Spatio-temporal dynamics of an Ips acuminatus outbreak and implications for management

Agricultural and Forest Entomology (2013), 15, 34–42 DOI: 10.1111/j.1461-9563.2012.00589.x

Spatio-temporal dynamics of an Ips acuminatus outbreakand implications for management

Fernanda Colombari∗, Martin L. Schroeder†, Andrea Battisti∗ and Massimo Faccoli∗∗Department of Agronomy Food Natural resources Animals and Environment-DAFNAE, Universita di Padova, Viale dell’Universita, 16-35020

Legnaro, Italy and †Department of Ecology, Swedish University of Agricultural Sciences, PO Box 7044, S-750 07 Uppsala, Sweden

Abstract 1 Understanding spatio-temporal processes of bark beetle infestations is crucial forpredicting beetle behaviour and aiding management decisions aiming to prevent ormitigate tree mortality. We recorded the spatial and temporal distribution of killedtrees during the 5-year period of an Ips acuminatus outbreak.

2 Killed trees were always grouped in well-defined patches (infestation spots). In yearsof high population density, infestation spots were large and aggregated, whereas,in years of low density, infestation spots were small and weakly aggregated orrandomly distributed within the study area.

3 Most trees were killed in the spring by beetles that had hibernated but, in someyears, trees were also killed in the summer by new-generation beetles originatingfrom spring attacks. Spring-killed trees always formed new infestation spots atnew locations (i.e. spot proliferation). By contrast, summer-killed trees alwaysoccurred at the edge of active spots established in the spring, thus resulting in spotgrowth.

4 With regard to management strategies, the results obtained in the present studysuggest that areas located in close proximity to infestations of the previousyear should be prioritized for risk assessment. Because large spots accountfor most of the observed tree mortality, the cut-and-remove method should befocused on these spots as soon as crown discoloration appears in the summer.If applied timely, this strategy will remove the new-generation beetles originatingfrom the spring attacks before they emerge and also reduce the risk of spotgrowth.

Keywords Engraver beetle, monitoring, sanitation cutting, spot growth, spotproliferation, Pinus sylvestris.

Introduction

Many bark beetle species have a very pronounced spatial pat-terning linked to their cooperative attack strategy, resulting ingroups of trees killed (Turchin et al., 1998) commonly beingknown as ‘spots’ because of their appearance during aerialsurveys (Ayres et al., 2011). The new generation of beetlesproduced in such an infestation spot may disperse and initiatea new infestation (spot proliferation) or continue to enlarge theold infestation by attacking nearby trees (spot growth) (Turchinet al., 1998; Safranyk, 2004; Robertson et al., 2007; Sullivan,2011). The spatial patterns of spot distribution may mainly

Correspondence: Fernanda Colombari. Tel.: +39 049 8272807; fax:+39 049 8272810; e-mail: [email protected]

depend on factors such as the occurrence of susceptible hosttrees, beetle density and behaviour (Ayres et al., 2011; Booneet al., 2011), and could be highly heterogeneous through bothspace and time. At low population density, damage is limitedto severely stressed trees scattered within the forest, whereas,during outbreaks, apparently healthy trees may also be killed,generally aggregated into spots within a much larger area ofunaffected forest landscape (Raffa et al., 1993; Birt, 2011). Het-erogeneity of the forest landscape itself might contribute to thepatchiness of damage (Birt, 2011). Nonetheless, the clumpeddistribution of the infested trees and the distance travelledby dispersing beetles, despite homogeneous stand conditions,suggest that biotic interactions and beetle-specific biologicalcharacteristics are involved in regulating the system (Turchin

© 2012 The Royal Entomological Society

Spatio-temporal dynamics of I. acuminatus 35

et al., 1998; Robertson et al., 2007). Understanding landscape-scale spatio-temporal processes of bark beetle infestations isimportant when modelling and/or predicting bark beetle activityand behaviour (Nelson et al., 2003) aiming to develop adequatemanagement strategies to prevent or mitigate beetle damage(Stephen & Taha, 1979; Hain, 1980).

Ips acuminatus (Gyllenhal) (Coleoptera: Curculionidae,Scolytinae) has a wide-ranging distribution across Europe andAsia, covering a large part of the range of its main host tree:Scots pine (Pinus sylvestris L.) (Novak et al., 1976). Treemortality as a result of I. acuminatus historically has beenlow (Bakke, 1968) and limited to weakened trees (Chararas,1962; Rudnew, 1965). Recently, however, it has increasedsignificantly, affecting apparently healthy trees during epi-demics (Romanyk, 1977; Lieutier et al., 1988). For this rea-son, I. acuminatus is now recognized as an important pestof European Scots pine stands (Gregoire & Evans, 2004). Inaddition, it has been associated with the Scots pine declinerecently observed in the Alps (Wermelinger et al., 2008) andthe intensification in frequency and severity of outbreaks hasbeen linked to a number of adaptations to local abiotic andbiotic factors, possibly driven by climate change (Colombariet al., 2012).

