Bark Beetle Attacks on Ponderosa Pine Following Fire in Northern ...

512 ENVIRONMENTAL ENTOMOLOGY Vol. 32, no. 3Fig. 1. Study sitelocations. TheSideStudy Area (spring wildÞre) burned in thespring 1996, theBridger Knoll Study Area(summer wildÞre) burned in the summer 1996, and the Dauber Study Area (fall prescribed Þre) burned in the fall 1995.Fire Behavior and Intensity. Becauseof theopportunisticnature of this study, direct observations of Þrebehavior characteristics, such as ßame length, are notavailable. Instead, BEHAVE version 4.4 (Andrews1986), a Þre behavior prediction model, was used topredict thepossiblerangeof Þrecharacteristics thatoccurred across each site. We used BEHAVE to predictÞrebehavior in theßaming front based on thefollowing inputs: Northern Forest Fire Laboratory(NFFL) Þre behavior fuel models (Anderson 1982),percent fuel moisture content of the 1-, 10-, and 100-htime lag fuels (Fosberg and Deeming 1971, Rothermel1983), midßame windspeed (Rothermel 1983), andpercent slope (Table 1). Required fuel moisture andwind data for the model were obtained from USDAForest Service records for each Þre.Fireline intensity, because of its relation to ßamelength, is best used to express Þre effects on materialsaffected by convective heating, such as foliage (VanWagener 1973, Finney and Martin 1993). Agee (1993)provides ranges of Þreline intensity for surface Þre(0Ð258 kW m 1 ), understory Þre (258Ð2,800 kWm 1 ), and crown Þre(2,800 kW m 1 ). Based onthesecriteria, thefall Þrewas primarily a surfaceÞrethat reached the crowns of occasional trees, whereasthe spring and summer Þres varied from surface tocrown Þres (Table 1).The fall Þre was a prescribed burn of both naturaland activity fuels (fuels generated from logging activity).Strip ignitions designed to create strip headÞres were initially used to ignite the area. Later intothe ignition phase, lighting patterns were changed tocause low-intensity backing Þres. The 1-, 100, 100-, and1,000-h time-lag moisture classes (Fosberg and Deeming1971, Rothermel 1983) were within normal rangesfor prescribed burning and winds were light (Table 1).Table 1. Fuel model, fuel moisture, slope, and windspeed values used in the fire behavior model BEHAVE to predict the range offire characteristics experienced across the study sitesParameter Fall prescribed Þre Spring wildÞre Summer wildÞreFireline intensity (kW m 1 ) 44 to 234 338 to 3,726 118 to 4,132Fuel model a 9,11 2,9 2,91-h fuel moisture content (%) 9 3 210-h fuel moisture content (%) 10 4 3100-h fuel moisture content (%) 14 5 41,000-h fuel moisture content (%) 18 8 6Slope(%) 5 5 5Ð30Windspeed b (km h 1 ) 3.2Ð8.0 8.0Ð12.9 3.2Ð32.2a Following Anderson (1982).b Midßame windspeed (Rothermel 1983).

514 ENVIRONMENTAL ENTOMOLOGY Vol. 32, no. 3Table 2. Summary statistics for three fires in northern Arizona for diameter breast height (DBH) and crown damage variables (means SEM). ANOVA was performed to test for differences in DBH and crown damage among firesFall prescribed Spring wildÞre Summer wildÞreANOVA resultsdf F PDBH (cm) 24.2a 0.44 40.1b 1.20 51.7c 0.57 2, 1364 256.760 0.001Crown scorch (%) 45.9%a 2.23 55.3%b 1.78 27.3%c 1.07 2, 1364 102.952 0.001Crown consumption (%) N/A 10.3%a 1.17 4.4%b 0.55 1, 1143 26.406 0.001Total crown damage(%) 45.9%a 2.23 65.6%b 1.86 31.5%c 1.22 2, 1364 113.367 0.001Means followed by different letters are signiÞcantly different based on FisherÕs protected LSD test at 0.05.each Þre. Logistic regression models of tree mortalitywere developed for each Þre using SPSS Version 8.0(SPSS 1998). The independent variables consideredfor these models were DBH, crown scorch, crownconsumption, TCD, and IAR. Data on IAR werepooled over all Dendroctonus and Ips species for eachtree because most trees were attacked by several insectspecies. Also, the small number of occurrences ofindividual insect species made pooling over speciestheonly feasibleapproach in our modeling.For each Þre, independent variables were screenedfor their inßuence on tree mortality by comparingvalues for dead versus live trees using two samplet-tests. Screening consisted of using only independentvariables that differed between live and dead trees(P 0.10), were not strongly correlated (r 0.50)with other independent variables, and were signiÞcantly(P 0.10) related to tree mortality. Modelgoodness-of-Þt was assessed based on the followingdiagnostic statistics for each model: Studentized residual,Deviance and CookÕs distance values, and a testof the model null hypothesis that all model coefÞcientsare0 except theconstant, which is comparableto theoverall F test for regression (Hosmer and Lemeshow1989, Norusis 1994). The model used to predict treemortality was:P m 1/(1 exp((b 0 b 1 x 1 ... b n x n )))where P m is theprobability of treemortality, b 0 ,b 1 ,and b n are regression coefÞcients, and x 1 and x n arerepresentative independent variables.Receiver operating characteristic (ROC) curveswere used to assess the overall accuracy for eachlogistic regression model (Saveland and Neuenschwander1990, Finney and Martin 1993, Regelbruggeand Conard 1993, Finney 1999, Stephensand Finney 2002). TheROC curveis a plot of theprobability of a truepositiveprediction, or hit rate(tree is classiÞed as dead when it is dead), versus theprobability of a falsepositive, or falsealarm rate(typeII error; tree is classiÞed dead when it is alive) byvarying the decision criterion from 0 to 1 for groupmembership (Saveland and Neuenschwander 1990,Bradley 1996). TheROC curvevaluecan vary from0.50, which is no better than chance, to 1.0, in whichall predictions are correct (Saveland and Neuenschwander1990, Swets 1996). ROC values between0.50 and 0.70 indicate low accuracy, values