Gen. Tech. Rep. PSW-10l Berkeley, CA. Pacific Southwest Forest and Range
Experiment Station, Forest Service, U.S. Department of Agriculture; 1987.
Wildland Fire Prevention: 1
Today, Intuition--Tomorrow, Management
Albert J. Simard and Linda R. Donoghue2
Abstract:
Describes, from a historical
elements in the development and analysis of fire perspective, methods used to characterize fire
prevention programs. First devised to pinpoint
prevention problems and evaluate prevention
how fires started, fire-cause categories were p r o g r am s a n d d i sc u ss e s pa s t r e se a rc h e f fo r ts to
later expanded to include persons responsible bolster these analytical and management for wildfire ignitions. efforts. Highlights research on the sociological perspectives of the wildfire These data, combined with other fire report
problem and on quantitative fire occurrence
information and summarized in tables and on
prediction and program evaluation systems. charts and maps, have been used for decades to
Focuses on current and future advances in fire
develop fire occurrence histories.
prevention management due to research in four t h i s i n f o r m a t i o n , f o r e x a m p l e , f i r e m a n a g e r s i n
B a s e d o n
critical areas: modeling fire occurrence, N o r t h A m e r i c a k n o w t h e t e m p o r a l d i s t r i b u t i o n o f
measuring prevention effectiveness, measuring wildfires by the month, week, day, or hour economic efficiency, and optimizing program mix. (Haines and others 1975; Simard and others 1979). The major problem is that temporal fire
occurrence distributions average many individual
YESTERDAY occurrence patterns, most of which are decidedly
n o t a v e r a g e . K n o w i n g w h a t h a p p e n e d y e s t e r d a y o r
last year provides little information about
Over the years, fire managers and tomorrow or next year. Likewise, the same data
researchers have developed a number of methods
to characterize fire prevention problems and disclose where fires occur, from a strategic evaluate prevention programs. Perhaps their
scale (Simard 1975) to a management scale efforts began as far back as 1905 when, on the
(Haines and others 1978) to a local scale (Meyer
first Forest Service fire report form, field 1 9 8 6 ) . T h e s e d a t a a l s o i n d i c a t e t h e t y p e s o f
personnel documented the fire cause in addition fuels in which these fires burn (Haines and
to 14 other items of information (Donoghue others 1975). 1982). Fire causes were, and continue to be, key Although spatial and fire cause
distributions are not constant over time, they
are relatively conservative (particularly compared to weather) and, unlike temporal data, can provide many insights into the near future. 1
From fire report data and consequent analyses,
Presented at the Symposium on Wildland F i r e 2000, April 27-30, 1987, South Lake Tahoe, managers have been able to characterize their California. fire prevention problems, determine the actions n e e d e d t o s o l v e t h e m , a n d a l l o c a t e r e s o u r c e s t o
2
implement their prevention programs.
Project Leader and Research Forester, respectively, North Central Forest Experiment These analytical and management efforts have Station, USDA Forest Service, East Lansing,
been bolstered by research that developed along Michigan. 187
t w o p a r a l l e l t r a c k s - - o n e d e a l i n g w i t h
suppression costs for a patrol area for any
sociological perspectives of the wildfire given day. By comparing expected suppression problem and the other with quantitative costs to actual costs, they calculated the fire-occurrence prediction and program E x p e c t e d M o n e t a r y V a l u e ( E M V ) o r t h e
evaluation systems. A major thrust of the "estimated gain" of added prevention efforts. s o c i o lo g ic a l r e se a rc h o ve r th e y e ar s h a s b ee n t o There are no statistics, however, on the
develop demographic profiles of select accuracy of their method when applied to
subpopulations that differ significantly from localized areas on a daily basis. the general public. The attitudes and Attacking the problem from a different characteristics of these groups (such as
arsonists, hunters, children, and rural p e r sp e ct i ve , N i ck e y ( 19 80 ) de s cr i be d a me t ho d t o residents), their knowledge of fire, and/or
q u a n t i t a t i v e l y a n a l y z e f i r e o c c u r r e n c e u s i n g
their reasons for setting fires have been c control charts. This graphic method is used documented in numerous reports spanning four to determine whether changes in fire occurrence d e c ad e s o f re s e ar c h ( e. g. , Sh e a 1 94 0; K er r 1 9 58 ; are due to chance or to changes in external
Folkman 1963; Siegelman and Folkman 1971; factors such as weather or population. The Bertrand and Baird 1975). One of the underlying method is founded on the Poisson distribution objectives of this research was to supply w h i ch pr o vi de s a m od e l of t he ex p ec te d nu m be r o f m a n ag e rs wi th u se f ul in fo r m at i on ab ou t hi g h- r is k wildfires within a given week. According to
groups likely to start wildfires through
Nickey, control charts provide a way to carelessness, ignorance, or arson, thus q u a nt i ta t iv el y ev a lu a te i n c re a si n g or d ec r ea s in g providing a basis for developing effective fire fire occurrence patterns over time, to forecast prevention programs. expected future fire occurrence under stable e x t er n al co nd i t io n s, to d e t er m in e t he e ff e ct s o f unusual weather conditions, and to evaluate the To a lesser extent, researchers also studied performance of fire prevention programs. the attributes of fire prevention message sources as well as the characteristics and uses To "provide a definitional and conceptual of prevention messages and channels of dissemination.