An outbreak of I. acuminatus, which was initiated in thesouth-eastern Italian Alps in 2005 (Colombari et al., 2012),provided an opportunity to study the spatial patterns of killedtrees over time. We analyzed a 5-year data set of beetle-killedtrees to test two hypotheses: (i) population density may affectthe number, size and spatial dependence of the infestationspots and (ii) infestation spots occur and continue to enlarge(spot growth) when, at a location, there is a high density ofbeetles. Our results are used to improve the knowledge ofoutbreak dynamics and to provide indications for managementrecommendations to reduce tree mortality.

Materials and methods

Model system

Ips acuminatus has two distinct attack periods in the studyarea. The first attack is conducted in the spring by adults(hereafter referred to as ‘spring adults’) that emerge fromoverwintering sites as soon as the daily maximum hightemperature reaches 14 ◦C (generally in April). The crownsof successfully colonized trees show signs of colonization(i.e. turn red) approximately 2 months after the attack, andshortly thereafter the new generation beetles are mature. Partof these new generation beetles (hereafter referred to as‘summer adults’) may attack new suitable host trees, whereasthe remaining beetles stay in the bark and do not reproduceuntil after hibernation. Irrespective of the generation, adultsare known to overwinter either under the bark of trees wherethey have developed (Bakke, 1968; Romanyk, 1977) or in theforest litter (Francke-Grosmann, 1963; Lekander et al., 1977;Romanyk, 1977), with the latter case being more frequent inthe study area. Killed trees are grouped in well-defined spots(Romanyk, 1977; Lieutier et al., 1988; Colombari et al., 2012).The size of the spots (i.e. the total number of killed trees) mayvary greatly within and among years.

Study area

The study area is located in the Dolomites (south-eastern ItalianAlps) (46◦27′N, 12◦12′E; approximately 1000 m a.s.l.) andconsists of 2100 ha of Scots pine stands, at least 3 km awayfrom the nearest other pine stand. The terrain is mountainous,marked by steeples and pinnacles, with vertical walls and sheercliffs separated by narrow and deep valleys. Mature Scots pinestands, with an age >100 years (mean diameter at breast height22 cm) and with a mean density of approximately 300 stems perha, cover almost 100% of the forested land on south-westernfacing slopes, on shallow soils, generally poor in water andnutrients, underlain by dolomite and limestone bedrock.

Monitoring of infestation spots

Over a 5-year-period (from autumn 2006 to autumn 2010), allgroups of trees infested by I. acuminatus (spots) were recordedin summer (trees colonized in the spring) and autumn (treescolonized in the summer) as soon as trees turned red.

The monitoring procedure was divided into two stages. Inthe first stage, the approximate location, number and sizeof infestation spots were estimated by observing the pineforest with binoculars from the opposite slope of the valley.In the second stage, all infestation spots were visited in aground survey and the centre of each infestation spot wasrecorded using the Global Positioning System (GPS) (accuracyof 3.3 m; eTrex Legend® H; Garmin, Olathe, Kansas). Red-crowned trees in each spot were examined for signs of recentI. acuminatus colonization and counted to determine spot size(i.e. the number of killed trees). Killed trees were defined tobelong to the same infestation spot as long as they were ≤12 mfrom the spot edge (USDA Forest Service, 2003). Infestationspots were grouped by size into four commonly used classes(1–10, 11–25, 26–50 and >50 killed trees) (Ayres et al., 2011)to analyze whether the occurrence of each category varied overthe outbreak period. Mean tree diameter at breast height andheight were assessed within each spot by measuring some felledtrees and visually estimating the size of the others. Diameterestimates were made at intervals of 5 cm. This monitoringprocedure was followed in autumn 2006, in both summer andautumn 2007, and in summer of all the subsequent years,whereas, in autumn 2008, 2009 and 2010, the survey was onlybased on binocular observations of infestation spots from theopposite slope.

GPS coordinates were used to plot spots in a layeredmap using the coordinate system UTM 33N Zone withprojection WGS 1984 and calculate distances between spotsoccurring within the same year and between consecutiveyears (arcgis, version 10; ESRI, Redland, California). Themap of the irregularly shaped area where the infestationsoccurred (20.74 km2) was obtained by merging and smoothingthe different polygons correspondent to the Scots pine foresttypes, gained from an existing regional vegetation/land-use map (Regione del Veneto – Segreteria Regionale per leInfrastrutture – Unita di Progetto per il Sistema InformativoTerritoriale e la Cartografia, Italy). To analyze whether thedistribution of the spots among distance classes changed duringthe outbreak, new spots formed in each year were divided into

© 2012 The Royal Entomological Society, Agricultural and Forest Entomology, 15, 34–42

36 F. Colombari et al.

six classes (100-m classes from 0 to 500 m and over 500 m)based on their distance from the nearest spot formed in theprevious year.

Statistical analysis

All values are reported as the mean ± SE, unless otherwisespecified. An independent factor (year) was tested by a one-wayanalysis of variance for each of five dependent variables (heightand diameter at breast height of killed trees, elevation and sizeof the spots, distances from each new spot of the year t + 1 tothe nearest spot formed in the previous year t). Variables werelog transformed whenever necessary. Regression analysis wasused to test relationships between pairs of variables (size andenlargement of a spot; size of a new spot and size of the nearestold spot). P < 0.05 was considered statistically significant.

