which is equivalent, in the RAND model, to water being half as effective as retardant, a case that barely changes the mix of air tankers versus scoopers.

Q: Are there examples of scoopers being used as a primary wild land fire aerial attack platform?

A: Yes. At the time this study was conducted, scoopers were the primary aircraft used to fight wild land fires in Canada, France and Spain. They also are used widely in several other European countries, including Italy, Croatia and Russia. The United States also has experience fighting wild land fires with scoopers. The U.S. Bureau of Land Management in Alaska operates two scoopers, the largest aircraft in their fleet. Los Angeles and San Diego counties lease scoopers. Minnesota also uses scoopers.

Q: Did RAND assume that air tankers would be new? Wouldn't it be cheaper for the Forest Service to lease older aircraft?

A: RAND examined expected costs of both new and used aircraft. However, the values the study uses in the reported model runs use annualized costs for new scoopers and tankers, because of the Forest Service's emphasis to RAND that it wished to consider newer, safer aircraft built for the air tanker mission or missions similar in maneuver loads and low-level flight. Helicopter costs were assumed to be equivalent to current Forest Service lease rates.

It is true that one would expect some reduction in annualized costs if leasing older aircraft. In the study, Figure 4.1, for instance, shows that the very old air tankers the Forest Service has been using have lease costs about 80 percent of what the study expects the annualized cost of new large air tankers would be.

The RAND cost estimates are reasonable for purposes of the modeling exercise, but not sufficiently precise to support acquisition decisions. It might be that used aircraft (air tankers and scoopers) are available on the market for somewhat less than the study's annualized cost estimates. With more precise cost data on market rates and availability, the models could be re-run to provide fleet mix estimates matching any preferred acquisition approach.

Q: Why doesn't the study give a fixed number of aircraft, rather than a range, for an optimal fleet?

A: There are key uncertainties in the RAND model parameters that limit the precision of the estimates. One significant finding of the RAND study is how sensitive fleet size requirements are to the fire intelligence available to Forest Service aircraft dispatchers. The study shows, for instance, how the central importance of efficient prepositioning of aircraft to meet the next day's firefighting needs and what the researchers termed "dispatch prescience," or the accuracy with which dispatchers can send aircraft to just those fires where they will mean the difference between a contained fire and an escaped large fire.

When aircraft can be correctly and flexibly prepositioned, fewer are needed. If aircraft dispatch can be optimized, fewer aircraft would be needed. The Forest Service could reduce its aviation costs if it could increase dispatch prescience and prepositioning accuracy.

However, current levels of dispatch prescience and prepositioning accuracy have not been measured, so were estimated roughly for this analysis. Because the results are sensitive to these parameters, additional research on them is needed in order to reduce the plausible range of optimal fleet size and mix.

Q: Why does the base case estimate of RAND's National Model call for 43 scoopers while the base case estimate of RAND's Local Resources Model calls for 15 scoopers? That seems like a big difference.

A: The two RAND models have different underlying assumptions, strengths and weaknesses that lead to somewhat different model results. A key weakness of the National Model is that it treats all large fires as costing what RAND researchers estimate as the mean cost for large fires, $3.3 million. In fact, however, the study shows that a large majority of large fires actually cost less than the mean, which is high because of a small number of very expensive fires that occur each year. Since the National Model uses an "average" cost that is quite high, prevention of just a couple large fires can justify additional aircraft. Thus, the National Model overestimates the number of aircraft that would be cost effective.

In contrast, the Local Resources Model allows differential valuation of fires, e.g., larger fires and those near populated areas are more costly. The mean value of large fires is still $3.3 million, but now many large fires would cost much less, so many more would have to be prevented to justify additional aircraft. The Local Resources Model's search algorithm exploits this information and focuses its aircraft on fires likely to become expensive, namely those near populated areas.