between0.70 and 0.90 indicatemoderateaccuracy, and values0.90 indicate very high accuracy (Swets 1996).ResultsTree Characteristics by Fire, Insect Attack Level,and Mortality Group. Average tree DBH was signiÞcantlysmaller for thefall Þrecompared with thespring and summer Þres (Table 2). DBH averagedover live and dead trees was similar for attacked trees(with conÞrmed attacks of Dendroctonus or Ips species)and unattacked trees at the fall and spring Þres,whereas attacked trees were signiÞcantly larger thanunattacked trees at the summer Þre (Table 3). However,for dead trees (Table 4), DBH of attacked treeswas signiÞcantly greater than unattacked trees at thefall Þre(t 2.715; df 38; P 0.010) and thesummerwildÞre(t 2.453; df 114; P 0.016), but not at thespring wildÞre(t 1.296; df 99; P 0.198). For livetrees (Table 4), DBH was signiÞcantly larger for attackedversus unattacked trees only at the summer Þre(t 2.822; df 715; P 0.005). Live unattacked treeshad signiÞcantly larger DBH than dead unattackedtrees at the fall and spring Þres, but not at the summerÞre(Table4).Attacked trees were divided into those that died bytheend of thestudy and thosethat lived. Attacked liveand dead trees had similar DBH at the fall and summerÞres, whereas live attacked trees (i.e., attacked but stillliving at theend of thestudy) weremarginally largerthan dead attacked trees at the spring Þre (Table 4).Attacked live trees were attacked only by red turpentinebeetle, whereas attacked dead trees usually wereattacked by primary bark beetles (e.g., western pinebeetle, mountain pine beetle, roundheaded pine beetle)or Ips species in combination with wood borers.Average crown scorch differed signiÞcantly amongÞres and was lowest on the summer Þre, intermediateon thefall Þre, and highest on thespring Þre(Table2). Attacked trees at all Þres had signiÞcantly morecrown scorch than unattacked trees (Table 3), andattacked dead trees had signiÞcantly more crownscorch than attacked live trees (Table 4). Crownscorch of unattacked dead trees was signiÞcantlyhigher than unattacked live trees for all Þres and wasnearly equal to crown scorch levels of attacked deadtrees (Table 4).Crown consumption occurred only in the springand summer Þres, and was signiÞcantly lower in thesummer Þrecompared with thespring Þre(Table2).Crown consumption of attacked trees was signiÞcantlyhigher than unattacked trees for spring andsummer wildÞres, and crown consumption of attacked

June2003 MCHUGH ET AL.: BARK BEETLE ATTACKS 515Table 3. Means (SEM) of tree variables by season of fire (fall, spring, summer) for attacked and unattacked ponderosa pines 3 yrpostfire in northern Arizona. P values are results of t-tests between attacked versus unattacked trees. Attacked trees have attacks byDendroctonus and Ips species, and unattacked trees do notSeason/variable Attacked Unattacked t PFall n 25 n 197 df 220DBH (cm) 25.8 1.53 24.1 0.46 1.280 0.202Crown scorch (%) 58.0 6.81 44.5 2.35 1.927 0.055Crown consumption (%) 0 (0) 0 (0) N/A N/ATotal crown damage(%) 58.0 6.81 44.5 2.35 1.927 0.055Insect attack rating 1.3 (0.01) 0 (0) 39.832 0.001Spring n 127 n 185 df 310DBH (cm) 40.6 2.05 39.7 1.45 0.334 0.738Crown scorch (%) 66.2 2.42 47.8 2.35 5.286 0.001Crown consumption (%) 18.6 2.25 4.5 1.03 6.274 0.001Total crown damage(%) 84.8 1.73 52.4 2.48 9.774 0.001Insect attack rating 1.4 0.004 0 (0) 39.211 0.001Summer n 154 n 679 df 831DBH (cm) 55.8 1.32 50.7 0.63 3.515 0.001Crown scorch (%) 46.7 2.68 22.8 1.09 9.116 0.001Crown consumption (%) 7.1 1.54 3.7 0.58 2.419 0.016Total crown damage(%) 53.8 2.91 26.5 1.27 9.123 0.001Insect attack rating 1.2 0.003 0 (0) 82.938 0.001trees in the spring Þre was 2.6 times higher than attackedtrees in the summer Þre (Table 3). Crownconsumption of dead trees was signiÞcantly greaterthan crown consumption of live trees regardless ofattack level (Table 4).TCD was signiÞcantly different among Þres and waslowest in the summer Þre, intermediate in the fall Þre,and greatest in thespring Þre(Table2). For all Þres,TCD was signiÞcantly higher for attacked trees versusunattacked trees (Table 3). For all Þres, TCD of deadtrees was signiÞcantly higher than live trees regardlessof attack level (Table 4).The percentage of trees attacked by Dendroctonusand Ips species was lowest in the fall Þre (11%, 25 of222 trees), intermediate in the summer Þre (19%, 154of 833 trees), and highest in the spring Þre (41%, 127of 312 trees). The average IAR was nearly equal for allÞres and ranged from a low of 1.2 in the summer Þreto a high of 1.4 in thespring Þre(Table3). Dead treeshad a signiÞcantly higher IAR than live trees for allÞres (Table 4).Distribution and Occurrence of Insect Taxa. Totalinsect occurrence pooled over all Þres was lowest inall years for mountain pine beetle and roundheadedpine beetle (Fig. 2). Mountain pine beetle was notfound on any trees in the fall or spring Þres and on onlyeight trees on the summer Þre 3 yr postÞre. Twenty-Þve percent of all trees attacked by mountain pinebeetle on the summer Þre were attacked in the ÞrstpostÞre year, 1997, and attacks increased gradually in1998 and 1999 (Fig. 2). The mean DBH for treesattacked by mountain pine beetle in the summer Þrewas 60.2 cm with a mean TCD of 71%. Few trees wereattacked by roundheaded pine beetle for the spring(4) and summer Þres (5), and no trees were attackedby this species for the fall Þre 3 yr postÞre. For round-Table 4. Means (SEM) of tree variables by season of fire (fall, spring, summer) for live and dead attacked and unattacked ponderosapines 3 yr postfire in northern Arizona. Significance values are results of t-tests between live and dead trees for attacked and unattackedtrees. Attacked trees have attacks by Dendroctonus