A set of spatial statistics tools were used to analyzeand evaluate the within- and between-year spatial pattern ofinfestation spots. For the within-year analyses, the AverageNearest Neighbour tool (Mitchell, 2005) was used to calculatea nearest neighbour index (based on the mean distancefrom each spot to its nearest neighbouring spot) to assesswhether the spot pattern of each year exhibited clustering ordispersion. A confidence level of 95% was selected to rejectthe null hypothesis of complete spatial randomness (CSR). Todescribe the characteristic of the pattern (spot clustering or spotdispersion) over a range of distance scales, the univariate O-ring statistic (Wiegand & Moloney, 2004) was performed overa maximum radius of 1500 m by selecting a cell size of 20 mand a ring width of 100 m. The confidence envelope (5%) wasobtained with 99 randomizations of the null model that assumesno interactions between the points of the pattern (CSR).Hot spot analysis (Nearest Neighbour Hierarchical Clusteringmethod, NNH; Smith & Bruce, 2008) allowed identification of‘hot spot areas’ (i.e. concentrations of spots within a limitedgeographical area) by imposing, as conditions for clustering, aminimum number of five spots and a fixed threshold distanceof 500 m. A spatial autocorrelation tool (Global Moran’s I ;Mitchell, 2005) was used to test, under the null hypothesis ofCSR (confidence level of 95%), whether the spatial distributionof the size of the spots was spatially clustered, dispersed orrandom. The inverse distance method was selected and the scaleof analysis was set by choosing a threshold distance ensuringthat every spot had at least one neighbour.

Between-year nearest neighbour distances (i.e. distancesfrom each new spot of the year t + 1 to the nearest spotformed in the previous year t) were determined for each offour subsequent years by using the Generate Near Table tool(arcgis, version 10). The bivariate O-ring statistic defined thespatial pattern of spots of year t + 1 in relation to the fixeddistribution of spots of the year t over a maximum radius of1500 m (cell size 20 m; ring width 100 m). The confidenceenvelope (5%) was obtained with 99 randomizations of thenull model ‘antecedent condition’ (Wiegand et al., 2006).

Statistical analyses were performed using the software r,version 2.9.1 (R Foundation for Statistical Computing, ViennaAustria) and statistica, version 8 (Statsoft Inc., Tulsa. Okla-homa). Spatial analyses were performed using arcgis, version

Figure 1 Number of Ips acuminatus-killed trees, infestation spots andnumber of killed trees per spot (mean ± SE) in the five study years. In the2006 survey, the killed trees were not divided into spring and summerattacks and thus the bar represents the total number of killed trees forthis year.

10 (Average Nearest Neighbour, Global Moran’s I) and thesoftware programita (O-ring statistic; Wiegand et al., 2006)and crimestat, version 3.3 (NNH; Levine, 2010).

Results

Temporal patterns of tree mortality

In total, 195 infestation spots, including 6415 I. acuminatus-killed trees, were recorded in the years 2006–2010. Duringthe 2006 survey, approximately 500 trees colonized duringthe previous year (2005) were noted. Because no data wererecorded for these trees, they were not included in the analyses.Tree mortality peaked in 2007, decreased sharply over thenext 2 years and increased again in 2010 (Fig. 1). Meandiameter and height of attacked trees were similar for the studyperiod, with the exception of 2009 when attacked trees weresignificantly smaller (F4,190 = 4.52, P < 0.005; F4,190 = 8.63,P < 0.001, for diameter and height, respectively) (Table 1).

Most trees were killed in the spring by adults that hadhibernated (Fig. 1). Such spring attacks always took placein new locations and thus resulted in the formation of newinfestation spots separated by some distance from old spots(i.e. spot proliferation).

Table 1 Mean ± SE diameter at breast height and tree height of the Ipsacuminatus infested trees during the five study years

Year Diameter (cm) Height (m)

2006 22.8 ± 0.8 11.5 ± 0.52007 22.6 ± 0.6 12.6 ± 0.42008 21.6 ± 0.7 11.6 ± 0.42009 17.7 ± 1.4∗ 7.2 ± 0.3∗

2010 19.9 ± 1.3 10.1 ± 0.6

∗Values for 2009 significantly differed from values for all other years(P < 0.05).

© 2012 The Royal Entomological Society, Agricultural and Forest Entomology, 15, 34–42

Spatio-temporal dynamics of I. acuminatus 37

In some years, trees were also killed during the summer bynew generation beetles originating from the spring attacks. Incontrast to the spring attacks, summer attacks always occurredat the edge of spots formed in the spring of the same yearand thus resulted in spot growth (i.e. enlargement of the spots).Summer attacks occurred in 2007 and 2008, although not in2009 and 2010 (no data for 2006). Spot growth was particularlyhigh in summer 2007 when 62% of the spots enlarged andapproximately 30% of the total number of trees was killed inthe summer (Fig. 1). The magnitude of the enlargement wasdirectly proportional to the size of the source spot (R2 = 0.30,F1,31 = 13.26, P < 0.001). No spot belonging to the smallestsize class (1–10 trees killed by the spring beetles) enlargedduring the summer.

