This file was created by scanning the printed publication. Errors identified by the software have been corrected; however, some errors may remain. Where and When to Measure Forest-Fire Danger G. LloydHayes • This article presentsthe results of a study to determine the place, time, and number of measurements that shouldbe made to obtain dependableratings of "average-bad"fire conditionswithout an excessive number of stations or observations. The author concludes that under the conditions prevailing in the Priest River ExperimentalForest in northern Idaho a single measurementtaken daily at noon at either a valley-bottomor a south-slope stationis adequatefor the rating of fire danger. OREST-fire danger a is nowmeasured in in manningfor fire control'in the longleafpine everyforestregionof theUnitedStates.type. Severalsystemsof measuringand evaluatIt was early recognizedthat the sameprining dangerhavebeehdeveloped (3), all of which ciple that governs fire-suppression practices are of value in the current supervisionof fire shouldalsogovernthe placeandtime of making suppression and in fire-controlplanning. danger measurements.Fire danger is known Studies.at the Northern Rocky Mountain For- to vary greatly from place to place and from est a•d RangeExperiment.Stationhave aimed time to time at the sameplace. For example, consistently to determine(1) what factorsaffect in the samefuel type the dangeris obviously fire danger,(2) howtheymaybe measured, (3) greateron souththan on'north slopes,at lower wherethey shouldbe measured,(4) whenthey than at higherelevations, and duringthe midshouldbe measured,and (5) how to integrate afternoonthan at night. Therefore,if an areais thesemeasurements into a reliable,practical,.and "manned"to copewith the averagefire danger readily usablescale. Thesedistinctcomponents of all the aspects, elevations, and hours,the orof theproblemconstitute the"what,how,where, ganizationshouldbe able to handlewith ease and when" of fire danger. The "why" of fire- all fireson the less-dangerous-than-average north dangermeasurement is primarilyto indicatethe slope,especiallythoseoccurringat night; but it size of the controlorganizationneededto cope would be entirely inadequateto control fires with existingfire danger. starting on the more-dangerous-than-average Presentpoliciesrequirethat preparations be southslope, particularly duringthecriticalaftermade for controllingall fires that start under noon period. These important considerations "averagebad" or easierconditions.This means were not adequatelytaken into accountin the a fire-controlorganizationcapableof suppress- statisticalanalysesof fire-weatherstation dising .all fires,exceptthosethat might start in the tribution madeby Morris (11). worst ]uela under the very worst burning conIn the early stageof measuringfire dangerin ditions,and recognizes the importance of the law the NorthernRockyMountainRegion,the availof diminishingreturnsin fire-controlplanning. ability of men at an established stationand conHornby (7) accepted the factthat beyonda cer- veniencein makingthe ,measurements were the tain pointit is moreeconomical to takethe loss controllingcriteria as to where and when the than it is to bearthecostof preventing that loss. average-bad shouldbe measured.Theseare still RecentlyKnorr (8) recognized the application dominantfactorsin actualpractic&.By 1934, of the average-bad principlein selecting the loca- however,progressin otherphasesof evaluating tionsof fire-dangerstationsto be usedasa guide fire dangerhad showntheneedfor a betterbasis. •Formerly attached to the Division of Forest Protec- Therefore,very appropriatelyat that time, a tion, Northern Rocky Mountain Forest and Range Ex- study was startedat the Priest River Experiperiment Station, Missoula, Montana; now in charge of howthe torest-fi•e research at the Appalachian Forest Experi- mentalForestto find out moreprecisely various fire-dangerfactois differed in mounment Station, Ashevill½, iN/. C. a"Forest-fire danger" is used in its broad senseas defined in the Glossaryo! Terms Used in Forest Fire Control, U.S. Department of Agriculture, Forest Service, 1•39, namely,"A general term expressingthe sum total tainous terrain in order to acquire more com- plete and basicknowledgeof where and when to measurefire danger. ef both the constant and variable factors which deterThe resultsof that altitudeand aspectstudy mine whether fires will start, spread, and do damage, have been given in two previouslypublished and which determine their difficulty of control." 