framework for putting wildfire prevention Informal sources of fire
prevention messages such as personal contactors management on a badly-needed logical
and local opinion leaders in the rural South foundation," J. M. Heineke and S. Weissenberger were characterized by several researchers (e.g.,
(1974) developed a stochastic model of Dickerson and Bertrand 1969; Doolittle and human-caused ignitions that, together with their
others 1975; Doolittle 1979). The mass media, model for fire damages and decision costs, could
including television, radio, newspapers, signs, be used to determine prevention decisions that
and billboards, were also recognized as primary minimize the expected value of fire prevention
sources of prevention information. Scientists costs plus fire losses. studied how and when these media were used,
As Wetherill (1982) states so pointedly,
their content, their ability to change however, despite all these research efforts and attitudes, and their impact on message retention
(e.g., Griessman and Bertrand 1967; Bernardi "the easily recognized benefits of program 1970; Doolittle and others 1976; Folkman 1973). evaluation. . .logical, documented evaluation of
By examining the sociological perspectives of the fire prevention programs of forestry
the wildfire problem, these scientists have
agencies is seldom done. . .Prevention personnel
d e v e l o p e d a f u n d o f k n o w l e d g e c r i t i c a l t o
a r e a w a r e o f t h e l a c k o f e v a l u a t i o n m e t h o d s , b u t
designing effective fire prevention programs. are unsure how to go about evaluating a program without the sole reliance on fire occurrence statistics." Furthermore, he notes, "Fire Other scientists have tried to develop fire
occurrence forecasting and program evaluation occurrence alone is an inadequate indicator of
systems that go beyond descriptive statistics prevention program value even though it is the
and analyses. Nickey and Chapman (1979), for most commonly used indicator. The vast e x a m p l e , d e v e l o p e d a p r o b a b i l i t y m o d e l t o
j u d g m e n t a l g a p b e t w e e n p r e v e n t i o n a c t i v i t i e s a n d
evaluate the Red Flag Alert Program applied
the benefits of those activities cannot be during periods of Santa Ana winds in Southern bridged by intuition alone. Why must forest fire
California. From estimates of the probability prevention be unscientific when the rest of
of human-caused fire occurrence and the o u r forestry practices are governed by conditional probability of fire size at s c i e n t i f i c p r i n c i p l e s ? E v a l u a t i o n i s t h e k e y t o
suppression, they determined the expected unlocking this understanding." It is this alone 188
that will take us from an era characterized by
intuitive judgments and ad hoc decisionmaking i n t o on e o f s o u nd ma n ag em e n t s ho r ed b y a s tr o ng foundation built on scientific principles and practices. TODAY T h e f o un d at io n of pr e ve nt i o n m an a ge me n t w i ll be laid on four sequentially related
cornerstones necessary for fire prevention program evaluation: - modeling fire occurrence - measuring prevention effectiveness
- measuring economic efficiency
- o p t i m i z i n g p r o g r a m m i x .
In the 1980's, research has begun to tackle each Figure 1--Probability of a fire day (per million
of the above and early results are just hectares) vs. the NFDRS IC-G for grouped data.
beginning to emerge. Horizontal bars represent the mid-point and 95
percent confidence limits for each group (Haines
and others 1983). Modeling Fire Occurrence
The primary cause of the great temporal variability in fire occurrence is weather. It follows, therefore, that we must understand
weather effects on fire occurrence--from daily
prediction to annual normalization. Weather Influences
D a i l y f i r e o c c u r r e n c e p r e d i c t i o n a n d
seasonal normalization can be tied to most fire-danger rating systems. For example, Lynham and Martell (1985) explained 47 percent of the
variability of annual resident-caused fires in
five regions of Ontario by using a daily
prediction model (based on the Canadian Fine Fuel Moisture Code) and integrating the daily predictions over a season. Haines and others (1983) developed a similar model, using the
Figure 2--Percent of fire occurrence variability
Model G Ignition Component (IC-G) of the U.S. explained vs. length of integration period. National Fire-Danger Rating System (NFDRS) (fig. 1). When integrated over a year, using seven Northeastern weather stations (each representing an area of 6,170 square
p r e d i c t i o n . S p e c i f i c a l l y , o n l y 1 0 p e r c e n t o f
kilometers), this model similarly explained 46
daily fire occurrence variability in the
percent of the average annual variation in fire Northeastern U.S. is explained by IC-G. occurrence. Although nonweather-related stochastic elements presumably play an increasingly important role
As the length of the integration period at shorter time intervals (and for smaller decreases, however, the amount of variability areas), weather effects are also more complex.
explained by simple predictors decreases
For example, failure to incorporate the bimodal markedly (fig. 2). In other words, the shorter
eastern fire season significantly impacts daily the integration period, the more difficult the fire-occurrence prediction. To overcome this 189
d e f ic i en c y, L y n ha m a n d Ma r t el l ( 1 98 5) s tr a ti f ie d cells (to coincide with standard base maps). the fire season into five sub-seasons. More
The system employs "Bayesian analysis" of the r e c en t ly , M ar t e ll ad d ed t r i go n om e tr ic r eg r es s io n historical data base.
to a logistic occurrence model to accomplish that observed fire occurrence departs from the
this purpose.