The total number of I. acuminatus infestation spots decreasedslightly from 2006 to 2008, strongly in 2009 and then increasedsomewhat in 2010 (Fig. 1). Generally, smaller spots were morefrequent than larger spots (Fig. 2). The only exception wasin 2007, when the largest spots (>50 trees) were the mostcommon class and the smallest spots (1–10 trees) were themost uncommon. At the end of the outbreak (2009 and 2010),the smallest spot class comprised approximately 57% of thetotal number of spots, whereas the largest one comprised lessthan 5%. The percentage formed by the largest spot class ineach year followed the same pattern as the total number of

Figure 2 Percentage of spots in each size class and total number ofinfested trees during the five study years.

killed trees, whereas the opposite was true for the smallest spotclass (Fig. 2). Over the study period, the different size classesrepresented 36.2% (1–10 trees), 30.6% (11–25 trees), 17.6%(26–50 trees) and 15.6% (>50 trees) of all infestation spotsdetected and accounted for 5.6%, 16.0%, 20.3% and 58.1% oftotal tree mortality, respectively.

The size of the infestation spots varied greatly, rangingfrom 1 to 390 killed trees and with a maximum radius ofapproximately 70 m. Single-tree-spots were rare (only fourrecords in the 5 years). The mean spot size peaked in 2007 (65.6trees ± 11.0; F4,190 = 14.83, P < 0.001) and then decreased,remaining more or less constant (14.1 ± 1.8, 11.0 ± 2.0, 13.0 ±3.3 in 2008, 2009 and 2010, respectively) (Fig. 1). No spatialautocorrelation, with respect to spot size, was found in any ofthe 5 years (Moran’s Index near 0; Z-score between −1.96 and+1.96; P > 0.05).

Within-year spatial pattern of infestation spots

Point pattern analyses revealed a strong aggregation amongspots of killed trees only at high population levels whereasthe relationships weakened or disappeared during the popula-tion decline. The average nearest neighbour distance slightlyincreased over the years and was less than the expected meanfor a hypothetical random distribution in each year, with theonly exception of 2009 (Table 2). The O-ring statistic showedthat spots in 2006, 2007 and 2008 exhibited statistically sig-nificant clustering at small scales (0–500 m), although theO-function became progressively weaker and jagged as insuf-ficient spots fell into the different distance classes, but theywere well within the confidence envelope (CSR) at larger scales(>500 m; Fig. 3). In 2009, spots showed repulsion at all scales,whereas, in 2010, there was only one significant aggregationpeak at distances of between 400 and 500 m.

Hot spot areas, revealed by NNH clustering, were first-orderclusters located in approximately the same zones for threeconsecutive years, whereas, in 2010, there was only one hotspot area in a different position (Fig. 3).

Between-year spatial pattern of infestation spots

Eighty-one percent of the new spots appeared within a 500-mradius of the nearest old spot of the previous year, with a higherfrequency in the smaller distance classes (Table 3). The mean

Table 2 Observed and expected mean values of the nearest neighbour distance (distance from each spot to its nearest neighbouring spot) of eachyear in the study site (20.74 km2; with the exception of 2009, when the area considered was 1.19 km2)

Year Observed mean distance Expected mean distance Nearest neighbour ratio Z-score P

2006 188.64 304.31 0.62 −5.44 <0.012007 225.20 312.80 0.71 −3.90 <0.012008 267.68 318.88 0.83 −2.19 0.032009 166.06 150.99 1.01 0.69 0.492010 281.71 496.93 0.56 −3.80 <0.01

The nearest neighbour ratio is calculated as observed/expected mean distance: if <1, the pattern exhibits clustering; if >1, the trend is towarddispersion. The Z-score and P-value are measures of statistical significance and indicate whether or not the null hypothesis should be rejected (spotsare randomly distributed).

© 2012 The Royal Entomological Society, Agricultural and Forest Entomology, 15, 34–42

38 F. Colombari et al.

Figure 3 Maps of the distribution of Ips acuminatus infestation spots (black dots) in the five study years together with graphs showing the univariateO-ring statistic over a maximum radius of 1500 m. In the maps, the grey area represents the area of the Scots pine forest used for the spatial analyses,with the exception of 2009 when the outlier located in the southern part was excluded. Locations of the hot spot areas are indicated by ellipses. Aminimum number of five spots per cluster within a fixed search radius of 500 m was the condition imposed in performing the hot spot area analysis.In the graphs, the grey curved lines indicate the confidence envelope (5%) for the specific null model chosen (complete spatial randomness) andthe horizontal ones indicate the mean intensity of the pattern. The O-ring statistics for the spot pattern in each year are represented by black solidcircles: at a given distance r, points above the confidence envelope indicate clustering (aggregation), points below repulsion and points within indicaterandomness.

Table 3 Distribution of distances between new infestation spots andtheir nearest neighbour spot from the previous year

Distance classes (m)

Year 0–100 100–200 200–300 300–400 400–500 >500

2007 (n = 53) 35.8 24.5 17.0 15.1 1.9 5.72008 (n = 51) 27.5 19.6 9.8 19.6 9.8 13.72009 (n = 14) 42.9 42.9 14.2 0.0 0.0 0.02010 (n = 21) 9.5 0.0 4.8 0.0 9.5 76.2Total (n = 139) 29.5 20.9 12.2 12.9 5.8 18.7

Data are presented as percentages of spots in each distance class foreach of four subsequent years.n, total number of new infestation spots.

distances between new spots and the nearest old ones, for eachpair of years, were of approximately the same magnitude in

2007–2009 (2006–2007: 207.3 ± 25.7; 2007–2008: 299.2 ±39.3; 2008–2009: 123.1 ± 19.7) and increased significantly in2010 (1,942.3 ± 295.6) (F3,135 = 25.88, P < 0.001).