744 WHERE AND WHEN TO MEASURE FOREST-FIREDANGER 745 reports(5, 6). Thispaperwill show,in sofar .as Figure 1 may be likenedin somerespectsto wind and the .atmospheric factorsaffectingfuel- a contourmap savethat its "contours"are lines moisture conditions are concerned, how the of equalburningindexratherthan linesof equal average-bad fire conditionin ruggedtopography elevation.At 6 a.m., for example,there are may be evaluatedand whereand whenit can be "valleys"of low burning indexesat both low measured. and high elevations with "ridges"of high burning indexbetween. Theseculminatein the highest"peak"around2 p.m. The 'data are readily usableto approximate the burningindexat any elevationon northand Fire-weatherdatawereobtainedat sevenfully south slopesbetween2,200 and 5,500 feet and exposedstations. One of these,whichmight be called the valley-bottomcontrol becauseof its at anyhourof thedayor night. For example, similarityto a dangerstationat any rangersta- the figures'on the verticalline for 6 a.m. in tion, waslocatedin a flat valleybottomat 2,300 Part A showthat the burningindex increases feetelevation. The remainingsixwereestablishedfrom valleybottomup to 3,500feetandthendetowardthe mountaintop, as follows: in pairs locatedon true north and southslopes creases BURNING INDEXESs AS A MEASURE OF AVERAGEBAD FIRE DANGEa at elevationsof 2,700, 3,800, and 5,500 feet on Elevation, a ridge rising from east to west. All pairs werewithin 100 feet elevationof the ridge crest. At each stationwind velocityat 7• feet above thegroundandmoisture contentof duffandhalf- feet inch fuel moisture indicator sticks were mea- suredand recordedcontinuously by an anemo- Burningindex, class 2,4OO 3,000 3,500 4,000 2.6 3.2+ 3.5 3.4• 4,500 3.3+ 5,000 5,5OO 3.2 3.1 hygrograph. Similarly, the horizomalline for 3,000 feel The fuel moistureand wind velocitydata of elevation on thissameslopeshowsthattheburneach station were integratedby the Northern Rocky Mountain fire dangermeter (model 5) ingdndexratingsfor varioushoursof the day into burningindexeson a scaleof sevenclasses. were as follow.s: Theseembr.ace theentirerangeof burningcondiHour•f Burning index, day class tions, from no spreadof fire under Class1 to 10 p.m. 3.8 spreadsof 1,500 to 2,000 acresper hour when 2 a.m. 3.4 the explosive conditi6ns underClass7 prevail. Figure 1 showsthe altitudinaldistributionof the burning indexeson both north and south 6 a.m. 10 a.m. 2 p.m. 6 p.m. 3.2+ 4.4 5.2-4.6 slopesfor 24 hoursof .themedianday4 in JulyIf average-bad fire dangeris concretelyde. August,1938. The conditions shownare typical or normalfor the clear,dry fire weatherthat is fined in termscapableof numericalexpression so characteristic of northern Idaho summers. (10), theinformation in FigureI obviously can Median valuesare used,sincemean valuesor be used to determine when and where to measure arithmeticaverages aretoomuchaffected by the ".average-bad" with respectto aspect,elevation, occasional, unusualstormyweather(1, 4, 7) to and time. The chartclearlyshowsthat at idenhigherburninginbe typicalof the summer climateof thisregion tical hoursand elevations, withitshightemperatures, low relativehumidity, dexes existonthesoutherly aspect!A) thanon the northerlyaspect(B). Likewise•at all elevascantyrainfall,andlongperiods, of drought. tionsonthesoutherly aspect thedangeris great. aThe Committeeon ForestryTerminologyof ;he Soer between 10 a.m. and 6 p.m., the so-called ciety of AmericanForesters has definedburningindex as "a relative number denotingthe combinedevaluation burningperiod,than at any other hour of the of the inflammability of forest fuels and the rate of entireday. Thesefactsleadto onelogicalconspreadof fire in suchfuels for specificcombinations of themeasure of the average-bad mustbe fuel-moisturecontent,herbaceousstage,and wind ve- clusion: upontheaverage of fire-danger conditions ß 1ocity." In someformerpublications (5, 6), burning based indexes have been termed "fire-behavior classes." asrepresented by theburningindexeswhichpre•Median day is the day on which conditionswere such that half of the July-Augustdays showedworse fire danger and the other half less fire danger. vail on the southerly aspectandvalleybottom between thehoursof 10 a.m. 