3
Essentially, for each day historical trend, occurrence probabilities are
As part of a separate study,
gradually shifted upwards or downwards w e normalized monthly occurrence per 100 units
of IC-E (fig. 3) and adjusted the daily accordingly, thus giving greater weight to the
o c c ur r en c e pr e d ic t io n m od e l o f H a in es a nd ot h er s most recent data. Such an analysis could be run (1983) (fig. 2).
annually or even monthly or weekly to monitor This simple improvement short-term changes in human-caused fire doubled the explained daily variability to 2 1 percent. A more fundamental approach to fire o c c ur r en c e pa t t er n s. St at i s ti c al an al y s is of th e
seasons would be to model the controlling plant original model indicates that it is a good predictor of annual fire occurrence over an
phenology process directly. 1 1 , 0 0 0 s qu a re k il o me t er d i s tr i ct (C un n i ng h am an d
Martell 1973); no data are available on daily Kourtz (1984) employs a stochastic fire accuracy within a 600 square kilometer cell. occurrence model developed by Cunningham and Martel (1973f).
Five classes of the Canadian
Fine Fuel Moisture Code, two seasons, and a
When integrated over a year, most fire historical data base are used to predict
occurrence predictors are highly expected daily fire occurrence (Ward 1985).
intercorrelated. High annual average values for Occurrence probabilities (in thousandths of
a long-term component are normally associated f i r es ) a r e ca l c ul a te d f or 2 0 x 3 0 k il o m et e r g ri d with high average short-term component values.
This is much less so for daily prediction, however. For example, adding a long-term
(Palmer Z-Index) threshold to an NFDRS IC-O
threshold provides a background signal that
y i e ld s n o ta bl e im p ro v em en t in th e d is c r im i na t io n power of a daily Extreme Fire Potential Index (fig. 4) (Simard and others 1987a). An ideal Figure 3--Monthly fire occurrence probability per 100 units of NFDRS IC-E. Figure 4--Percent of U.S. extreme fires identified (composite score) vs. Extreme Fire 3
Potential Index (EFPI); percent of Northeastern Unpublished data on file, University of
Toronto; Forestry Faculty; Toronto, Ontario, U.S. days below EFPI threshold vs. EFPI (Simard Canada. and others 1987a). 190
weather-driven daily fire occurrence model would
I n e s s e n c e , a s l a t i t u d e i n c r e a s e d f r o m s o u t h t o
incorporate short-, medium-, and long-term north, fire occurrence decreased. Latitude components as well as phenological and other integrates several factors that can influence seasonal adjustments. fire occurrence such as length of fire season,
fuel types, and cultural attitudes towards fire. For example, because Southern States have There is also a class of global weather variables that show some promise for seasonal longer fire seasons, there are more fire occurrence prediction in some regions. For opportunities for fires to occur. As a e x a m p l e , b y J a n u a r y 1 , g l o b a l m e t e o r o l o g i s t s a r e
corollary, the effect of a unit of prevention able to predict the occurrence and strength of
effort is diluted over a longer period, and
an El Niño during the coming year with some
should, therefore, have less impact on total skill. Adding the January 1 state of the
fire occurrence. To test the former hypothesis, Quasi-Biennial Oscillation of the upper we normalized fire seasons for 27 Eastern stratospheric zonal winds permits us to predict N a t i o n a l F o r e s t s b y d i v i d i n g f i r e a c t i v i t y 4
h a l f o f t h e v a r i a b i l i t y o f f i r e a c t i v i t y f o r t h e
f o r each week by the highest weekly activity. coming fire season in six Southern States We then defined the seasons as 10, 15, 20, and
(Simard and others 1987b). 25 percent of maximum activity but found that the results are independent of how the seasons
Examined individually, these daily and are defined. Our investigation showed that seasonal fire occurrence prediction models are
latitude alone explains half of the variability disjointed and fragmentary. They tend to be
in length of fire season in the Eastern U.S. related to specific areas and are (f i g . 6 ) .5 Fu r t he r i n ves t i ga t io n s s h o ul d data-dependent. When seen as a whole, however,
y i e ld additional measurable processes that can
a pattern of increasing knowledge can be
be used to more directly link latitude and fire discerned. We are beginning to understand occurrence. individual components of the weather-fire occurrence process and to aggregate individual
results.