and Ips species, and unattacked trees do notVariableAttackedUnattackedLiveDead t P LiveDead t PFall n 14 n 11 df 23 n 168 n 29 df 195DBH (cm) 26.4 1.54 25.1 2.96 0.427 0.674 25.0 0.47 18.5 0.10 5.397 0.001Crown scorch (%) 36.4 7.23 85.5 5.62 5.119 0.001 37.6 2.322 84.5 2.96 8.187 0.001Crown consumption (%) N/A N/A N/A N/A N/A N/A N/A N/ATotal crown damage(%) 36.4 7.23 85.5 5.619 5.119 0.001 37.6 2.32 84.5 2.96 8.187 0.001Insect attack rating 1.0 0.00 1.6 0.152 4.748 0.001 N/A N/A N/A N/ASpring n 45 n 82 df 125 n 166 n 19 df 183DBH (cm) 45.6 3.44 37.8 2.52 1.830 0.070 40.8 1.52 30.5 4.31 2.186 0.030Crown scorch (%) 60.7 3.71 69.3 3.12 1.711 0.090 45.7 2.44 66.8 7.22 2.783 0.006Crown consumption (%) 9.1 2.29 23.8 3.11 3.235 0.002 1.9 0.48 26.8 7.49 8.701 0.001Total crown damage(%) 69.8 3.28 93.1 1.28 7.811 0.001 47.7 2.51 93.7 1.57 6.185 0.001Insect attack rating 1.02 0.002 1.7 0.005 8.703 0.001 N/A N/A N/A N/ASummer n 98 n 56 df 152 n 619 n 60 df 677DBH (cm) 55.9 1.69 55.6 2.13 0.128 0.898 52.8 0.64 47.8 2.35 1.461 0.144Crown scorch (%) 35.3 3.02 66.6 3.91 6.300 0.001 19.1 1.00 60.7 4.44 11.859 0.001Crown consumption (%) 1.4 0.63 17.1 3.73 5.349 0.001 0.695 0.19 34.8 4.65 21.813 0.001Total crown damage(%) 36.7 3.10 83.8 3.09 9.953 0.001 19.8 1.05 95.5 1.15 22.365 0.001Insect attack rating 1.07 0.003 1.3 0.006 3.990 0.001 N/A N/A N/A N/A

516 ENVIRONMENTAL ENTOMOLOGY Vol. 32, no. 3Fig. 2. Cumulativeoccurrenceof insect taxa on ponderosapine over three years (1997, 1998, 1999) after Þre innorthern Arizona. Cumulativeoccurrenceis thenumber oftrees attacked by each insect totaled over years. WPB western pine beetle, MPB mountain pine beetle, RPB roundheaded pine beetle, IPS Ips species, RTB redturpentine beetle, WB wood borers in the Buprestidae andCerambycidae families. Data are pooled over fall, spring, andsummer Þres.headed pine beetle, 44% of total attacks occurred in1997 with small increases in 1998 and 1999 (Fig. 2).Mean DBH of trees attacked by roundheaded pinebeetlewas 50.7 cm for thespring Þreand 62.0 cm forthe summer Þre. Mean TCD for trees attacked byroundheaded pine beetle was 82% for the spring Þreand 90% for the summer Þre.Western pine beetle was found on only three treesat the fall Þre, 19 trees at the spring Þre, and 13 treeson the summer Þre 3 yr postÞre. Sixty-eight percent oftotal attacks by this species occurred in 1997, withslight increases in 1998 and 1999 (Fig. 2). The meanTCD of trees attacked by western pine beetle was 92%on thespring Þreand 83% on thesummer Þre. All threetrees attacked by this species on the fall Þre had 100%TCD. Average DBH of trees attacked by western pinebeetle was 22.8 cm for the fall Þre, 53.1 cm for thespring Þre, and 56.6 cm for the summer Þre.Only three trees on the fall Þre and nine trees on thesummer Þre were attacked by Ips species 3 yr postÞre.However, the spring Þre had 50 trees attacked by Ipsspecies. For the fall Þre, 100% of all Ips attacks occurredin the Þrst postÞre year. For the spring Þre, 72%of all Ips attacks occurred in theÞrst postÞreyear. Forthe summer Þre, 22% of all Ips attacks occurred in theÞrst postÞre year (Fig. 2). Over all Þres, 66% of Ipsattacks occurred in 1997, with only slight increases in1998 and 1999 (Fig. 2). The mean DBH of trees attackedby Ips was 18.4 cm for thefall Þre, 23.3 cm forthe spring Þre, and 42.4 cm for the summer Þre. Themean TCD for trees attacked by Ips was 80% for thefall Þre, 93% for the spring Þre, and 83% for the summerÞre.Red turpentine beetle was the most common speciesof Dendroctonus at all Þres. In the fall and springÞres, 95% and 87% of attacks by this species occurred1 yr postÞre (1997), whereas for the summer Þre only27% of attacks occurred 1 yr postÞre. Over all Þres, 56%of red turpentine beetle attacks occurred in 1997, andattacks increased in 1998 and 1999 (Fig. 2).Red turpentine beetle was present more often onlive than on dead trees in the fall and summer Þres. Ofthe 19 trees attacked by this insect in the fall Þre, 14were alive and 5 were dead 3 yr postÞre. Live and deadtrees attacked by red turpentine beetle in the fall Þrehad a similar DBH (t 1.154; df 17; P 0.142), butlive trees had a lower TCD (36%) than dead (90%)trees (t 4.300; df 17; P 0.001). In the summer Þre,127 trees were attacked by red turpentine beetle, ofwhich 87 were alive and 40 were dead 3 yr postÞre.Similar to the fall Þre, live and dead trees attacked byred turpentine beetle had similar DBH in the summerÞre(t 0.018; df 125; P 0.986), and live trees hadlower TCD (39%) than dead trees (82%) (t 7.673;df 125; P 0.001). Occurrence of red turpentinebeetle on the spring Þre was similar between live (45)and dead (49) trees 3 yr postÞre. Again, DBH did notdiffer between live and dead trees attacked by redturpentinebeetleon thespring Þre(t 0.867; df 92;P 0.388), and live trees had a lower TCD (70%) thandead (92%) trees (t 6.294; df 92; P 0.001).Wood borers in the Cerambycidae and Buprestidaefamilies were the most common insect found at all Þres(Fig. 2). Most colonization by wood borers occurredin theÞrst postÞreyear (1997) for thefall (94% of allcolonized trees) and spring (87% of all colonizedtrees) Þres, whereas only 58% of total colonizationoccurred in this year for the summer Þre. However, by1998, 98% of all wood borer colonization had occurredin the summer Þre. Over all Þres, 73% of all trees withwood borers were colonized in 1997, and 98% in 1998(Fig. 2). Wood borers colonized both live and deadtrees; however, in live trees they were generally foundin dead portions of tree cambium as a result of girdlingfrom Þre. In dead trees wood borers colonized isolatedareas of dead cambium and phloem sections of the treebole in addition to areas previously attacked