'and6 p.m. 746 JOURNAL OF FORESTRY NIGHT PERIOD MORNING DAY PERIOD EVENING 5400 5200 5000 48,00 48,00 4400 4200 4000 38,00 3800 3400 3200 3000 2800 2800 2400 5400 ,5200 ,5000 4800 4600 4400 4200 4'000 3coo 3800 3400 3200 3000 2800 2600 2400 I0 MDNT. 2 4 6 A.M. 8 I0 NOON 2 4 8 8 I0 P.M. Fig. L--Burning-index ratings for different altitudes, different hours of the day and night, and on south (Part A) and north (Part B) alopesfor the median day of July-August, 1938, on the Priest River Experimenta]Forest. The isograms,or "contour lines," are lines of equal burning index. WHERE AND WHEN TO MEASURE FOREST-FIRE DANGER 747 measurements at anindividual staOn the medianday of July-August, 1938,the agingseveral severalvaluesfrom different burningindexof the average-bad condition was tion or combining class 4.8. This was obtained as follows: (1) For the two monthsthe averageburningindex of thecriticalperiod(10•a.m. to 6 p.m.) was stations ? AccuRAcY OF INDIVIDUALSTATIONRATINGS with whichthe calculated separately for eachof thethreesouth- Table1 showsthe accuracy burningindexon anyJuly-August slopeand the onevalley-bottom stations.(2) average-bad Next,thepercentage of all fiats.andsouthslopes daywasapproximated by singledailyratingsat includedin the studyarea (6,000 acres)which eitherthevalley-bottom or the mountain-top stafell within the altitudinalzonebestrepresentedtion.Anyofthethreesingle measurements rated by oneof eachof thefourstations wasestimated average-bad withan accuracy sufficient for pracfroma topographic map.(3) The.average burn- ticalpurposes, but the oneat 2 p.m. on the ing indexof eachstationobtainedin step1 w.as southslopenearthe lookoutstationwasbest. then multipliedby the appropriatepercentage Whilethepractice of measuring onlyonceeach fromstep2 andthe four products wereadded together.(4) Their sum,class4.8, was the weighted-average burning index of all south slopesandvalley-bottom areasembraced by the study. Figure 1 (A) showsthat on the normalday, between10 a.m. and 6 p.m., a burningindex of class4.8 occurredtwice in the valley bottom and at all elevationson the southslopeexcept at thehighestpoint,whereit prevailedonlyonce at 2 p.m. Consequently, the average-bad condition, as defined,could be measuredat any placein thevalleybottomor onthe southslope. In the valleybottomthesemeasurements could be made either at noon or 5 p.m. Measurements.at any other hours would give indexes eitherdefinitelybelowor above"averagebad" for the area. On the southslope,similar measurementswould have to be made at 3,000 and 4,000 feet around 11 a.m. or 5 p.m., and at 5,500 feet around2 p.m. SOME PRACTICAL CONSIDERATIONS dayandthenonlyat a lookout station wouldbe readilyfeasibleon mostForestServiceranger districtsin July and August,it wouldnot serve duringearlierandlatermonths whenthelookout stationsare oftenunoccupied. RELATIVE ACCURACYOF THE THREE SETS OF DATA As shownin Table 1, the 2 p.m. burning indexesfrom the south-slope station at 5,500foot elevation were most accurate. There on 50 percentof all daystested,the ratingsdiffered from the average-had burningindex by only ___0.1 classor less;on 66 per cent of the days the difference was +--0.24 class or less. How- ever, at the valley-bottom stationthe readings were almost as reliable. There the noon data varied from the average-badrating by ___0.15 classor less on half the days,and by ___0.34 classor lesson two-thirdsof the days. Furthermorethan 5 p.m. data at this samestationdeviatedfrom .average-bad by not over ___.20 and +-0.33 classeson half and two-thirdsof the days, respectively. The largestoverestimate exceeded the averagebad ratingby 1.3 classes and the largestunder- As broughtoutabove,the average-bad canbe measured at morethan oneplace,and at different hoursof the day in someplaces. Usually menfor this purposeare readilyavailableonly estimate was 1.0 class too low. These errors but at valley-bottomranger stationsor mountain-top mightappearto be of practicalimportance, lookoutpoints. The valley-bottom