BIOLOGICAL CONTROL-PARASITOIDS AND PREDATORS Establishment of the Predator, nigrinus (Coleoptera: ), in the Eastern United States

1 D. L. MAUSEL, S. M. SALOM, L. T. KOK, AND G. A. DAVIS

Department of Entomology, Virginia Tech, 216A Price Hall MC 0319, Blacksburg, VA 24061-0002

Environ. Entomol. 39(2): 440Ð448 (2010); DOI: 10.1603/EN09088 ABSTRACT The hemlock woolly adelgid, Adelges tsugae Annand (Hemiptera: Adelgidae), native to western North America and Asia, was accidentally introduced from Japan to the eastern United States. To potentially establish biological control of A. tsugae, we released a predator endemic to western North America, Laricobius nigrinus Fender (Coleoptera: Derodontidae), from 2003 to 2005, in 22 localities from Georgia to Massachusetts. Release sites spanned the invasive range of the adelgid across Þve United States Department of Agriculture plant hardiness zones (5a to 7a). Release sizes were 75, 150, 300, 600, or 1,200 adult L. nigrinus per site in the fall, winter, early spring, or sequentially (i.e., fall or winter and early spring). We monitored establishment by annual sampling for L. nigrinus adults with beat sheets and for L. nigrinus larvae by branch clipping. At the end of 3 yr, L. nigrinus was established in 13 of the 22 sites. The following variables were evaluated for their correlation with the numbers of L. nigrinus larvae and adults recovered and for their effect on establishment (scored as

F3 presence/absence): (1) Minimum winter temperature at the release site, (2) A. tsugae density at the time of release, (3) release size, and (4) release season. Only minimum winter temperature was correlated with larval recoveries and no variables were correlated with adult recoveries. Logistic regression modeling found that establishment was positively related to minimum winter temperature and release size. We recommend smaller release sizes in warm areas where establishment probability was high (i.e., zones 7a, 6b, and 6a) and larger release sizes in cold areas where establishment probability was low (i.e., zones 5b and 5a). Releases during fallÐearly spring and across the range of A. tsugae densities tested were successful.

KEY WORDS , Adelges tsugae, Laricobius nigrinus, release strategy, establishment

To improve the establishment rate of classical biolog- mott et al. (1998) discovered that releases of Ͻ100 ical control agents, better understanding is needed of gorse thrips, Sericothrips staphylinus (Haliday) (Thys- variables that affect colonization (Memmott et al. anoptera: Thripidae), on gorse were superior to 1,000, 1998). Climate in particular is recognized frequently the size formerly advocated. as an important variable, whose inßuence can be pre- Classical biological control is the option of primary dicted through the use of climate-matching methods interest for suppression of the hemlock woolly adel- (Sutherst et al. 2004). An issue of practical importance gid, Adelges tsugae Annand (Hemiptera: Adelgidae), is the optimal number of individuals to release per site. in native eastern hemlock, Tsuga canadensis (L.) Since the relative importance of demographic stochas- Carrie`re, and Carolina hemlock, T. caroliniana En- ticity (e.g., sex ratio), environmental variability (e.g., gelmann, forests of the eastern United States (Cheah temperature), and Allee effects (e.g., mate Þnding in et al. 2004). The adelgid is native to northwestern small populations) are typically unclear during ex- North America and Asia and was discovered in Rich- ploratory releases, a mixed release strategy has been mond, VA, in the 1950s after accidental introduction advocated (Shea and Possingham 2000). A mixed strat- from Japan (Havill et al. 2006). It has spread widely egy is one using a structured set of different release (Anonymous 2009) and caused extensive mortality of approaches, which can facilitate a better understand- hemlock trees (Orwig and Foster 1998, Eschtruth et al. ing of variables that inßuence establishment and im- 2006). Methods exist for monitoring A. tsugae (Costa prove efÞciency and success of subsequent releases and Onken 2006, Fidgen et al. 2006), chemically con- (Campbell 1976, Grevstad 1999). For example, Mem- trolling it (Steward and Horner 1994), and ameliorat- ing hemlock forest decline (Ward et al. 2004). Unfor- tunately, there does not appear to be any effective 1 Current address: Department of Plant, Soil, and Sciences, natural enemies of A. tsugae in eastern United States University of Massachusetts, Agricultural Engineering Bldg., 250 Natural Resources Road, Amherst, MA 01003-9295 (e-mail: dmausel@ forests (Wallace and Hain 2000). However, Laricobius psis.umass.edu). rubidus LeConte (Coleoptera: Derodontidae), an ex-