by Dendroctonusand Ips species.Mortality Models. The logistic regression model developedfor the fall prescribed Þre was highly significantand Þt the data well based on a test of the modelnull hypothesis (X 2 104.921; df 2; P 0.001).Diagnostic statistics also indicated that overall modelÞt was acceptable. The ROC value of 0.93 indicatesvery high accuracy in model discrimination betweenlive and dead trees. The model suggests that probabilityof tree mortality increased as TCD and IARincreased. Moreover, insect attacks reduced theamount of TCD associated with high levels of treemortality (Fig. 3a). For example, for trees with a TCDof 50%, the probability of mortality was 0.02 for treeswith no insect attacks, 0.14 for partial attacks, and 0.51for mass attacks (Fig. 3a).The logistic regression model developed for thespring prescribed Þre was also highly signiÞcant and Þtthe data well based on a test of the model null hypothesis(X 2 241.066; df 2; P 0.001) and diag-

June2003 MCHUGH ET AL.: BARK BEETLE ATTACKS 517Fig. 3. Distribution of predicted probability of ponderosa pine mortality for logistic regression models using total crowndamage (TCD) and insect attack rating (IAR) (none, partial, mass; Dendroctonus and Ips species) for three Þres in northernArizona. Fall prescribed Þre (a), spring wildÞre (b), and summer wildÞre (c). See Table 5 for model equations.nostic statistics (Table5). Similar to thefall Þre, themodel for thespring Þresuggests that theprobabilityof tree mortality increased as TCD and IAR increased,and higher levels of mortality at equal TCD wheninsect attacks occurred (Fig. 3b). For example, fortrees with 70% TCD, probability of mortality was 0.03for trees with no attacks, 0.18 with partial attacks, and0.63 with mass attacks (Fig. 3b).Model results for the summer Þre were similar to thefall and spring Þres. The model developed for thesummer prescribed Þre was highly signiÞcant and Þtthe data well based on a test of the model null hypothesis(X 2 413.577; df 2; P 0.001). Diagnosticstatistics also supported acceptable model Þt (Table5). For the summer Þre, partial and mass attacks alsoincreased the probability of tree mortality at a givenlevel of TCD (Fig. 3c). For trees with partial or massattacks, TCD was generally lower in the fall comparedwith the spring and summer Þres (Fig. 3). For example,TCD associated with 50% mortality of trees with

518 ENVIRONMENTAL ENTOMOLOGY Vol. 32, no. 3Table 5. Logistic regression coefficients (SEM), -2 log likekihood ratio statistic (-2LL), and receiver operating characteristic curvevalue (ROC) for selected ponderosa pine mortality prediction equations following fire in northern Arizona for a fall prescribed fire, springwildfire, and summer wildfire. All model coefficients were significant at 0.05Model form: P m 1/1 exp B 0 B 1 TCD B 2 IAR aB 0 B 1 B 2 -2LL ROCFall prescribed Þre 8.826 1.4735 0.103 0.0186 1.864 0.5771 104.5 0.93Spring wildÞre 11.682 1.7693 0.116 0.0194 2.047 0.3093 151.8 0.96Summer wildÞre 7.979 0.8073 0.087 0.0091 1.321 0.2820 258.8 0.97a Where: P m , predicted probability of mortality; B 0 , intercept, and B n , model coefÞcients; TCD, total crown damage (scorch consumption)(0Ð100); IAR, insect attack rating (0, none; 1, partial attack; 2, mass attack).mass attacks was 50% for thefall Þre, 65% for thespringÞre, and 62% for the summer Þre (Fig. 3).DiscussionBased on precipitation data for Flagstaff, AZ (NationalOceanic and Atmospheric Administration,Stations 023010/03103 and 023009/9999, http://lwf.ncdc.noaa.gov/servlets/ACS), the postÞre years inour study included one extreme drought (1996),two mild droughts (1997, 1999), and one wet year(1998). Total yearly precipitation in 1996, the extremedrought, was 53% of normal. Years 1997 and 1999 weremilder droughts, with similar below-average precipitation(1997: 68% of normal; 1999: 69% of normal).Precipitation during year 1998 was 20% above normal.Thedrought conditions following theÞres in our studylikely reduced postÞre tree survival (Ryan 2000), andmay have increased tree susceptibility to bark beetles(Furniss and Carolin 1977). Our results should not begeneralized to other Þres or other postÞre weatherconditions without further evaluation.Mortality in Þre-damaged ponderosa pine as a resultof attacks by western pine beetle has been documentedfor parts of southern Oregon and northernCalifornia (Miller and Patterson 1927, Miller and Keen1960). Western pine beetle can breed in old, slowgrowingtrees as well as in dense overstocked stands ofeven-aged young trees (DeMars and Roettgering1982). These conditions existed at all our study sites,however, this insect was not a major contributor oftree mortality in our study because it occurred on only35 trees across all Þres. Average TCD of trees attackedby western pine beetle was 80%, which may havelimited insect performance and spread to other treesbecause of associated phloem damage (Miller andKeen 1960, DeMars and Roettgering 1982). However,themorelikely explanation for thelow occurrenceofwestern pine beetle in our study is that populationswere low near our study sites. This is consistent withÞndings by Sanchez-Martinez and Wagner (2002)who reported that populations of western pine beetlewere at endemic levels in the 1990s in the vicinity ofour fall and spring Þre study sites in northern Arizona.A review of Aerial Detection Survey (ADS) recordsmaintained by the USDA Forest Service, SouthwesternRegion Entomology and Pathology, Arizona ZoneOfÞceof Entomology and Pathology located in Flagstaff,AZ for the years 1995Ð1999 showed isolatedgroup clusters (Þve tree groups) of ponderosa pinekilled by western pine beetle within 0.4 km of thesummer Þre. However, these records suggest