Generally applicable weather-related
daily fire occurrence prediction and seasonal weather normalization will surely be a reality
before the year 2000. Other Influences
I n e x a m i n i n g r e l a t i o n s h i p s b e t w e e n
h u m a n - caused fires in the Eastern U.S. and
nonprevention influences, Donoghue and Main
( 1 9 8 5 ) f ou n d a st r on g l at i t ud e e f fe ct ( fi g . 5 ). Figure 6--Number of weeks with fire activity greater than 10 percent of peak weekly activity vs. latitude. 4
Each Class A fire received 1 point,
Class B=2, C=4, D=8, E=16, F=32 points. 5
Figure 5--Number of human-caused wildfires in the Eastern U.S. vs. latitude (Donoghue and Main
1985). Unpublished data on file, North Central
Forest Experiment Station, East Lansing,
Michigan. 191
Donoghue and Main (1985) also found a weak fire occurrence for weather differences, they (R2 =0.08) but statistically significant
measured two attributes of prevention (P<0.001) parabolic relation between average effectiveness--total impact and implementation
state nonmetropolitan population density and rate. In 1969, the State of Washington mandated human-caused wildfires in the Eastern U.S. installation of spark arrestors on locomotives
( f i g . 7). At one end of the scale, the relative by April 1970. Results (fig. 8) indicated that
fraction of wildland has decreased sufficiently w i t hi n 2 ye ar s , e x ha u st f i r es we r e re d u ce d b y 9 5 to reduce the risk that an ignition source will percent. The State also required installation start a wildfire. At the other end, there are of improved braking systems in the early fewer people (hence, potential ignition 1970's. The gradual decline in brake shoe fires sources)--again resulting in fewer wildland
( fig. 8) was attributed to a gradual replacement fires. Lynham and Martell (1985) found a
program. Overall, this modification reduced
similar relation for resident-caused fires in brake shoe fires by about 85 percent. Even with Ontario. We suspect that other demographic engineered improvements, however, fires cannot
variables or attributes of the local economy, be completely eliminated. Malfunctions and s u c h a s m e d i a n i n c o m e , u n e m p l o y m e n t r a t e , o r
maintenance (or lack of) can notably affect
types of industry, might also have merit in
system efficacy. Pottharst and Mar (1981) found quantifying resident-caused wildland fires. a s imilar overall effectiveness and decline rate for Oregon brake shoe fires, but exhaust fires
A s w i t h t h e w e a t h e r c o m p o n e n t o f f i r e
were reduced much more gradually than those in
occurrence, these are early, incomplete Washington. This was attributed to Oregon's
results. The methods show promise, however,
persuasive strategy vs. Washington's legal t h a t i n the future we will be able to compensate requirement.
for many of the nonprevention influences on fire cause-and-effect are directly linked and
occurrence. Eliminating such background "noise" adequate data with minimal "noise" are gathered,
It is clear that, when
will permit much more accurate measurements of
a classic intervention study can reliably prevention program effectiveness than are measure prevention effectiveness. currently possible. Outside the engineering field, cause-and-effect relationships are much more Measuring Prevention Effectiveness
tenuous; many are unknown. The noise level in the data increases markedly.
In such cases, a
cross-sectional analysis is often more P o t t h a r s t a n d M a r ( 1 9 8 1 ) s t u d i e d t h e
e f f e c t i v e n e s s o f e n g i n e e r i n g i m p r o v e m e n t s i n
productive. Donoghue and Main (1985) used such
reducing railroad fires in the Pacific a n a p p r o a ch t o ex a mi n e th e ef f ec t iv en e s s o f l aw Northwest. After normalizing annual railroad enforcement in preventing arson fires. They Figure 7--Number of human-caused wildfires in Figure 8--Number of railroad fires in Washington
the Eastern U.S. vs. nonmetropolitan population by year (data from Pottharst and Mar 1981). density (Donoghue and Main 1985). 192
collected data on the number of fire-related Ultimately, the cost of preventing fires must p r o se c ut i on s, c on v ic t io ns , an d s e tt le m e nt s i n 2 7 be compared with the savings of the fires that
Eastern States during 1972-1981. They partially did not occur. Recently, Donoghue and others. normalized fire occurrence based on latitude, (1987) adapted a four-quadrant fire economic monthly precipitation departures from normal, model (Simard 1976) to fire prevention. They and nonmetropolitan population density. used a case study employing enforcement and
arson fire data from Arkansas to demonstrate how the model could be applied. The model They found that, although enforcement was n o t s i gn i fi ca n t ly re l at ed t o a ll fi re s , i t w a s incorporates four functions (fig. 10). Quadrant s i g ni f ic a nt ly r el a te d t o a r so n f i re s ( f ig . 9 ) . I contains a loss function, defined by the A l t ho u gh th e r e la t io n i s w e ak (R 2 = 0. 0 4 ), th e relation between the number of fires and