0046-225X/10/0440Ð0448$04.00/0 ᭧ 2010 Entomological Society of America April 2010 MAUSEL ET AL.: ESTABLISHMENT OF Laricobius nigrinus 441 tant predator of the native pine bark adelgid, Pineus Effective biological control agents typically estab- strobi (Hartig), has recently been discovered com- lish easily and show some impact within three gener- pleting development on A. tsugae (Zilahi-Balogh et al. ations (Clausen 1951). Those that do not establish 2005, Mausel et al. 2008). Foreign explorations in Asia easily or show impact within this period often have and northwestern North America have identiÞed sev- little chance of success, but exceptions occur (Debach eral predators as potential biological control agents of and Rosen 1991). Centered on this theory, our objec- A. tsugae (Sasaji and McClure 1997, Yu et al. 2000, Yu tive was to use a mixed release strategy and determine and Montgomery 2007, Zilahi-Balogh et al. 2007, how well adult L. nigrinus releases in A. tsugae-infested Kohler et al. 2008). Three coccinellids from Asia and eastern hemlock forests established based on 3 yr of Laricobius nigrinus Fender from northwestern North systematic postrelease monitoring. To potentially in- America have been released (Cheah et al. 2004). crease the success rate of subsequent releases, the Surveys of western hemlock, Tsuga heterophylla following variables were evaluated for their associa- (Raf.) Sargent, have shown that L. nigrinus is wide- tion with the numbers of L. nigrinus larvae and adults spread and feeds on A. tsugae, at both low and high recovered, and for their effects on establishment density infestations in northwestern North America (scored as F3 presence/absence): Minimum winter (Kohler et al. 2008). Adults and larvae feed on A. temperature at the release site, A. tsugae density at the tsugae eggs, nymphs, and adults at the base of needles. time of release, release size, and release season. L. nigrinus eggs are laid in late winterÐearly spring in sistens ovisacs (summerÐearly spring A. tsugae gener- Materials and Methods ation), which contain progrediens eggs (spring A. tsugae generation). Larvae develop through four in- Selection of Release Sites and Trees. Releases were stars feeding on A. tsugae eggs and fourth instars drop made in eight states throughout the invasive range of A. tsugae in the eastern United States (Table 1; Fig. 1). to the forest ßoor to pupate. After eclosion, adults Ͼ undergo aestival diapause in the soil during summer Stands of 10 ha with a hemlock component, where and emerge in fall, synchronized with sistens Þrst A. tsugae density was low to moderate (1Ð75% of instars breaking aestival diapause. They feed on de- shoots infested), and crown health was not severely veloping adelgid nymphs and adults in fallÐwinter compromised (i.e., off color, premature needle loss, and branch tip dieback) were selected. Individual (Zilahi-Balogh et al. 2003). trees that L. nigrinus were released on had high live Because L. nigrinus feeds, oviposits, and undergoes crown ratios (Ͼ95%) to enable access to branches larval development during fall, winter, and early from the ground for beat sheet sampling. spring, cold temperature is likely an important abiotic Release Procedures. The L. nigrinus adults released mortality factor during the critical establishment were laboratory-reared from parents collected from phase when releases occur in the eastern United an A. tsugae-infested western hemlock seed orchard States. Survival of A. tsugae itself is positively related near Victoria, BC, Canada (48.48 N, 123.36 W) in late to minimum winter temperature (Trotter and Shields winter 2003. They were reared using established 2009). Winters are mild in the marine west coast cli- methods (Lamb et al. 2005b). Reared adults emerging mate of Victoria, BC, Canada (i.e., the source location after aestival diapause from soil Þlled containers in of L. nigrinus in this study) but are considerably colder fall 2003Ð2005. (i.e., F1 laboratory generation) were in many areas of A. tsugaeÕs invasive range in the placed in 3.8-liter containers in groups of up to 50 on eastern United States. Summers are also cooler and A. tsugae-infested eastern hemlock. Depending on the dryer in Victoria, BC; however, L. nigrinus is dormant season, adults were maintained at 2Ð10ЊC, seasonally in earthen soil cells at this time and unlikely to be adjusted photoperiods, and 70Ð90% RH in environ- affected by potentially higher soil temperatures and mental chambers before release. They were screened moisture. for missing appendages, poor-color, or small size. L. nigrinus was imported into the eastern United Three sub-samples of adults were examined for ento- States from Victoria, BC, for evaluation and was de- mopathogens before release by Dr. L. Solter, IL Nat- termined to be host-speciÞc (Zilahi-Balogh et al. ural History Survey; Dr. A. Lawrence, Mississippi State 2002). Subsequent to Federal and State approval for University; and Dr. J. Becnel, USDA-ARS Gainesville environmental release of L. nigrinus, several Þeld stud- and all results were negative. ies were carried out in southwest Virginia. Adults L. nigrinus adults were given fresh A. tsugae every 2 demonstrated high winter survival in sleeve cages, wk until they were released. A cluster of 10Ð15 A. signiÞcant feeding, and oviposition on A. tsugae (Lamb tsugae-infested eastern hemlock twigs (10 cm length) et al. 2005a). The Þrst release of L. nigrinus made in was inserted into wet Oasis ßoral foam (Smithers- early spring 2003 used an egg release strategy and F2 Oasis Co., Kent, OH) wrapped in paraÞlm. Care was adults were recovered in fall 2005 (Lamb et al. 2006). taken to ensure that the twigs did not harbor Fiorinia Adults released on planted A. tsugae-infested hem- externa Ferris (Hemiptera: Diaspididae), an invasive locks established, exhibited phenological synchrony scale insect. Twigs were placed in ventilated, escape- with A. tsugae, and signiÞcantly reduced the pestÕs proof plastic containers (950 ml) with wet Þlter paper. density (Mausel et al. 2008). Laboratory mass-rearing Various numbers of adult L. nigrinus (30, 35, 40, or 80) methods were developed to produce L. nigrinus adults were put in each container for release on individual for releases (Lamb et al. 2005b). trees. Larger containers (3.8 liters) and 20 cm long 442 ENVIRONMENTAL ENTOMOLOGY Vol. 39, no. 2