thatthere were no large populations of western pine beetlebecause the next largest group cluster (20 tree groups)was 16 km from the summer Þre. Important factorsaffecting the intensity of concentrations of westernpine beetles in and around Þres include the type ofburn, Þre timing, degree of crown injury, weather andstand conditions at thetimeof theÞre, and, lastly, thecurrent cycle of infestation in the area (Miller andKeen 1960). In our study, although many factors mayhave been ripe for an increased infestation, there wasnot a population availableto takeadvantageof thehabitat.Mountain pine beetle was not found on trees at thefall or spring Þres, and it occurred rarely at the summerÞre, which is consistent with the view that mountainpine beetle is not attracted to Þre-damaged trees(Ryan and Amman 1994, Rasmussen et al. 1996). Firemay not strongly promotepopulation outbreaks ofmountain pine beetle because it prefers fast-growingtrees rather than Þre-damaged, slow-growing trees(Rasmussen et al. 1996, Ryan and Amman 1996). However,another explanation for the low occurrence ofmountain pine beetle in our study is low populations,because it was considered to be at an endemic populationlevel during the 1990s in the vicinity of our falland spring Þres (Sanchez-Martinez and Wagner2002). ADS records from the USDA Forest Service,Southwestern Region Entomology and Pathology, ArizonaZoneOfÞceof Entomology and Pathology locatedin Flagstaff, AZ for the years 1995Ð1999 indicatedthat several small clusters (Þve tree groups) ofmountain pine beetle killed trees were located within0.4 km of the summer Þre, and that a large outbreak ofmountain pine beetle (5,000 trees) occurred in 1997.However, this outbreak was over 32 km away andnever expanded outside this area toward the summerÞre.Outbreaks of roundheaded pine beetle have causedconsiderablemortality of ponderosa pinein someArizonaand Utah forests (Negron 1997, Negron et al.2000). Massey et al. (1977) reported outbreaks ofroundheaded pine beetle ranging from several hundredtrees to 400,000 trees over a 60,700 ha area in NewMexico. However, this beetle was rare on our studysites in northern Arizona. Most outbreaks of this speciesoccur in slow growing, stagnated stands (Negron

June2003 MCHUGH ET AL.: BARK BEETLE ATTACKS 5191997, Negron et al. 2000). Tree diameters on the twoÞres in our study where it occurred (spring, summer)are within ranges reported for outbreaks of this speciesin trees not damaged by Þre (Massey et al. 1977,Negron et al. 2000). We were unable to Þnd publishedinformation on the relationship between roundheadedpine beetle and Þre damaged trees. Low populationsof this species in our study could have beencaused by a weak roleof Þrein thepopulation dynamicsof this species, or by low populations near ourstudy sites. Sanchez-Martinez and Wagner (2002) didnot report any occurrences of this species in theirstudy which occurred near our fall and spring Þres. Anevaluation of the USDA Forest Service, SouthwesternRegion Entomology and Pathology ADS records 1995Ð1999 showed no recorded observations of roundheadedpine beetle on the North Kaibab Ranger Districtwhereour summer Þrestudy sitewas located.Our measure of Ips populations may bean underestimatebecause we only examined the lower stem oftrees and not the upper stem or branches (Fischer1980). Sanchez-Martinez and Wagner (2002) reportedthat Ips attacks increased during the 1990s inthevicinity of our fall and spring study sites. In theareaof the summer Þre, only isolated clusters (Þve treegroups) of Ips populations were recorded in USDAForest Service, Southwestern Region Entomology andPathology ADS records for the years 1995Ð1999. Thelow occurrence of Ips in our study may bea result ofnegative effects of Þre on host materials for broodproduction, or a lack of populations to takeadvantageof suitablehabitat.Red turpentine beetle is often found in Þre-damagedtrees, yet it is rarely considered to be an importantcause of tree mortality (Herman 1954, Wagener1961, Mitchell and Martin 1980). Red turpentine beetlewas the most common subcortical insect except forwood borers in all Þres. It occurred much more oftenon live trees (i.e., trees that survived for 3 yr after Þre)than dead trees for the fall and summer Þres. In contrast,its occurrenceon thespring Þrewas similarbetween live and dead trees. Overall, our results forthe fall and summer Þres are similar to other studies ofponderosa pine, which concluded that attacks by thisspecies are not related to tree mortality (Herman 1954,Wagener 1961). However, our results from the springÞre suggest that red turpentine beetle may interactwith other bark beetles to inßuence death of Þredamagedtrees on some sites.Wood borers in the Cerambycidae and Buprestidaefamilies were the most common insects at all sites inour study. Consistent with our observations, woodborers are known to be abundant in Þre-damaged orÞre-killed trees (Schmid and Parker 1990, Powell et al.2002). While wood borers likely are not importantsources of tree mortality, their economic damage canbelargebecauseof lumber degradation and introductionof wood decay fungi (Mitchell and Martin 1980,Schmid and Parker 1990). The contribution of woodborers to decay and breakdown of Þre-damaged timbermay be a concern to resource managers, especiallyif salvageoperations areplanned.Trees attacked by Dendroctonus and Ips species asa group had morecrown damagefrom Þrethan unattackedtrees. This Þnding suggests that heavy crowndamage by Þre reduced tree carbohydrate allocationto resin defenses that repel bark beetles, as reportedrecently for Þre-damaged ponderosa pines at one Þrein northern Arizona located near our study sites (Wallinet al. 2003). Another