b r e ad t h o f th e da t a b as e, g en e ra l c on s i st e nc y o f suppression cost plus net value change r e s ul t s w it hi n in d iv i du al S ta t es , a nd (CS + NVC). Quadrant II contains the previously c o n fo r ma n ce w i t h w ha t w ou l d b e e x pe ct e d l e nd d e s c r i b e d e n f o r c e m e n t p r o d u c t i o n f u n c t i o n o r
c r e di b il i ty . A t l o w e nf or c e me n t l ev el s , s m al l the relation between units of enforcement and c h a ng e s y ie ld l ar g e d ec re a s es in th e n u mb e r o f number of arson wildfires. Quadrant III a r s on fi r es . A t h i gh e r le v e ls , m a rg in a l contains the enforcement cost function or the p r o du c ti v it y i s n o ta b ly l e s s. Go i ng f r o m n o l aw r e l a t i o n b e t w e e n c o s t s a n d u n i t s o f
e n f or c em e nt t o a h ig h l ev e l o f e n fo rc e m en t enforcement. Quadrant IV contains a cost
r e d uc e s a rs on f ir e o c cu rr e n ce by ab ou t ha l f transform--in other words, a line that simply ( c o mp a re d t o a n a v er a ge o f 90 pe r ce nt f or equates the two cost axes. Graphically, the
e n g in e er i ng i m p ro v em e nt s a n d r ai l ro ad f ir e s) . nomogram facilitates transforming the A l t ho u gh on ly a n e xp l or at o r y s tu d y, t h e se e n f o r c e m e n t c o s t f u n c t i o n i n Q u a d r a n t I I I t o
r e s ul t s i nd ic a t e t ha t i t i s p o ss i bl e t o Quadrant I. Summing the two functions in
q u a nt i ta t iv el y li n k a s oc i o lo g ic a l pr e v en t io n Quadrant I yields the traditional least cost a c t iv i ty to w i l dl a nd fi re o cc u rr e nc e, e ve n plus loss presentation (fig. 10). t h o ug h m a ny o f th e c a us at i v e p at h wa ys b et w ee n Lacking detailed information, an average t h e m a re il l- d e fi n ed . C S + N V C p e r f i r e w a s u s e d i n Q u a d r a n t I . A n
average cost per unit of enforcement was
similarly used in Quadrant III. The enforcement M e a s u r i n g E c o n o m i c E f f i c i e n c y
production function for the Eastern States was S i m p l y c o u n t i n g n u m b e r s o f f i r e s p r e s e n t s a n
incomplete evaluation of prevention programs. Figure 9--Number of arson fires per State vs.
level of law enforcement (sum of prosecutions +
convictions + settlements) in the Eastern U.S.
Figure 10--Four-quadrant economic model of fire
(Donoghue and Main 1985).
prevention (Donoghue and others 1987).
193
mathematically calibrated fob Arkansas and for
quickly becomes cumbersome with a four-quadrant
two regions within the State6
model.) We started with the production function
(Donoghue and Main 1985):
Individual solutions were calculated for
these two regions (fig. 11). As expected, the
efficient solution involves substantially more
A = AO (E)
enforcement expenditures in higher value
-P
E > 1
(1)
loblolly-shortleaf pine than in lower value
oak-hickory areas. Sensitivity analyses of
enforcement costs disclosed that substantial
where:
A = no. of arson fires
expenditures were justified in the
l o b l o l l y -shortleaf type across a wide cost
AO = no. of arson fires with no
range. In contrast, increasing enforcement
enforcement
costs in the oak-hickory type resulted in
notable reductions in the efficient level of
E = units of enforcement (prosecutions
expenditures.
+ convictions + settlements)
We extended the model of Donoghue and others
p = productivity coefficient (-0.134)
(1987) to derive a mathematical solution to the
problem.
(Even a simple sensitivity analysis
We also used the total cost function from
Quadrant I:
C = Cf (A) + Ce (E) + Cp
where:
C
(2)
= total cost
Cf = suppression cost per fire + net
value change per fire
Ce = enforcement cost per unit
Cp = presuppression cost
Equation (2) assumes that the costs per fire and
per unit of enforcement are constant. Although
not strictly true, conceptually we can only
prevent the average fire, and economies of scale
are not a major factor in enforcement. These
considerably lessen the significance of this
assumption and simplify the mathematics.
Although we bypass the mathematical development
here, it is straightforward.
Substitute (1)
into (2), take the first derivative of C, set it
to 0 (the point of minimum total cost), and
solve for E:
Figure 11--Optimum law enforcement levels in
 1 


 A ο pCf   p + 1 
E=

 Ce 
loblolly-shortleaf and oak-hickory forests in
Arkansas (Donoghue and others 1987).
6
(3)
Eastern States values of AO and a more detailed description of methods for calculating Note that one need not know the
within-state values can be found in an presuppression cost (Cp) to find the efficient
unpublished report on file, North Central solution.