The bivariate O-ring statistic function confirmed the presenceof a positive association between the spot patterns of years 2006and 2007 and the spot patterns of years 2007 and 2008 becausethe results showed that there were significantly more new spotsin the 600-m neighbourhood of an old spot than expected bya random distribution (Fig. 4). The pattern changed in 2009,when the low number of new spots displayed only two mainaggregation peaks within a 300-m radius of the old spots. Spotsfrom 2010 did not show any spatial dependence on the spotpattern of the previous year and followed a random distribution.

No relationships were found between the size of a newspot and the size of the nearest old spot (from the previ-ous year) (2006–2007: R2 = 0.002, F1,51 = 0.10, P = 0.75;2007–2008: R2 = 0.001, F1,49 = 0.06, P = 0.80; 2008–2009:

© 2012 The Royal Entomological Society, Agricultural and Forest Entomology, 15, 34–42

Spatio-temporal dynamics of I. acuminatus 39

Figure 4 Bivariate O-ring statistic (black solid circles) showing, over a maximum radius of 1500 m, whether the spatial pattern of spots of each year(2007–2010) is influenced by the fixed distribution of spots of the previous year. Grey curved lines indicate the confidence envelope (5%) for thespecific null model chosen (antecedent condition) and horizontal ones indicate the mean intensity of the pattern. At a given distance r, points above theconfidence envelope indicate positive spatial association of the two patterns, points below negative association (segregation) and points within indicateno spatial association (independency).

R2 = 0.23, F1,12 = 3.58, P = 0.08; 2009–2010: R2 = 0.02,F1,19 = 0.52, P = 0.48). No changes in spot elevation wereobserved between years (F4,190 = 1.46, P = 0.22).

Discussion

In the present study, we showed that, during the peak yearsof the I. acuminatus outbreak, infestation spots increased insize and were strongly aggregated, whereas, during yearswith lower tree mortality, infestation spots were smallerand randomly distributed. In addition, infestation spots werespatially concentrated in certain areas (‘hot spots’) of the studysite. In some years, attacks by summer adults also contributedto tree mortality. These attacks always occurred at alreadyactive infestation spots and thus resulted in enlargement of spotsinitiated by spring adults. By contrast, trees killed by springadults always resulted in the formation of new infestation spotsin new localities separated by some distance from old spots(i.e. spot proliferation).

The outbreak spread out through 20 km2 of forest and lastedfor 6 years, which represents a period slightly longer thanreported for any earlier I. acuminatus outbreak (Romanyk,1977; Lieutier et al., 1988). The outbreak started in 2005,peaked in 2007 and then declined strongly in 2008, graduallyreturning to low population densities in the last 2 years.Previous studies indicate that a high availability of weakenedtrees is the trigger for outbreaks and leads to a dramatic increase

in the beetle population, which can then successfully exhaustthe defences of healthy trees as well (Romanyk, 1977; Lieutieret al., 1988; Hernandez et al., 2004). In the site used in thepresent study, a possible trigger was the record summer droughtof 2003, although precise data about tree stress are not availableand other interacting factors might also have been involved inexplaining the switch to high population densities.

During the peak years of the outbreak, infestation spots wereclustered, whereas, during the low density years, the meandistance between spots increased and the spatial distributionchanged to random (Fig. 3). Both landscape structure (i.e. thepresence of potentially suitable habitats) (Powers et al., 1999)and the fact that new infestations tended to occur mainlynear the infestations of the previous year (Fig. 4) may havecontributed to the observed clustered distribution of infestationspots during the peak years of the outbreak. By contrast, duringthe declining phase, new spots displayed only jagged (irregular)and weak aggregation (2009) or absence of spatial dependence(2010) in relation to old spots (Fig. 4). We hypothesize that,in 2009, when population densities were lower, many attacksfailed and the beetles were forced to fly longer distances tofind some scattered weakened trees in which they succeeded inreproducing. This hypothesis is in accordance with our resultthat the trees colonized in 2009 were significantly smallercompared with the trees colonized during the other years.In the subsequent year (2010), most of the new spots wereformed in the neighbourhood of these scattered trees. Thismight explain the partly clustered distribution observed in

© 2012 The Royal Entomological Society, Agricultural and Forest Entomology, 15, 34–42

40 F. Colombari et al.

2010 (Fig. 2), the peak in the number of spots of the biggestdistance class (Table 3) and the absence of spatial dependencein relation to old spots (Fig. 4). Our results are in accordancewith studies conducted on I. typographus, showing that, underhigh density conditions, killed trees mainly occur within aradius of 500 m from source populations (Wichmann & Ravn,2001; Jakus et al., 2003; Kautz et al., 2011). Also, in theDouglas-fir beetle Dendroctonus pseudotsugae Hopkins, thebetween-year average nearest neighbour distances drop below600 m at high population density, with a resulting within-yearclustered distribution of infestations (Dodds et al., 2006). Bycontrast, at low density, killed trees are spatially less predictableand the within-year spatial pattern may be dispersed, randomor clustered as the between-year nearest neighbour distancesincrease.