explanation is that these insectsare attracted to Þre-damaged trees. Althoughlarger trees may offer more food for beetle broodproduction because of thicker phloem (Amman andPasek 1986), differences in DBH between attackedand unattacked trees were not large or consistent forall Þres. Thus, we found little evidence for strongselection of speciÞc size classes of trees for attack byDendroctonus and Ips species as a group under ourstudy conditions. Unfortunately, small sample sizes forsomeinsects, and theoccurrenceof multipleinsectspecies on many trees, prevented comparison of treesize between attacked and unattacked trees by individualinsect species.Our logistic regression models showed low probabilityof mortality over 3 yr after Þre for trees notattacked by Dendroctonus and Ips species when TCDwas less than 70% consistent with several other studiesof ponderosa pine (Herman 1954, Dieterich 1979,Stephens and Finney 2002). At a TCD 70Ð80% (dependingon the Þre), the probability of mortality increaseddramatically for unattacked trees. No otherstudies of ponderosa pine have included insect attacksin models of tree mortality after Þre. We found thatattacks of Dendroctonus and Ips species as a groupincreased the probability of mortality compared withunattacked trees. Partial attacks increased tree mortalitycompared with unattacked trees when TCDexceeded 50%. Mass attacks had a stronger inßuence,and increased tree mortality when TCD exceeded40%. These Þndings suggest an importantroleof attacks by Dendroctonus and Ips beetles as agroup in killing ponderosa pines with moderate toheavy amounts of crown damage from Þre. Salvagelogging operations after Þre may need to considerponderosa pines for harvest with lower levels of crowndamagethan previously thought in stands in whichattacks by Ips and Dendroctonus species are occurringor populations areincreasing.Our results show that Ips and Dendroctonus speciesas a group preferred ponderosa pines that wereheavily damaged by Þre compared with lightly damagedtrees on three Þres that burned in northernArizona in 1995 or 1996 when these insects were at anendemic population level. Such a preference may promotepopulation increases of these insects if Þre damagedtrees are common. In northern Arizona, wildÞresin ponderosa pine forests are currently increasing inseverity and size as a result of increasing fuel loads(Covington and Moore1994, Covington et al. 2001)and frequent droughts since 1996. Thus, currenttrends suggest growing numbers of Þre-damaged ponderosapines available for Ips and Dendroctonus speciesin northern Arizona forests. However, we havelittle information on fecundity and brood perfor-

520 ENVIRONMENTAL ENTOMOLOGY Vol. 32, no. 3mance of these insects in ponderosa pines with heavyÞre damage. Such information is needed to betterevaluate effects of Þre in ponderosa pine forests onbark beetle populations, and should be emphasized infuture research.AcknowledgmentsThepaper was written whilethesenior author was aforester with the Coconino National Forest. This researchwas supported by the Coconino and Kaibab National Forestsand the Arizona Zone Forest Pest Management Group of theU.S. Forest Service, Southwestern Region. The authors thankthe Peaks Ranger District Marking Crew, Coconino NationalForest, and the South and North Kaibab Ranger DistrictTimber Marking Crews, Kaibab National Forest for theirassistance in Þeld data collection; and Steve Dudley of theArizona ZoneOfÞceof Entomology and Pathology, U.S.Forest Service for providing the necessary aerial detectionsurvey records in a GIS ready format. One anonymous reviewerprovided helpful comments that improved the qualityof themanuscript. Wethank J.A. Anhold and J. D. McMillin,Arizona ZoneOfÞceof Entomology and Pathology, U.S.Forest Service, for critiquing early versions of the manuscript.References CitedAgee, J. K. 1993. Fire ecology of paciÞc northwest forests.Island Press, Washington, D. C.Amman, G. D., and K. C. Ryan. 1991. Insect infestations ofÞre-injured trees in the Greater Yellowstone Area. U.S.Dep. Agric. For. Serv. Research Note INT-398. Ogden,UT.Amman, G. D., and J. E. Pasek. 1986. Mountain pine beetlein ponderosa pine: effects of phloem thickness on egggallery density. U.S. Dep. Agric. For. Serv. Intermount.For. Range Exp. Stn. Res. Pap. No. INT-RP-367, Ogden,UT.Anderson, H. E. 1982. Aids to determining fuel models forestimating Þre behavior. U.S. Dep. Agric. For. Serv. Gen.Tech. Rep. INT-122, Ogden, UT.Andrews, P. A. 1986. BEHAVE: Fire behavior predictionand fuel modeling systemÑBURN subsystem Part 1. U.S.Dep. Agric. For. Serv. Gen. Tech. Rep. INT-194, Ogden,UT.Beatty, J. S. 1986. Forest insect and disease guide. U.S. Dep.Agric. For. Serv. R3-TP-1, Forest Pest Management,Southwestern Region, Albuquerque, NM.Bentz, B. J., J. A. Powell, and J. A. Logan. 1996. Localizedspatial and temporal attack dynamics of the mountainpine beetle in lodgepole pine. U.S. Dep. Agric. For. Serv.Inter. Res. Stat. Res. Pap. INT-RP-494, Ogden, UT.Bradley, A. P. 1996. ROC curves and the X 2 test. PatternRecognition Letters 17: 287Ð294.Bradley, T., and P. Tueller. 2001. Effects of Þre on barkbeetle presence on Jeffrey pine in the Lake Tahoe Basin.For. Ecol. Mgmt. 142: 205Ð214.Brewer, D. G., R. K. Jorgensen, L. P. Munk, W. A. Robbie, andJ. L. Travis. 1991. Terrestrial ecosystems survey of theKaibab National Forest. U.S. Dep. Agric. For. Serv. S. W.Region Report, Albuquerque, NM.Connaughton, C. A. 1936. Firedamagein theponderosapinetypein Idaho. J. For. 34: 46Ð51.Covington, W. W., and M. M. Moore. 