Experiment Station, East Lansing, Michigan. 194
next decade or two, a set of productivity curves
A value of AO can be calculated for each
State or management area to be analyzed can be developed to cover most aspects of
(D o n o gh u e an d ot h er s 19 8 7 )6 . T h e Ea s t er n wildland fire prevention.
v a l u e of p is fixed at -0.134 (unless examining the sensitivity of a solution to the error Optimizing Program Mix
inherent in the production function). Assuming that the average cost per fire (Cf) and cost per unit of enforcement (Ce) are known, equation A key part of this activity lies in using
( 3 ) y ields the economically efficient number of Operations Research (OR) techniques to solve
enforcement units for the management area being
prevention problems. The past two decades are
analyzed. Equation (1) yields the expected rich with OR analyses of wildland fire
number of arson fires, equation (2) yields total
management problems. Martell (1982) reviewed
system cost, and each part of equation (2) more than 200 articles on the topic. The list
yields the cost of each component of the system.
of analyses is long: prevention planning, fuel
management, strategic and tactical detection
By rearranging terms into two ratios
planning, resource acquisition and strategic
( A/E and Cf/Ce), we can present the locus of all
deployment, resource mobilization, initial
economically efficient solutions on a single
attack dispatching, extended attack management,
two-dimensional graph (fig. 11).
fire impact management, and training. Martell
In the
process, the number of fires without enforcement
(1982) concludes, however, that "although many
(AO) conveniently drops out of the solution.
of the studies. . .have produced valuable
In addition, by using the 2-standard deviation
insights into complex fire problems, fire
limits of the productivity coefficient, we also
managers seldom rely on the results of OR
show the 95 percent confidence band for the
studies to guide their decisionmaking." He
result. Put another way, this region delineates
lists several possible causes for this
the relative fuzziness of the answer (presumably
applications gap, one of which is particularly
computer-calculated to a precision of four
germane here--"efforts to implement OR continue
decimal places).
t o b e hampered by difficulties in predicting the
physical and economic consequences of
alternative courses of action." In essence, OR
These results provide a glimpse into the
future, when economic analyses of fire
provides powerful tools for optimizing mixes of
prevention (and by extension, of fire
things when their relative costs, benefits, and
management) will be the norm. Armed with a
substitutabilities are known.
calculator and equations 1, 2, and 3 or with
In the business of preventing wildfires, we
f i g ur e 1 1 a nd a r u le r , on e ca n g e ne ra t e s e ts of
efficient solutions so that the sensitivity of
can estimate relative costs (surprisingly
the results to what is and isn't known can be
crudely in most cases), we know something about
determined. Such information can bolster
the benefits for just one or two cases, and
confidence in implementation when large input
virtually nothing about substitutabilities. For
changes have relatively little impact on the
example, we compared law enforcement exclusively
output. Conversely, it can quantify the
to arson fire occurrence, yet there is a weaker
benefit to be derived from more precisely
(statistically marginal) relation to debris
measuring the value of one or more inputs.
burning fires. Total law enforcement
effectiveness will somehow have to be prorated
between these two and possibly other causes.
The power embodied in this process lies in
eliminating intervening calculations. Once a
Similarly, rural population density explained
production function (equation 1) is defined
21 percent of the variability in debris fires
experimentally and costs are measured, all else
and 8 percent of the variability in arson fires
is mathematically derived. This permits us to
(Donoghue and Main 1985)--again, necessitating
g o directly to the desired solution and analyze
some form of proration. There may also be an as
it, rather than expending considerable energy
yet unexplored relation to other causes. At one
crunching numbers just to obtain a single
end of the substitution scale, railroad fires
solution. Today, the technique is only
and related prevention activities may be
available for one fire cause and one prevention
sufficiently independent of other causes that
activity in one part of the country. We
they can be treated as such. At the other end
suspect that a modicum of research could
of the spectrum, measuring and distributing the
develop a production function for engineering
effectiveness of multi-media Smokey Bear
and railroad fires. We hope that during the
campaigns among the various fire causes would
195
seem to be a truly challenging problem. Once we
Demonstrating the feasibility of such analyses
understand the technical relationships between
also points out the need for much better
things that we do and things that we want to
information on prevention costs and net value
accomplish, there are powerful analytical tools
change than are currently available. We expect
available to help us do more of the latter and
that by the year 2000, many organizations will
less of the former.
have made considerable progress in collecting
and analyzing such information, and that by
the year 2010, economic analyses of prevention
programs will be as common as such analyses of
TOMORROW
fire management are today.
What does the future hold for wildland fire
prevention management? Before the year 2000, we
Powerful analytical techniques for program
expect that daily fire occurrence prediction
optimization are currently available, but the
will be as commonplace as fire-danger rating is
interrelationships and substitutabilities
today. Predictions will be quantitative (e.g.,
between prevention activities and programs are
number of fires, probability of large fires),
unknown and will likely remain so for the rest
and they should account for half of the observed
of this century. Although we can solve the
daily fire occurrence variability in management
problem, current solutions tend to be of
areas on the order of a few thousand square
academic interest rather than operationally
kilometers. Within the same time frame, it
useful due, in part, to questionable inputs.
should also be possible to explain two-thirds
As economic evaluations are completed for an
to three-quarters of the interannual variation
increasing number of prevention programs, the
in fire occurrence. The data are at hand,
usefulness of program optimization will
efficient computer processing techniques for
increase. Significant progress in this final
large data bases are available, and we have a
area is not likely until well into the first
reasonable (though incomplete) knowledge of
quarter of the next century.
what's going on. What remains to be done is a
decade of research in which the individual bits
Quantitative prevention management could be
and pieces of what we've learned are put
a reality by the year 2000. All that would be
together in one generally applicable package.
needed is a significant infusion of resources.