Table 1. Summary of adult L. nigrinus releases in 22 A. tsugae-infested eastern hemlock forests between Nov. 2003 and Mar. 2005 and postrelease recoveries of L. nigrinus (Ln) and the native L. rubidus (Lr)

Release variables Postrelease recoveries Site No. adults No. larvae (Name, state) A. tsugae Total Season Ln Ln Lr Ln resultd Min. Hard. den.c no. Ln temp.a zoneb Date(s) Fall Win. Spr. F1 F2 F3 F2 F3 Yr 1 Yr 2 Mt. Tom, MA Ϫ25.5 5b 40.0 150 19 Nov 04 150 Ñ Ñ 0 0 0 0 0 0 0 Not established Rothrock, PA Ϫ18.3 6a 19.5 600 4 Dec 03 Ñ 300 Ñ 0 0 8 30 343 7 61 Established 20 Apr 04 Ñ Ñ 300 Bear Run, PA Ϫ18.3 6a 16.9 300 24 Mar 05 Ñ Ñ 300 0 1 18 125 311 9 23 Established Finzel, MD Ϫ21.7 6a 42.3 300 18 Dec 03 Ñ 300 Ñ 0 0 0 0 0 0 0 Not established Rocky, MD Ϫ16.7 6b 38.4 1,200 23 Nov 04 1,200 Ñ Ñ 31 49 37 517 440 64 54 Established Frederick, MD Ϫ18.9 6b 18.2 75 23 Nov 04 75 Ñ Ñ 0 1 15 1 14 1 5 Established Seneca, WV Ϫ22.8 5a 18.8 600 28 Nov 03 300 Ñ Ñ 0 0 0 0 0 10 158 Not established 10 Mar 04 Ñ Ñ 300 Watoga, WV Ϫ22.8 5a 18.9 300 28 Nov 03 300 Ñ Ñ 0 0 0 0 0 16 114 Not established Mon, WV Ϫ22.8 5a 19.0 300 10 Mar 04 Ñ Ñ 300 0 0 0 0 0 0 7 Not established North, VA Ϫ23.3 5b 43.1 600 8 Dec 03 Ñ 300 Ñ 0 0 0 23 32 5 7 Established 28 Mar 04 Ñ Ñ 300 Big Stony, VA Ϫ23.3 5b 23.7 300 8 Dec 03 Ñ 300 Ñ 0 0 0 0 0 2 44 Not established Lick, VA Ϫ21.1 5b 34.6 150 4 Nov 04 150 Ñ Ñ 1 0 0 5 0 5 0 Not established Highland, VA Ϫ18.3 6a 58.0 1,200 4 Nov 04 1,200 Ñ Ñ 0 0 9 0 62 6 0 Established Hurricane, VA Ϫ18.3 6a 41.7 300 30 Mar 03 Ñ Ñ 300 1 0 1 19 1 14 0 Established Dickey, VA Ϫ16.1 6a 61.8 75 8 Feb 05 Ñ 75 Ñ 0 0 0 0 0 52 27 Not established Hem Hill, NC Ϫ16.7 6a 35.6 300 31 Dec 03 Ñ 300 Ñ 3 13 93 10 314 0 0 Established Holloway, NC Ϫ16.7 6a 32.9 150 27 Oct 04 150 Ñ Ñ 1 9 9 2 2 24 20 Established Ivy, NC Ϫ13.9 6b 58.1 1,200 14 Mar 05 Ñ Ñ 1,200 0 1 0 42 868 0 0 Established Locust, NC Ϫ15.5 6b 39.9 75 12 Jan 05 Ñ 75 Ñ 0 0 3 42 109 42 109 Established Middle, NC Ϫ15.5 6b 48.2 600 12 Jan 05 Ñ 300 Ñ 0 0 8 180 1,163 12 74 Established 13 Mar 05 Ñ Ñ 300 Laurel, TN Ϫ14.4 6b 23.6 300 17 Feb 04 Ñ Ñ 300 0 0 0 118 31 31 67 Established Overßow, GA Ϫ12.2 7a 41.1 150 26 Oct 04 150 Ñ Ñ 0 0 0 0 0 0 0 Not established