Large size spots (26–50 and >50 killed trees) dominated inthe peak years of the outbreak (Fig. 2), whereas the years withlower tree mortality were characterized by spots of few trees(up to 10), similar to that reported by Dodds et al. (2006) forD. pseudotsugae. Furthermore, the largest spot class accountedfor most of the total number of killed trees (Fig. 3). A positiverelationship between total tree mortality and size of infestationspots has been demonstrated also for I. typographus (Schroeder& Lindelow, 2002) and may be expected as a result of morebeetles being attracted to newly-formed infestation spots athigher population densities.

Similarly to that observed in other systems of moreaggressive beetles (Dendroctonus spp.; Robertson et al., 2007;Ayres et al., 2011; Sullivan, 2011), in I. acuminatus, bothspot growth and spot proliferation guide the spread of theinfestations and the resulting spatial patterns. In 2007, theyear with the largest mean spot size, attacks by the summeradults contributed substantially to tree mortality by enlargingthe existing spots. Interestingly, trees killed by the summeradults were only found at spots already established by the springadults and, to our knowledge, this is the first observation ofsuch a type of behaviour in the genus Ips. Furthermore, theenlargement was directly proportional to the size of the sourcespot and spots of the 1–10 trees class did not grow at all. Thispattern can be explained by the large number of beetles requiredto successfully colonize the neighbouring trees. Below a certaindensity threshold of new generation beetles, spots expand nofurther (Hedden & Billings, 1979; Ayres et al., 2011). ForDendroctonus frontalis Zimmermann, other factors, such ashabitat edges, age and basal area of pine stands, tree defences,abundance of natural enemies, interactions with phoretic mitesand fungi, climate, as well as movement of beetles into andout of spots, have all been shown to influence the rate of spotgrowth (Ayres et al., 2011).

No spot growth was observed between years (i.e. adultscoming from hibernation always formed new spots in thespring, often hundreds of metres away from the spots of theprevious year). This is a different pattern from that observed forI. typographus, which often continues to kill trees adjacent toinfestation spots of the previous year if they are over a certainsize (Hedgren et al., 2003). One explanation contributing tothis difference may be hibernation behaviour. Ips typographusadults hibernate under the bark of the killed trees or in thelitter at the base of the killed trees. Consequently, there is a

high local density of beetles in the spring at larger infestationspots. Most I. acuminatus instead hibernate outside the killedtrees (Colombari et al., 2012), likely spread out in the litter at acertain distance from trees in which they have developed, and,in the spring, there will be no high local density of beetles veryclose to old infested trees. Another explanation could be thatthe pioneer beetles of I. acuminatus are more dependent onweakened trees for initiating a mass attack than pioneer beetlesof the more aggressive bark beetle species I. typographus.

The results obtained in the present study provide somesimple rules of thumb regarding management strategies duringoutbreaks of I. acuminatus, with particular attention to thepossible application of control measures. These consist mainlyof sanitation cutting, a technique used to limit the negativeimpact of ongoing outbreaks in high-value areas by removinginfested and green susceptible trees (Hansen et al., 2010).(i) Most new infestation spots are likely to be in close proximityto previous year infestations and this information could be usedto prioritize regions for risk assessment. (ii) Most of the killedtrees are found in larger infestation spots (>50 trees) althoughthe smaller spots (1–10 trees) are more numerous. Duringthe entire outbreak period, small spots constituted 36% of allspots but accounted for less than 6% of tree mortality, whereaslarge spots constituted only 15% of the spots but accounted forapproximately 60% of the tree mortality (also including attacksby the summer adults). Thus, sanitation cuttings should befocused on larger spots. This will also generally reduce the costper tree for the cuttings. By setting the limit to spots comprising>10 trees, the spots that are at risk of growing during yearswith attacks by the summer adults are eliminated. (iii) Giventhat most trees are killed by the spring adults, surveys fornew infestation spots should be concentrated through a limitedperiod, from June to early August, when trees fade during thedevelopment of the summer generation. (iv) Sanitation cuttinghas to be carried out as soon as discoloration appears and shouldbe extended to the nearby apparently healthy trees; cuttingcarried during the autumn or early winter has proved to beineffective because most of the adults hibernate outside thebark (Colombari et al., 2012). Although the time lag betweenneedle discoloration and beetle emergence may vary greatlywith weather and the physiological condition of the pine trees,it is in any case short (up to 2–3 weeks according to Romanyk,1977). A quick and effective monitoring procedure, possiblywith the help of remote sensing, would be essential for thesuccess of control treatments.

Acknowledgements

This work could not have been accomplished without thecontribution of many individuals. Thorsten Wiegand providedthe software programita, which was used to perform theO-ring statistic. Valerio Finozzi and Edi D’Ambros (VenetoRegion’s Forest Service) helped with the surveys and cuttingdata. Attilio Egitto and Gino Da Col provided support forthe field activities. Tommaso Anfodillo, Fausto Fontanella andRoberto Menardi (Centre of Studies of Alpine Environment‘Lucio Susmel’ – University of Padova – San Vito di Cadore,Italy) assisted with logistics in the field. Edoardo Petrucco

© 2012 The Royal Entomological Society, Agricultural and Forest Entomology, 15, 34–42

Spatio-temporal dynamics of I. acuminatus 41

Toffolo and Lorenzo Marini commented on earlier versionsof the manuscript. Four anonymous reviewers significantlyimproved the quality of the manuscript. The research wassupported by the PRIN project, 200774ENMR (Climatic changeand Italian pine pests: a model study) and by a grant fromthe Veneto Region’s Phytosanitary Service and Forest Service(FITFOREST).