Given the relatively low profile of prevention
in current research planning, however, we see
Progress on normalizing the weather
component of fire occurrence will enable us to
some (but not substantial) progress in the next
remove much of the "noise" from prevention
decade or so. From a "half-full glass
effectiveness data. This will, in turn, open up
perspective," even some progress will make
new opportunities to measure the effectiveness
quantitative prevention management a reality by
of more complex, diffused, and
the end of the first quarter of the next
difficult-to-quantify prevention activities. We
century. That's about half the time that it has
also expect considerable progress in normalizing
taken us to get this far. On the other hand,
other nonprevention components of fire
who knows? The concept might just take hold
occurrence such as sociological, economic, and
with a few innovative managers. These people
demographic effects. By the year 2000, or
might then decide to drive the system rather
slightly beyond, our knowledge in this field
than letting it drive them. When that starts to
could equal our knowledge of fire-danger rating
happen, great things are often achieved. In our
today. We emphasize that it is not necessary to
opinion, prevention management is inevitable.
understand exactly how or why a prevention
We can make it happen sooner or let it happen
activity leads to fewer fires to measure program
later--the choice is up to us.
effectiveness. This is fortunate because the
historical trend of progress in understanding
the how and why of wildland fire prevention
REFERENCES
leads to the conclusion that we will be well
beyond even the futuring horizon of this
Bernardi, Gene C. 1970. Three fire prevention
conference before substantial progress is made
television films varying in "threat"
in this area.
c o n te n t ... their effectiveness in changing
attitudes. Res. Paper PSW)63. Berkeley, CA:
As happened for enforcement and arson fires,
Pacific Southwest Forest and Range Exp.
economic evaluation will quickly follow the
Sta., Forest Service, U.S. Department of
measurement of prevention effectiveness.
Agriculture; 26 p.
196
Folkman, William S. 1973. Roadside fire
Bertrand, Alvin L. ; Baird, Andrew W. 1975.
Incendiarism in southern forests: a decade
prevention signs ... a restudy of their
of sociological research. Bulletin 838.
effectiveness. Res. Note PSW-282. Berkeley,
Mississippi State, MS: Mississippi
CA: Pacific Southwest Forest and Range Exp.
Agriculture and Forest Exp. Sta.,
Sta., Forest Service, U.S. Department of
Mississippi State University; 40 p.
Agriculture; 6 p.
Griessman, B. Eugene; Bertrand, Alvin L. 1967.
Cunningham, A. A.; Martell, D. L. 1973. A
stochastic model for the occurrence of
Factors related to communication of forest
m a n - caused forest fires. Canadian Journal of
fire prevention messages: a study in
selected rural communities. Bulletin No.
Forest Research 3(2): 282-287.
6 2 3 . Baton Rouge, LA: Louisiana Agricultural
Dickerson, Ben E. ; Bertrand, Alvin L. 1969.
Exp. Sta., Louisiana State University; 31 p.
Potential roles of local opinion leaders in
Haines, Donald A.; Johnson, Von J.; Main,
the communication of forest fire prevention
messages. Bulletin No. 639. Baton Rouge, LA:
William A. 1975. Wildfire & atlas of the
Louisiana Agricultural Exp. Sta., Louisiana
Northeastern and North Central States. Gen.
Tech. Report NC-16. St. Paul, MN: North
State University; 20 p.
Central Forest Exp. Sta., Forest Service,
Donoghue, Linda R. 1982. Classifying wildfire
U.S. Department of Agriculture; 25 p.
causes in the USDA Forest Service: problems
Haines, Donald A.; Main, William A.; Frost, John
and alternatives. Res. Note NC-280. St.
Paul, MN: North Central Forest Exp. Sta.,
S.; Simard, Albert J.
Forest Service, U.S. Department of
rating and wildfire occurrence in the
1983. Fire-danger
Agriculture; 5 p.
Northeastern United States. Forest Science
29(4): 679-696.
Donoghue, Linda R. ; Main, W. A. 1985. Some
Haines, Donald A.; Main, William A.; McNamara,
factors influencing wildfire occurrence and
measurement of fire prevention
Eugene F. 1978. Forest fires in
effectiveness. Journal of Environmental
Pennsylvania. Res. Paper NC-158. St. Paul,
Management 20(1): 87)96.
MN: North Central Forest Exp. Sta., Forest
Service, U.S. Department of Agriculture;
Donoghue, Linda R.; Simard, Albert J. ; Main,
20 p.