a Absolute min. winter temp was recorded over 3 yr at a weather station within 20 km from each site. b Hardiness zones are based on avg. ann. min. temp (ЊC): 5a (Ϫ26.2 to Ϫ28.8), 5b (Ϫ23.4 to Ϫ26.1), 6a (Ϫ20.6 to Ϫ23.3), 6b (Ϫ17.8 to Ϫ20.5), and 7a (Ϫ15.0 to Ϫ17.7) (Cathey 1990). c A. tsugae density was calculated as the % of new hemlock shoots infested with at least one sistens A. tsugae (Fidgen et al. 2006). d Establishment deÞned as recovery of F3 adults or larvae. twigs were used to hold 80 predators. The containers Determination of A. tsugae Density on Release with predators were hand-carried, kept at ambient Trees. We determined the percentage of shoots in- temperature, and shaded before release, typically fested on each release tree at the time of release as a within 24 h. Clusters of twigs containing predators proxy for sistens absolute population density (Fidgen were tied to the most heavily A. tsugae-infested branch et al. 2006). One branch was selected from each car- on each release tree with surveyorÕs ßagging in the dinal point and a 30 cm section of each branch (30Ð60 morning. Releases were made from 2003 to 2005. cm from the distal shoot tip toward the trunk of the Adults were released at selected sites within the A. tree) was demarcated with plastic cable-ties. Within tsugae-infested portion of eastern hemlockÕs range to this demarcated area, the total number of new shoots explore the effects of climatic variation on establish- (i.e., produced the previous growing season) and the ment (Fig. 1). Release sizes were 75 (n ϭ 3), 150 (n ϭ number that were infested with at least one adelgid 4), 300 (n ϭ 8), 600 (two sequential releases of 300, n ϭ were counted in winter. The average percentage of 4), and 1,200 (n ϭ 3) predators per site. The number shoots infested was then calculated for each tree at a released per tree at each locality varied as follows: (1) site. 75 adults, each of two trees received 35 or 40 predators; Determination of Establishment. Standardized (2) 150 adults, 30 per tree released on Þve trees; (3) sampling to detect L. nigrinus life stages using two 300 adults, 30 per tree released on 10 trees; (4) 600 different methods (beat sheet for adults and branch adults, 60 per tree released on 10 trees (as a sequential clipping for larvae) started the year after release and release of 30 per tree in fall or winter and early spring), continued for 3 yr at each site. Establishment was and; (5) 1,200 adults, 80 per tree released on 15 trees. considered successful if F3 adults or F3 larvae were Season of release was fall, winter, early spring, or recovered. sequential. During fall releases (Oct. to Nov., n ϭ 8), During fall 2004Ð2007, beat sheet sampling was con- sistens A. tsugae stages were developing early instars. ducted six times a year from Ϸmid-September through For winter releases (Dec. to Feb., n ϭ 5) they were December to collect L. nigrinus adults using canvas late instars and adults, and for early spring releases (71 cm2) beat sheets (Bioquip, Rancho Dominguez, (Mar. to Apr., n ϭ 5) they were sistens adults and CA). Every 2-wk the lower canopy of the release trees progrediens eggs. Sequential releases consisted of re- and a minimum of 10 neighboring nonrelease trees leases in both fall (or winter) and early spring (n ϭ 4). were sampled per visit. Sampling was timed to avoid April 2010 MAUSEL ET AL.: ESTABLISHMENT OF Laricobius nigrinus 443