References

Ayres, M.P., Martinson, S.J. & Friedenberg, N.A. (2011) Southern pinebeetle ecology: populations within stands. Southern Pine Beetle (ed.by I. I. R. N. Coulson and K. D. Klepzig), General Technical ReportSRS-140, pp. 75–89. USDA Forest Service, Southern ResearchStation, Asheville, North Carolina.

Bakke, A. (1968) Ecological studies on bark beetles (Coleoptera:Scolytidae) associated with Scots pine (Pinus sylvestris L.) inNorway with particular reference to the influence of temperature.Meddelelser Fra Det Norske Skogforsøksvesen, 21, 443–602.

Birt, A. (2011) Population dynamics of southern pine beetle in forestlandscapes. Southern Pine Beetle II (ed. by R. N. Coulson andK. D. Klepzig), General Technical Report SRS-140, pp. 91–108.USDA Forest Service, Southern Research Station, Asheville, NorthCarolina.

Boone, C.K., Aukema, B.H., Bohlmann, J., Carroll, A.L. & Raffa, K.F.(2011) Efficacy of tree defense physiology varies with bark beetlepopulation density: a basis for positive feedback in eruptive species.Canadian Journal of Forest Research, 41, 1–15.

Chararas, C. (1962) Etude biologique des scolytides des coniferes.Encyclopedie entomologique, Serie A, XXXVIII. Lechevalier, France.

Colombari, F., Battisti, A., Schroeder, L.M. & Faccoli, M. (2012) Life-history traits promoting outbreaks of the pine bark beetle Ipsacuminatus (Coleoptera: Curculionidae, Scolytinae) in the south-eastern Alps. European Journal of Forest Research, 131, 553–561.

Dodds, K.J., Garman, S.L. & Ross, D.W. (2006) Landscape analysesof Douglas-fir beetle populations in northern Idaho. Forest Ecologyand Management, 231, 119–130.

Francke-Grosmann, H. (1963) Die Ubertragung der Pilzflora bei demBorkenkafer Ips acuminatus Gyll: Ein Beitrag zur Kenntnis derIpiden-Symbiosen. Zeitschrift fur Angewandte Entomologie, 52,355–361.

Gregoire, J.C. & Evans, H.F. (2004) Damage and control of BAWBILTorganisms – an overview. Bark and Wood Boring Insects in LivingTrees in Europe, a Synthesis (ed. by F. Lieutier, K. R. Day,A. Battisti, J. P. Gregoire and H. F. Evans), pp. 135–180. Kluwer,The Netherlands.

Hain, F.P. (1980) Sampling and predicting population trends. TheSouthern Pine Beetle (ed. by R. C. Thatcher, J. L. Searcy, J. E.Coster and G. D. Hertel), Technical Bulletin 1631, pp. 71–105.USDA Forest Service, Washington, District of Columbia.

Hansen, E.M., Negron, J.F., Munson, A.S. & Anhold, J.A. (2010) Aretrospective assessment of partial cutting to reduce spruce beetle-caused mortality in the Southern Rocky Mountains. Western Journalof Applied Forestry, 25, 81–87.

Hedden, R.L. & Billings, R.F. (1979) Southern pine beetle: factorsinfluencing the growth and decline of summer infestations in EastTexas. Forest Science, 25, 547–556.

Hedgren, P.O., Schroeder, L.M. & Weslien, J. (2003) Tree killingby Ips typographus (Coleoptera : Scolytidae) at stand edges withand without colonized felled spruce trees. Agricultural and ForestEntomology, 5, 67–74.

Hernandez, R., Perez, V., Sanchez, G., Castella, J. & Palencia, J.(2004) Ensayos de atraccion y captura de Ips acuminatus (Coleo-ptera: Scolytidae). Ecología, 18, 35–52.

Jakus, R., Grodzki, W., Jezik, M. & Jachym, M. (2003) Definitionof spatial patterns of bark beetle Ips typographus (L.) outbreakspreading in Tatra Mountains (Central Europe), using GIS. Pro-ceedings: Ecology, Survey and Management of Forest Insects (ed.by M. L. McManus and A. M. Liebhold), General Technical ReportNE-311, pp. 25–32. Northeastern Research Station, USDA ForestService, Newtown Square, Pennsylvania.

Kautz, M., Dworschak, K., Gruppe, A. & Schopf, R. (2011) Quantify-ing spatio-temporal dispersion of bark beetle infestations in epidemicand non-epidemic conditions. Forest Ecology and Management, 262,598–608.

Lekander, B., Bejer-Petersen, B., Kangas, E. & Bakke, A. (1977)The distribution of bark beetles in the Nordic countries. ActaEntomologica Fennica, 32, 1–36.