1994. Southwesternponderosa pine forest structure and resource conditions:changes since Euro-American settlement. J. For. 92: 39Ð47.Covington, W. W., P. Z. Fule, S. C. Hart, and R. P. Weaver.2001. Modeling ecological restoration effects on ponderosapine forest structure. Res. Ecol. 9: 421Ð431.DeMars Jr., C. J., and B. H. Roettgering. 1982. Western pinebeetle. Forest insect and disease Leaßet 1. U.S. Dep.Agric. For. Serv., Washington, D.C.DeNitto, G., B. Cramer, K. Gibson, B. Lockman,T. McConnell, L. Stripe, N. Sturdevant, and J. Taylor.2000. Survivability and deterioration of Þre-injuredtrees in the Northern Rocky Mountains: A review of theliterature. U.S. Dep. Agric. For. Serv. State and PrivateForestry, Forest Health Protection Report 2000Ð13, Missoula,MT.Dieterich, J. H. 1979. Recovery potential of Þre-damagedsouthwestern ponderosa pine. U.S. Dep. Agric. For. Serv.Res. Note RM-379, Ft. Collins, CO.Finney, M. A. 1999. Fire-related mortality in ponderosapine in eastern Montana. U.S. Dep. Agric. For. Serv. FinalReport INT-93800-RJVA, Fire Sciences Laboratory, Missoula,MT.Finney, M. A., and R. E. Martin. 1993. Modeling effects ofprescribed Þre on young-growth coast redwood trees.Can. J. For. Res. 23: 1125Ð1135.Fischer, W. C. 1980. Prescribed Þre and bark beetle attackin ponderosa pine forests. Fire Manage. Notes 41: 10Ð12.Fosberg, M. A., and J. E. Deeming. 1971. Derivation of the1- and 10- hour timelag fuel moisture calculations forÞre-danger rating. U.S. Dep. Agric. For. Serv. Res. NoteRM-207, Ft. Collins, CO.Furniss, R. L., and V. M. Carolin. 1977. Western forest insects.U.S. Dep. Agric. For. Serv. PaciÞc NW For. RangeExp. Stn. Misc. Publ. 1339, Washington, D.C.Geiszler, D. R., R. L. Gara, C. H. Driver, V. F. Gallucci, andR. E. Martin. 1980. Fire, fungi, and beetle inßuences ona lodgepole pine ecosystem of south-central Oregon. Oecologia(Berl.). 46: 239Ð243.Hanula, J. L., J. R. Meeker, D. R. Miller, and E. L. Barnard.2002. Association of wildÞrewith treehealth and numbersof pine bark beetles, reproduction weevils and theirassociates in Florida. For. Ecol. Manage. 170: 233Ð247.Harrington, M. G. 1987. Ponderosa pine mortality fromspring, summer, and fall crown scorching. West. J. Appl.For. 2: 14Ð16.Harrington, M. G. 1993. Predicting Pinus ponderosa mortalityfrom dormant season and growing season Þre injury.Int. J. Wild. Fire. 3: 65Ð72.Herman, F. R. 1954. A guide for marking Þre-damaged ponderosapine in the southwest. U.S. Dep. Agric. For. Serv.Res. Note No. 13, Ft. Collins, CO.Hosmer Jr., D. W., and S. Lemeshow. 1989. Applied logisticregression. Wiley, New York.Littke, W. R., and R. I. Gara. 1986. Decay of Þre-damagedlodgepole pine in south-central Oregon. For. Ecol.Manage. 17: 279Ð287.Massey, C. L., D. D. Lucht, and J. M. Schmid. 1977. RoundheadedPine <strong>Beetle</strong>. Forest Insect and Disease Leaßet155. U.S. Dep. Agric. For. Serv., Washington, D.C.McHugh, C. W., and T. E. Kolb. 2003. Ponderosa pine mortalityfollowing Þrein northern Arizona. Int. J. Wild. Fire12: 7Ð22.McCullough, D. G., R. A. Werner, and D. Neumann. 1998.Fire and insects in northern and boreal forest ecosystemsof North America. Annu. Rev. Entomol. 43: 107Ð127.Menges, E. S., and M. A. Deyrup. 2001. PostÞresurvival insouth Florida slash pine: interacting effects of Þre intensity,Þre season, vegetation, burn size, and bark beetles.Int. J. Wild. Fire10: 53Ð63.

June2003 MCHUGH ET AL.: BARK BEETLE ATTACKS 521Miller G., N. Ambos, P. Boness, D. Reyher, G. Robertson,K. Scalzone, R. Steinke, and T. Subirge. 1995. Terrestrialecosystems survey of the Coconino National Forest. U.S.Dep. Agric. For. Serv. SW Region, Albuquerque, NM.Miller, J. M., and J. E. Patterson. 1927. Preliminary studieson the relation of Þre injury to bark beetle attack inwestern yellow pine. J. Agric. Res. 34: 597Ð613.Miller, J. M., and F. P. Keen. 1960. Biology and control of thewestern pine beetle: a summary of the Þrst Þfty years ofresearch. U.S. Dep. Agric. For. Serv. PaciÞc SW For.RangeExp. Stn. Misc. Publ. 800, Washington, D.C.Mitchell, R. G., and R. E. Martin. 1980. Fireand insects inpinecultureof thePaciÞc Northwest, pp. 182Ð190. InMartin, R. E. [ed.], Proceedings, 1980 Sixth Conferenceon Fire and Forest Meterorology. Washington D. C.,Society of American Foresters, Washington, D.C.Morris, W. G., and E. L. Mowat. 1958. Some effects ofthinning a ponderosa pine thicket with a prescribed Þre.J. For. 56: 203Ð209.Negron, J. 1997. Estimating probabilities of infestation andextent of damage by the roundheaded pine beetle inponderosa pine in the Sacramento Mountains, New Mexico.Can. J. For. Res. 27: 1936Ð1945.Negron, J. F., J. L. Wilson, and J. A. Anhold. 2000. Standconditions associated with roundheaded pine beetle(Coleoptera: Scolytidae) infestations in Arizona andUtah. Environ. Entomol. 29: 20Ð27.(NOAA) National Oceanic and Atmospheric Administration.1997. Local climatological data: annual summarywith comparativedata. Flagstaff, AZ.Norusis, M. J. 1994. SPSS Advanced statistics 6.1. SPSS Inc.,Chicago, Illinois.Ott, R. L. 1992. An introduction to statistical methods anddata analysis, 4th ed. Wadsworth Inc., Belmont, CA.Peterson, D. L. 1985. Crown scorch volumeand scorchheight: estimates of postÞre tree condition. Can. J. For.Res. 15: 596Ð598.Peterson, D. L., and M. J. Arbaugh. 1986. PostÞresurvival inDouglas-Þr and lodgepole pine: comparing the effects ofcrown and bole damage. Can. J. For. Res. 16: 1,175Ð1,179.Powell, H.D.W., S. J. Hejl, and D. L. Six. 2002. Measuringwoodpecker food: a simple method for comparing woodboringbeetle abundance among Þre-killed trees. J. FieldOrnithol. 73: 130Ð140.Raffa, K. F., and A. A. Berryman. 1982. Physiological differencesbetween lodgepole pines resistant and susceptibleto themountain pinebeetleand associated microorganisms.Environ. Entomol. 11: 486Ð492.Raffa, K. F., and A. A. Berryman. 