William A. 1987. Determining the economic
Heineke, J. M.; Weissenberger, Stein. 1974. An
relationship between law enforcement
activities and arson wildfires: a
analysis of wildfire prevention. IEEE
feasibility study in Arkansas. Unpublished
Transactions on Systems, Man, and
Cybernetics SMC-4(6): 513-522.
manuscript in review.
Kerr, Ed. 1958. Southerners who set the woods on
Doolittle, M. L. 1979. A scale for rating
fire. Harper's Magazine 217 (1298): 28-33.
fire-prevention contactors. Res. Note
Kourtz, Peter. 1984. Decision)making for
SO-249. New Orleans, LA: Southern Forest
centralized forest fire management. Forestry
Exp. Sta., Forest Service, U.S. Department
Chronicle 60(6): 320-327.
of Agriculture; 2 p.
Lynham, T. J.; Martell, D. L. 1985. An analysis
Doolittle, M. L.; Eller, Francis; Jackson,
Robert S. 1976. Strategies for reducing
of annual people-caused forest fire
incendiary fire occurrence in the south.
occurrence in Ontario. In: Proceedings of
Southeastern Area S. & P.F.-3. Athens, GA:
the Eighth Conference on Fire and Forest
Southeastern Area State and Private
Meteorology; 1985 April 29-May 2; Detroit,
F o r es t ry , F or e s t S er v ic e, U .S . D e pa rt m e nt o f
MI. Bethesda, MD: Soc. of American
Agriculture; 9 p.
Foresters; 200-207.
Martell, D. L. 1982. A review of operational
Doolittle, M. L. ; Kootsher, M. H.; Mercer, C. V.
1975. Relationship of personnel
research studies in forest fire management.
characteristics to fire prevention
Canadian Journal of Forest Research 12(2):
effectiveness. Res. Note SO-191. New
119-140
Orleans, LA: Southern Forest Exp. Sta.,
Meyer, Earl. 1986. Evaluating arson-caused
Forest Service, U.S. Department of
forest fires in Wisconsin, 1982-85. Fire
Agriculture; 3 p.
Management Notes 47(3): 22-25.
Folkman, William S. 1963. Levels and sources of
Nickey, Bradley B. 1980. Fire occurrence
forest fire prevention knowledge of
analysis by c control charts. Unpublished
California hunters. Res. Paper PSW-11.
draft supplied by author.
Berkeley, CA: Pacific Southwest Forest and
Nickey, Bradley B.; Chapman, Clifford B. 1979.
Evaluating fire prevention effectiveness
Range Exp. Sta., Forest Service, U.S.
Department of Agriculture; 22 p.
197
1987a. Predicting extreme fire potential.
through a probability model. Fire Technology
In: Proceedings of the Ninth Conference on
15(4): 291-306,.
Fire and Forest Meteorology; 1987 April
Pottharst, Ed; Mar, Brian W. 1981. Wildfire
21-24; San Diego, CA. Boston, MA: American
prevention engineering systems. Canadian
Meterological Society (in press).
Journal of Forest Research 11(2): 324-333.
Simard, A. J.; Graham, J. D.; Young, G. A.;
Shea, John P. 1940. Getting at the roots of
man-caused forest fires: a case study of a
Redmond, R. D. 1979. Canadian forest fire
southern forest area. General Report. Forest
statistics: 1961)1966. Information Report
Service, U.S.
FF-X-70. Ottawa, Ontario: Forest Fire
Department of Agriculture;
Research Institute, Canadian Forestry
78 p.
Service, Environment Canada; 42 p.
Siegelman, Ellen Y.; Folkman, William S. 1971.
Simard, A. J.; Haines, D. A.; Main, William A.;
Youthful fire-setters ... an exploratory study
in personality and background. Res. Note
1987b. Global teleconnections and wildland
PSW-230. Berkeley, CA: Pacific Southwest
fire activity in the United States.
Unpublished manuscript in review.
Forest and Range Exp. Sta., Forest Service,
Ward, P. C. 1985. Structure and function of four
U.S. Department of Agriculture; 6 p.
computer programs for decision support. AFMC
Simard, A. 1975. Wildland fire occurrence in
Pub. No. 201. Sault Ste. Marie, Ontario:
Canada. Map. Ottawa, Ontario: Canadian
Ontario Ministry of Natural Resources,
Forestry Service, Environment Canada.
Canada; 61 p.
Simard, A. J. 1976. Wildland fire management:
Wetherill, G. Richard. 1982. Forest fire
the economics of policy alternatives.
Forestry Tech. Report 15. Ottawa, Ontario:
prevention programs and their evaluation.
Forest Fire Research Institute, Canadian
Res. Note SO)282. New Orleans, LA: Southern
Forest Exp. Sta., Forest Service, U.S.
Forestry Service, Environment Canada; 52 p.
Department of Agriculture; 5 p.
Simard, A. J.; Eenigenburg, J. E.; Hobrla, S. L.
198