and 5,500- they all occurredon dayswhenthe average-bad foot south-slope stationsusedin this studywere burningindex was low. On suchdays,when typicalof suchlocalities.Measurements of aver- fire will not give serioustrouble,errors of this aswhendanage-badmade at thesestationswere testedto amountarenotof suchsignificance in efficientfire control answer several important practical questions. ger is high and success vitally on accurateestimationof condi(1) How accuratelydo singleratingsat an indi- depends dividualstationrepresentthe average-bad from tions. day to day; (2) for the stationsin question, The maximum overestimate of 1.3 classes ocwhichof the threeratingsapproximatedaverage- curred,for example,on July 6 whenaverage-bad badmostaccurately; and (3) howmuchcanthe for the 6,000-acrearea as a whol.ewas rated at a accuracyof daily ratingsbe increased by aver- burningclassof 2.4. At noonon that day,mea- 748 JOURNAL OF FORESTRY surementsat the valley-bottomstation indicated of about one-tenthof a class. Even its largest the burning index at class 3.8. However,be- singleerror, an overestimateof 0.4 class,would causeof rain which startedat 2 p.m. the same not be of seriousimportance to a protectiveor- day, the 5 p.m. index at this stationfell to class 1.6 or 0.9 classbelowthe average-bad.Similar• ly, the diurnal trend was abnormalon the day immediatelyfollowingowingto rapid drying of unusuallywet fuels. Only on days which were affectedby recentrain did errorsexceedingñ0.5 class occur. No error exceedingthat amount occurredon any day unaffectedby rain which ganization. Two measurements in the valley bottomand one on the mountaintop eachday thereforecan be used, if justifiable on other bases,to give maximumaccuracy. DISCUSSION The aboveanalyses indicate thattheoretically the average-bad burningindexcan be bestapratedclass3.6 (moderately bad) or more. In proximatedfrom combinationsof three daily fact, on the moderatelydangerousdays the rat- measurements at two stations. From the pracings were generallyaccuratewithin ñ0.3 class ticalpointof view,however, cananyfire-control and less. organizationreally make use of ratings as refined as this? To statethe questionfrom anINCREASING THE ACCURACY OF RATINGS otherangle:Will the additionalcostof making While suchaccuracyis reasonably good,anal- morethan one daily measurement be justified ysisshowsthat averagesof two or moremeasure- by the financial benefitsderived from the most mentsfrom a single station or from separate accurate butalsothemostcostlyestimate based stationsgive greateraccuracythan any individ- on three measurements? ual measurement. Three different combinations Suchquestions mustbe answered negatively of two daily estimatesand one additionalcom- for the NorthernRockyMountainRegionwheu bination of all daily measurements were tested. consideration is givento our knowledge of fire :Asis evidentfroin Table 2, averagesof three behavior, the uniformityof dangerfrom dayto daily measurements, two in the valley bottom day, andthe degreeof refinement employed in and one on the mountaintop, approximated present-day fire-controlwork. Analysisindiaverage-bad with remarkabledependability. The catesthat on the two-thirds of the daysthat a median and standard errors of such an estimate fire-controlorganizationwouldhave been most werereducedto the inconsequentially low points vitally concerned with the fluctuatingdanger, TABLE I.--AccuRACY WITH WHICH THE AVERaGE-BaBBURNINGINBZX WAS APPRGXlMaTED BY SINGLE,DAILY MEASUREMENTS AT DIFFERENTSTATIONS,JULY-AuGuST, 1938 Station and hour Valley bottom, 2,300 feet 12:00 m............................................................... 5:00 p.m......................................................... South slope, 5,500 feet 2:00 p.m.......................................................... Median error of estimate • Standard error of estimate Greatest over. estimate Greatest underestimate Class Class Class Class 4.0.15 4-0.20 4-0.34 4-0.33 +1.3 +0.9 --0.9 --1.0 4.0.10 4-0.24 +0.6 --0.6 XThemediandifferencebetweenthe measuredand the estimatedaverage.badburning index. TABLE2.•AccuRAcY WITH.WHICHTHEAVERScs-BiB BURNINg INBEXWASAPPROXIMATED BY COMBINING THE DAILY MEASUREMENTS OF TWO ANDTHREESTATIONS, JULY-AuGusT, 1938 Station and hours Valley bottomat 12 m. and 5 p.m........................ Valley bottom at 12 m. and south slope,5,500 feet, at 2 p.m..................................................... Valley battom at 5 p.m. and south slope, 5,500 feet, at 2 p.m................................................. Valley bottom at 12 m. and 5 p.m., and south slope,5,500 feet, at 2 p.m............................... Median error of Standard error of Greatest over- Greatest under- estimate estimate estimate estimate Class Class Class Class 4-0.10 4.0.19 +0.8 --0.4 4-0.08 4.0.18 +1.0 --0.4 4-0.09 4.0.16 +0.3 ---0.7 4-0.06 •0.11 +0.4 --0.2 WHERE AND WHEN TO MEASURE FOREST-FIREDANGER 749 was rated as 4.8 at noon. As the valley bottom singlemeasurements estimated the average-bad wasactually 5.2atnoononthedayofthis fromonly-q-0.05to ñ0.18 classlessaccuratelyrating or 0.3 class,mustbe added than did two observations, and only ___0.13to fire, the difference, +__0.23 classlesspreciselythan the medianof to the 5.1 shown for the fire location. The estithree measurements.No fire control force in this matedburningindexat th6 siteof the firstfire 5.1 plus0.3 or class5.4. regionis at presentso flexiblethat it can be wastherefore variedfor'a changeof only0.23 classin fire Similarly thedangerat thesiteof thesecond danger. fire as estimated from Figure1 and corrected Moreover,thevariationsin fire behdvior with- by the noonmeasurement wasclass4.0 (class ha the differentburning-index classeshave not 5.2 of valley-bottom stationat noonminusclass yet beendetermined to a degreethat economi- 1.2, thenormaldifference between ratingsof the cally justifiesincreasingor decreasinga fire- valley-bottom stationat noonand of a spoton controlforce to meet a changeof -q-0.34class, thenorthslope at 2,600feetelevation at8 p.m.). the largeststandarderror for estimatesbased Thelocalrangertherebyis ableto judgeandto uponsingledaily measurements. Until fire con- reportprobable fire behavior at thesetwofires trol actioncanbe refinedto thepointof actually far moredependably andspecifically thanwould makingchangesin the controlforceswith varia- otherwise be possible.The actionhe takesfor tionsof one-thirdclassof fire danger,onedaily a 5.1 conditionobviouslywill be both stronger measurement is therefore recommended as adeandfasterthanfor theclass4.0, eventhoughthe quateon anyunit similarto the onestudied. latterrepresents a spotat,lowerelevation, hence to burning. Thesetestsare strictlyapplicable, of course, withmoreareaaboveit andexposed onlyto the samplingof .average-bad as herede- While these are conditionsand actionsreadily fined and computed. There is no reason,how- apparent to the experienced forestofficer, they ever, if someother index were wanted,why it are not so obviousand are often overlookedby could not be approximated by similar methods the newmanto whomall firesmaylook equally and with similar accuracy. Average-bad danger sinister. The severalpossibilities of making.dailymeaon forest areaslarger than 6,000 acrescould wereanalyzedalsoto determine which not be determinedso accuratelyperhaps;but surements estimates an accuracywithin practicallimitationsshould wouldbe the bestbasisfor dispatchers' be obtainableduring settledweatheron an area of this kind. Each combination of station meamanytimeslargerthan 6,000 .acresprovidingit surements was used in turn with values shown in Figure1 to sstimate burningindexesat three is climaticallyuniform. Burning-indexmeasurements, in addition to differentsitesand widelydifferenthours. Furtheir value for rating dangeron an area as a thermore, additional charts were constructed a whole, also can be usedby fire dispatchers to similarto Figure1 but with eachrepresenting estimateburningconditionsat any spoton their certainlimited rangeof dangerat noonat the unit where a new fire occurs. The manner in valley-bottomstation. In using Figure 1 for Which this may be done can be illustratedby this test, the estimateswere confinedto those usingthevaluesin Figure1 to showtherelation- daysthatrated4.3to 5.3 at thevalley-bottom shipbetweenthe dangeractuallymeasuredand station. Other charts should be used for lower the dangerestimated to be existing"at the fire." and higher dangers. This limited the number in eachgroupto 30. Table3 