Fig. 1. Locations of L. nigrinus release sites in 22 A. tsugae-infested eastern hemlock forests between November 2003 and March 2005 superimposed on a scanned section of the USDA plant hardiness zone map (Cathey 1990) (ϩ, established; Ϫ, not established). periods with temperatures below 0ЊC, when adults are moderately-to-heavily infested (25Ð100% shoots in- not on branches (Zilahi-Balogh et al. 2003), moderate- fested) or consisted of mainly dead A. tsugae without to-heavy winds (Ͼ10 km/h), or rain or snow. Col- ovisacs, then another branch was cut from the same lected adults were identiÞed in the Þeld using L. ni- tree or from the nearest neighboring tree. If dense A. grinus reference specimens in 4ϫ magniÞer boxes tsugae ovisacs could not be recovered from these al- (Bioquip) and returned to the trees. The sample date ternative sources, sampling stopped, and the lack of a and total number of L. nigrinus recovered were re- suitable sample was noted. The branches from a tree corded. Reference specimens from each site were were pooled and placed in a labeled plastic bag. Ap- examined for species conÞrmation and placed in the proximately 7,000 cm3 of foliage were collected from Virginia Tech, Entomology Department Collection, each tree and samples were transported to the labo- Blacksburg, VA. L. rubidus adults also were collected ratory. Each sample was placed in a modiÞed Berlese but were rare and not recorded. funnel to rear L. nigrinus and L. rubidus larvae devel- During early spring 2005Ð2007, branch clipping oping on A. tsugae eggs (Lamb et al. 2005b). Mature sampling was conducted annually to collect L. nigrinus larvae that dropped into mason jars were counted eggs and larvae in sistens ovisacs. We timed sampling twice daily, pooled by site, and transferred to soil with peak A. tsugae egg abundance. Branch clipping containers for pupation and aestivation. Adults that was conducted at each site between March and May, emerged from the soil in the fall were identiÞed as L. when woolly ovisacs were 2Ð3 mm in diameter, indi- nigrinus or L. rubidus, counted by site, and the species cating peak egg abundance. SpeciÞc sampling dates ratio calculated to estimate the number of larvae col- were chosen based on elevation and latitude. In the lected of each species. L. nigrinus adults were distin- Þrst year of sampling, we noticed that ßowering east- guished from those of L. rubidus by their uniformly ern redbud, Cercis canadensis L. (Fabaceae), is a good black color (Lawrence and Hlavac 1979). A sub-sam- phenological indicator of peak progrediens egg abun- ple of larvae was preserved in 95% ethanol and if adults dance and used it to initiate sampling in subsequent failed to emerge, the larval sub-samples were identi- years. Two 0.5Ð1.0 m long branches from the canopy Þed by a polymerase chain reaction, restriction frag- of each release tree were cut with a 1.8 m Þberglass ment length polymorphism diagnostic assay using the pole-pruner and up to three 1.8 m attachments (Fred CO1 gene (Havill et al. 2010). Marvin Assoc., Akron, OH). Branches up to 10 m Statistical Methods. Simple correlation analyses above ground were clipped from all the sample trees. (Nonparametric SpearmanÕs rank correlation) were The branches were then cut down to 20Ð30 cm length used to measure the intensity of association between twigs with a hand-pruning shear. If a branch was not the numbers of L. nigrinus larvae and adults recovered 444 ENVIRONMENTAL ENTOMOLOGY Vol. 39, no. 2 and each independent variable: Minimum winter tem- Table 2. Spearman’s rank correlations between the numbers perature, A. tsugae density, release size, and release of F3 L. nigrinus recovered and min. winter temp, A. tsugae density (22 ؍ at the time of release, release size, and release season (n season using SPSS version 16.0. Larval numbers were analyzed as the total number of F3 larvae collected No. larvae No. adults divided by the number of release trees sampled per Correlation Correlation P value P value site. The total numbers of F3 adults recovered were coefÞcient coefÞcient analyzed. Minimum temp 0.471 0.027 0.254 0.253 To