1983. Theroleof host plantresistance in the colonization behavior and ecology ofbark beetles (Coleoptera: Scolytidae). Ecol. Mono. 53:27Ð49.Rasmussen, L. A., G. A. Amman, J. C. Vandygriff, R. D. Oakes,A. S. Munson, and K. E. Gibson. 1996. <strong>Bark</strong> beetle andwood borer infestation in the Greater Yellowstone Areaduring four postÞre years. U.S. Dep. Agric. For. Serv. Res.Pap. INT-RP-487, Ogden, UT.Regelbrugge, J. C., and S. C. Conard. 1993. Modeling treemortality following wildÞrein Pinus ponderosa forests inthe Central Sierra of California. Int. J. Wild. Fire. 3:139Ð148.Reinhardt, E. D., and K. C. Ryan. 1988. How to estimate treemortality resulting form underburning. Fire Manage.Notes 49: 30Ð36.Rothermel, R. C. 1983. How to predict the spread and intensityof forest and range Þres. U.S. Dep. Agric. For. Serv.Gen. Tech. Rep. INT-143, Ogden, UT.Ryan, K. C. 1982. Evaluating potential tree mortality fromprescribed burning, pp. 167Ð179. In D. M. Baumgartner[ed.], Site preparation and fuels management on steepterrain. Washington State University, Cooperative Extension,Pullman, WA.Ryan, K. C. 1983. Techniques for assessing Þre damage totrees, pp. 1Ð11. In J. E. Lotan [ed.], Proceedings of theSymposium: Fire its Field Effects. Intermountain FireCouncil, Missoula, MT.Ryan, K. C. 1990. Predicting prescribed Þre effects on treesin the interior west, pp. 148Ð162. In M. E. Alexander, andG. F. Bisgrove [tech. coors.], Proceedings, The Art andScience of Fire Management. Forestry Canada, InformationReport NOR-X-309, Edmonton, Alberta, Canada.Ryan, K. C. 1998. Analysis of therelativevalueof morphologicalvariables in predicting Þre-caused tree mortality,pp. 1511Ð1526. In D. X. Viegas [ed.], Proceedings, IIIInternational Conference on Forest Fire Research and14th Conference on Fire and Forest Meteorology. Luso,Coimbra, Portugal. ADAI Ð Associacao para o Desenvolvimentoda Aerodinamica Industrial, Coimbra, Portugal.Ryan, K. C. 2000. Effects of Þre injury on water relations ofponderosa pine, pp. 58Ð66. In W. K. Moser and C. F.Moser [eds.], Fire and forest ecology: innovative silvicultureand vegetation management. Tall Timbers FireEcology Conference Proceedings, No. 21. Tall TimbersResearch Station, Tallahassee, FL.Ryan, K. C., and G. D. Amman. 1994. Interactions betweenÞre-injured trees and insects in the Greater YellowstoneArea, pp. 259Ð271. In Despain [ed.], Plants and theirenvironments. Proceedings, of the First Biennial ScientiÞcConference in the Greater Yellowstone Ecosystem.U.S. Dep. of the Interior, National Park Service, NaturalResources Publication OfÞce, Denver, CO.Ryan K. C., and G. D. Amman. 1996. <strong>Bark</strong> beetle activity anddelayed tree mortality in the Greater Yellowstone Areafollowing the1988 Þres, pp. 151Ð158. In J. Greenlee [ed.],Ecological Implications of Fire in Greater Yellowstone.Proceedings, 2nd Biennial Conference on the GreaterYellowstone Ecosystem. International Association ofWildland Fire, FairÞeld, WA.Ryan, K. C., D. L. Peterson, and E. D. Reinhardt. 1988.Modeling long-term Þre-caused mortality of Douglas-Þr.For. Sci. 34: 190Ð199.Ryan, K. C., and E. D. Reinhardt. 1988. Predicting postÞremortality of seven western conifers. Can. J. For. Res. 18:1291Ð97.Santoro, A. E., M. J. Lombardero, M. P. Ayers, and J. J. Ruel.2001. Interactions between Þre and bark beetles in an oldgrowth pineforest. For. Ecol. Mgmt. 144: 245Ð254.Schmid, J. M., and D. L. Parker. 1990. Fireand forest insects,pp. 232Ð233. In J. S. Krammes [tech. coord.], Effectsof Þre management of southwestern natural resources.U.S. Dep. Agric. For. Serv. Rocky Mount. For. Rge. Exp.Stn., Ft. Collins, CO.Sanchez-Martinez, G., and M. R. Wagner. 2002. <strong>Bark</strong> beetlecommunity structurein four ponderosa pinestand conditionsin northern Arizona. For. Ecol. Manage. 170: 145Ð160.Saveland, J. M., and L. F. Neuenschwander. 1990. A signaldetection framework to evaluate models of tree mortalityfollowing Þredamage. For. Sci. 36: 66Ð76.(SPSS) Statistical Packages for Social Sciences. 1998. UserÕsmanual, version 8.0. SPSS, Inc., Chicago, IL.Stephens, S. L., and M. A. Finney. 2002. Prescribed Þremortality of Sierra Nevad mixed conifer tree species:

522 ENVIRONMENTAL ENTOMOLOGY Vol. 32, no. 3effects of crown damage and forest ßoor combustion. For.Ecol. Mgmt. 162: 261Ð271.Swets, J. A. 1996. Signal detection theory and ROC analysisin psychology and diagnostics: collected papers. LawerenceErlbaum Associates, Inc. Mahwah, NJ.Thomas, T. L., and J. K. Agee. 1986. Prescribed Þre effectson mixed conifer forest structure at Crater Lake, Oregon.Can. J. For. Res. 16: 1082Ð1087.[USDA] United States Department of Agriculture ForestService. 1997. Plant Associations of Arizona and NewMexico, vol. 1: Forests, 3rd Edition. U.S. Dep. Agric. For.Serv., SW Region, Albuquerque, NM.Van Wagener, C. E. 1973. Height of crown scorch in forestÞres. Can. J. For. Res. 3: 373Ð378.Wagener, W. W. 1961. Guidelines for estimating the survivalof Þre-damaged trees in California. U.S. Dep. Agric.For. Serv. PaciÞc SW Forest and Range Exp. Stn. Misc.Pap. 60, Berkeley, CA.Wallin, K. F., T. E. Kolb, K. R. Skov, and M. R. Wagner. 2003.Effects of crown scorch on ponderosa pine resistance tobark beetles in northern Arizona. Environ. Entomol. 32:652Ð661.Wood, D. L. 1972. Selection and colonization of ponderosapine by bark beetles. Symp. R. l Entomol. Soc. London. 6:101Ð107.Wood, D. L. 1982. The role of pheromones, kairomones, andallomones in the host selection behavior of bark beetles.Annu. Rev. Entomol. 27: 411Ð446.Wyant, J. G., P. N. Omi, and R. D. Laven. 1986. Fireinducedtree mortality in a Colorado ponderosa pine/Douglas-Þrstand. For. Sci. 32: 49Ð59.Received for publication 1 February 2002; accepted 10 January2003.