shows Let it be assumed that two fires are reported of estimates the dependability found for any one.of three on a day when the noon burning index of the different measurements when used as a basis for valley-bottomstation was measuredand found to be class 5.2, or 0.3 class above the index estimatingconditionson any July-Augustday shownby Figure 1 for sucha stationand hour. at threeotherspotswithin the area. In seven of the nine cases the estimated burn. Further, assumethe first fire as occurringat 3 p.m. on a southslopeat 3,800feetandthe sec- ing indexesdiffered from the actual by only classor less,on half of all the daystested. ond at 8 p.m. on a north slopeat 2,600 feet. ___0.2 Accordingto Figure 1 (A) the aspect-elevationIn the other cases the differences were even combination for the first fire givesa rating of smaller,from ___0.1to ___0.15classor less. This aboutclass5.1. But this chartshowsonlywhat uniformity of the median errors of estimate can be expectedif the valley-bottomcondition showsthat on at least50 per cent of the July- 750 JOURNAL OF FORESTRY TABLE 3.--AccURACY WITH WHICH THE BURNING INDEX AT THREE DIFFERENT PLACES ANn HouRs WAS ESTIMATED FROM MEASUREMENTSTAKEN AT THE VALLEY-BOTTOM STATION AT NOON AND AT 5 P.M., AND AT THE SOUTH-SLOPESTATION 5,500 FEET ELEVATIONAT 2 F.M. Situation and hour of measurement Valley bottom, noon Median error of of estimate estimate Valley bottom, 5 p.m. South, 3,800 feet, North, 2,700 feet, South, 5,500 feet, South, 3,800 feet, North, 2,700 feet, 2 p.m. 8 p.m. 10 a.m. 2 p.m. 8 p.m. South slope, 5,500 feet, 2 p.m. South, 5,500 feet, 10 a.m. South,3,800 feet, 2 p.m. North, 2,700 feet, 8 p.m. South, 5,500 feet, 10 a.m. Standard error of estimate Greatest .overestimate Greatest underestimate +0.15 ñ0.2 ___0.2 ñ0.2 ñ0.1 +0.27 ñ0.25 +0.40 ñ0.41 ñ0.22 70.7 -{-0.7 71.3 70.7 70.6 ---0.4 ---0.3 •0.7 ---0.7 -4).5 ñ0.2 ñ0.2 ñ0.2 ñ0.2 ñ0.39 ñ0.22 ñ0.40 ñ0.35 70.8 70.4 70.6 71.0 --0.9 ---0.6 --1.3 •0.9 thenthe Augustdaysconcerned, the distributionof burn- velocityand direction,or by cloudiness, of burningindex shouldand mustbe ing indexeswassosimilarto that shownby Fig- estimates ure 1 that it mattered little where or when the modified by common-sense allowancesbased measurements weremade. For example,on half upon experienced judgment. Good dispatching the daysthe burningindexesat 2 p.m. on the will alwaysdemandthe exerciseof soundjudgsouthslopeat 3,800feetelevationwereestimated ment. Thiscanbe guided,but neverreplaced,by just as accuratelyfrom measurements made 21 mechanical methodsof estimating. hoursearlierand 1,500feetbelow(valleystation For usein dispatching, burningindexes should at 5 p.m.) as from observations taken at the thereforebe measuredas near as practicalto sametime and 1,700 feet higher (south-slopebad fire zonesand preferablyjust prior to the station5,500 feet at 2 p.m.) In eachcasethe time of day whenfires usuallyoccur. In this median error was the same, ___0.2class. connection,noon measurementsshould be more The standarderrors in Table 3 show, however, valuablethan thosemadeat 2 p.m.; the latter that on dayswhenthe distributionof fire danger in turn shouldhave greatervaluethan 5 p.. m. wasnotsotypical,greateraccuracy wasobtained data. Both for useby the fire dispatcherand in when the measuredand estimatedburning in- over-allfire-danger rating,measurements should dexeswere made near each other with respect not be made during those before-noonhours to bothplaceandtime. Thus,the standarderror whenthe burningindexis changingrapidly,i.e., for estimates of 2 p.m. burningindexesat 3,800 in valleybottomsfrom 7 through11 a.m., andat feet on the south'slopewas only ___0.22class mid-elevationson south slopesfrom 8 through when based upon measurementsmade at the 10 a.m. Measurements made at such times are 5,500-footsouth-slope stationat 2 p.m. It was not so.reliablyindicativeof whatcondkionswill ___0.41 classwhen using5 p.m. measurementsbe elsewhere on the area as are measurements made after the burning index has cometo its of the valley-bottom station. Large overestimates and underestimates in ex- daily comparativeequilibrium. cessof ___0.5classoccurredoccasionallyat all SUMMARY of the stationstested(Table 3). Errors of that magnitudeare probablyless