Levine, N. (2010) CrimeStat: A Spatial Statistics Program for theAnalysis of Crime Incident Locations, Version 3.3. Ned Levine &Associates, Houston, Texas, and the National Institute of Justice,Washington, District of Columbia.

Lieutier, F., Faure, T. & Garcia, J. (1988) Les attaques de scolytes et ledeperissement du pin sylvestre dans la region Provence Cote d’Azur.Revue Forestiere Francaise, 40, 224–232.

Mitchell, A. (2005) The ESRI Guide to GIS Analysis, Volume 2:Spatial Measurements and Statistics, 252 pp. ESRI Press, Redlands,California.

Nelson, T., Boots, B. & Wulder, M.A. (2003) Spatial-temporal analy-sis of mountain pine beetle infestations to characterize pattern, risk,and spread at the landscape level. Mountain Pine Beetle Sympo-sium: Challenges and Solutions (ed. by T. L. Shore, J. E. Brooksand J. E. Stone), Information Report BC-X-399, 298 pp. NaturalResources Canada, Canadian Forest Service, Pacific Forestry Centre,Canada.

Novak, V., Hrozinka, F. & Stary, B. (1976) Atlas of Insects Harmfulto Forest Trees, Vol. I, 58 pp. Elsevier, The Netherlands.

Powers, J.S., Sollins, P., Harmon, M.E. & Jones, J.A. (1999) Plant–pestinteractions in time and space: a Douglas-fir bark beetle outbreak asa case study. Landscape Ecology, 14, 105–120.

Raffa, K.F., Phillips, T. & Salom, S. (1993) Strategies and mechanismsof host colonization by bark beetles. Interactions Among BarkBeetles, Pathogens, and Conifers in North American Forests (ed. byT. O. Schowalter and G. Filip), pp. 103–128. Academic Press, NewYork, New York.

Robertson, C., Nelson, T.A. & Boots, B. (2007) Mountain pine beetledispersal: the spatial–temporal interaction of infestations. ForestScience, 53, 395–405.

Romanyk, N. (1977) Contribucion al estudio de Ips acuminatus Gyll.en Espana. Metodos de prevencion y combate. Boletin de la EstacionCentral de Ecologia, 11, 49–61.

Rudnew, D.F. (1965) Schutz der Walder vor Schadlingen in derUkraine. Anzeiger fur Schadlingskunde, 38, 130–137.

Safranyk, L. (2004) Mountain pine beetle epidemiology in lodgepolepine. Mountain Pine Beetle Symposium: Challenges and Solutions(ed. by T. L. Shore, J. E. Brooks and J. E. Stone), InformationReport BC-X-399, 298 pp. Natural Resources Canada, CanadianForest Service, Pacific Forestry Centre, Canada.

Schroeder, L.M. & Lindelow, A. (2002) Attacks on living spruce treesby the bark beetle Ips typographus (Col. Scolytidae) followinga storm-felling: a comparison between stands with and withoutremoval of wind-felled trees. Agricultural and Forest Entomology,4, 47–56.

Smith, S.C. & Bruce, C.W. (2008) CrimeStat III User Workbook. TheNational Institute of Justice, Washington, District of Columbia.

Stephen, F.M. & Taha, H.A. (1979) Area-wide estimation of southernpine beetle populations. Environmental Entomology, 8, 850–855.

Sullivan, B.T. (2011) Southern pine beetle behavior and semio-chemistry. Southern Pine Beetle II (ed. by R. N. Coulson and

© 2012 The Royal Entomological Society, Agricultural and Forest Entomology, 15, 34–42

42 F. Colombari et al.

K. D. Klepzig), General Technical Report SRS-140, pp. 25–50.USDA Forest Service, Southern Research Station, Asheville, NorthCarolina.

Turchin, P., Reeve, J.D., Cronin, J.T. & Wilkens, R.T. (1998) Spatialpattern formation in ecological systems: bridging theoretical andempirical approaches. Modelling Spatiotemporal Dynamics in Ecol-ogy (ed. by J. Bascompte and R. V. Sole), pp. 199–213. Springer,Germany.

USDA Forest Service (2003) Southern Pine Beetle Infestation Sup-pression – Environmental Assessment. Southern Region, UwharrieRanger District, Troy, North Carolina.

Wermelinger, B., Rigling, A., Schneider Mathis, D. & Dobbertin, M.(2008) Assessing the role of bark- and wood-boring insects in the

decline of Scots pine (Pinus sylvestris) in the Swiss Rhone valley.Ecological Entomology, 33, 239–249.

Wichmann, L. & Ravn, H.P. (2001) The spread of Ips typographus L.(Col., Scolytidae) attacks following heavy windthrow in Denmark,analysed using GIS. Forest Ecology and Management, 148, 31–39.

Wiegand, T. & Moloney, K.A. (2004) Rings, circles, and null-modelsfor point pattern analysis in ecology. Oikos, 104, 209–229.

Wiegand, T., Kissling, W.D., Cipriotti, P.A. & Aguiar, M.R. (2006)Extending point pattern analysis to objects of finite size and irregularshape. Journal of Ecology, 94, 825–837.

Accepted 16 June 2012First published online 17 October 2012

© 2012 The Royal Entomological Society, Agricultural and Forest Entomology, 15, 34–42