importantin the Fire-dangerratings are necessaryto sound presentinstancethan they would be in rating fire-controlplanning. Among other uses,they generalfire danger. In actualdispatching they furnisha very practicalindexof the pres.uppreswouldbe largelyavoided. No dispatcher, for sionmanpowerneededon an area. Dangermeaexample,would acceptor use withoutmodifica- surements madeat onespotand usedto rate the tion an estimateof the burningindexfollowinga fire .dangerof the immediatelysurroundingarea rain which was basedupon measurements that must be representative of the more dangerous were taken before that rain started. Yet four of portions of the area and of the more crkical the largeit overestimates in Table 3 werecaused hoursof the day, but not the most dangerous by exactly such a combinationof conditions. exposuresat the very peak of the day. This Similarly,when the typical topographic dis- average-badconditioncan be sampledand extributionof fire dangeris disturbedby uneven pressednumericallythrough burning indexes distribution of rain,by unusualchanges in wind obtainedby integratingmeasurements of fuel WHERE AND WHEN TO MEASURE FOREST.FIRE moistureand wind velocity. For the area ineludedwithin the 6,000-acrePriestRiver Experi. mental Forest, it was found that these measurementscouldbe madeeither in the valleybottom at noonor 5 p.m., and on southslopes,depending upon elevation,at 11 a.m., 2 p.m., or 5 751 areas.andnearthe beginningof the daily critical period. Fromthe dispatcher's pointof view,the noon measurement is the best of the three time periodstestedand the 2 p.m. measurement is Single daily measurements made at a valleybottomstationat noonor 5 p.m., andat a 5,500foot south-slope stationat 2 p.m., represented the average-bad burningindexesfor the area as a wholewith an accuracywell within the limits of the practicalapplicationof fire-dangerratings. A combinationof two daily measurements gave more representative burning-indexvalues than did singleobservations, while three daily measurements improvedthe .accuracy of averagebadratingsstill further. Few,if any,fire-control organizationsare, however,sufficientlyflexible to use effectivelyburning indexesmore precise than thoseobtainedfrom singledaily measurements. Nor is present-dayknowledgeof fire behaviorso completethat minutedifferences in estimatedfire dangerhavepracticalimplications. Hence, there is no economicjustificationfor financingadditionalmeasurements in order to obtainrelativelymeagerfurther refinement. For dispatchingpurposesit matterslittle on normalor averageJuly-August daysin the Northern RockyMountainRegionwhereor whenthe burningindex is measured. Each of three sets of data taken at different sites and hours of the DANGER next best. LITERATURE CITED 1. Gishorne,H.T. not a normal. 171-175. 2. 1935. When a normal is Amer. Met. $oc. Bul. 16: . 1939. Hornby'sprinciplesof fire-controlplanning.-Jour. Forestry 37: 292-296. 3. . 1939. Fordstpyrology. $ci. Monthly49(7): 21-30, illus. 4. Hayes,G. Lloyd. 1935. Don't let a "normal" deceiveyou. U. $. Forest•$erv. Bul. 19:4. 5. 1941. Influence of altitude and aspecton daily variationsin factorsof forest-fire danger. U.S. Dept.Agric.Cir. 591. 38 pp., illus. 6. . 1942. Differences in fire dan- ger with altitude,aspect,and time of day. Jour. Forestry40:318-323, illus. 7. Hornby,L. G. 1936. Fire controlplanning in the Northern Rocky Mountain region. North. RockyMountainForestand Range Expt.Sta. (Multi.). 8. Knorr, Phillip. 1942. Variationsin firedangerfactorson a rangerdistrictin the longleafpine region. Jour. Forestry40: 689-692. day provided almostequally preciseestimates of the burningindexat otherplacesandhours. 9. Mindling, GeorgeW. 1940. Do climaOn lesstypicaldayssuchas whenrain occurred, tologicalnormalsserve adequatelyas nor- however, better estimateswere obtained when mMs? Amer. Met. Soc. Bul. 21:3-6. measuredand estimatedburning indexeswere 10. Morris, William G. 1939. What is "avermadeneareachotherwith respectto bothplace age bad?" Fire ControlNotes3(3):25-26. and time. Under these conditions the need of 11. 1940. Statisticalanalysisof forest-fire weather station distribution. Jour. the fire dispatcherwould be best servedwhen Forestry38:318-321. measurements are made closeto high-danger