accommodate the multivariate nature of our A. tsugae density 0.088 0.696 Ϫ0.217 0.333 releases, a stepwise binary logistic regression (Back- Release size 0.348 0.112 0.070 0.755 ward likelihood-ratio method) was used to model the Release season 0.246 0.269 Ϫ0.228 0.309 probability that varying combinations of the different independent variables (minimum winter tempera- ture, A. tsugae density, release size, and season) con- F to F generation). Five (36%) had stable or declin- tributed signiÞcantly to the response variable values 2 3 ing larval numbers. Seventeen (77%) of the 22 sites (F adult or larval L. nigrinus presence/absence). 3 also yielded recoveries of L. rubidus larvae. Of 1,114 Variables that had no effect at P ϭ 0.10 were excluded Laricobius spp. larvae collected and reared to adults in from the Þnal regression model. For locality analyses, 2005, 25% were L. rubidus and 75% were L. nigrinus the absolute minimum winter temperatures during the (i.e., F generation). In 2006, of 3,690 larvae collected, 3 yr postrelease at each site were recorded from the 2 34% were L. rubidus and 66% were L. nigrinus (i.e., F nearest weather station, within 20 km. Weather data 3 generation). At the three WV sites, Big Stony, VA, and were acquired through the Southeast Regional Cli- Dickey, VA, all larvae collected were L. rubidus while mate Center at North Carolina State University at Hem Hill and Ivy, NC, all larvae were L. nigrinus. (http://www.sercc.com/). The season of release was Typically, both species coexisted and L. nigrinus was modeled categorically with dummy variables (fall ϭ 1, regularly more abundant than L. rubidus on eastern winter ϭ 2, early spring ϭ 3, or sequential ϭ 4). Before hemlock. Overall, L. nigrinus was established in 13 of analysis, a preliminary multiple linear regression was the 22 sites (Fig. 1). conducted to identify outliers (MahalanobisÕ dis- Variables Affecting Recovery Numbers and Estab- tance) and to evaluate multicollinearity (Tolerance lishment. With respect to the power of the indepen- statistics) among the four independent variables. Po- dent variables explored (minimum winter tempera- tential outliers were evaluated by determining which ture, A. tsugae density, release size, and season), cases exceeded the ␹2 criteria of ␹2 (4) ϭ 18.5 at P ϭ simple correlation indicated that no variables were 0.001. Logistic regression model responses were re- correlated with adult recoveries. The number of L. vealed by graphing model outputs across the range of nigrinus larvae recovered was positively correlated values for the predictor variables. with minimum winter temperature (Table 2). To expedite Þeld selection of future release sites To predict establishment of L. nigrinus (i.e., F adult based on minimum winter temperature, each release 3 or larval presence/absence), the stepwise (backwards site in this study was categorized into United States likelihood-ratio method) logistic regression showed Department of Agriculture (USDA) plant hardiness that a model with two variables (minimum winter zones (Cathey 1990) (Table 1). The “hardiness zone” temperature and release size) was the most parsimo- map was created from average annual minimum win- nious (Table 3). ter temperatures between 1974 and 1986 at 8,000 North American weather stations. Our L. nigrinus re- 1 leases were made in zones 5a (Ϫ26.2 to Ϫ28.8ЊC; n ϭ P ϭ Ϫ͓ ϩ ϩ ͔ 1 ϩ e 6.709 (0.428)(Minimum temp.) (0.006)(Release size) 3), 5b (Ϫ23.4 to Ϫ26.1ЊC; n ϭ 4), 6a (Ϫ20.6 to Ϫ23.3ЊC, n ϭ 8), 6b (Ϫ17.8 to Ϫ20.5ЊC; n ϭ 6), and 7a (Ϫ15.0 A. tsugae density and release season had no effect on to Ϫ17.7ЊC; n ϭ 1). the probability of establishment. The model was sta- tistically reliable in distinguishing between estab- lished and not established L. nigrinus populations and Results made predictions that were not signiÞcantly different

Determination of Establishment. F1,F2, and F3 L. from the observed data (Hosmer and Lemeshow test, nigrinus adults were recovered with beat sheets at Þve ␹2 (8) ϭ 7.6, P ϭ 0.47). Overall, the Þnal model cor- (23%), six (27%), and 10 (45%) of the 22 sites, respec- rectly classiÞed 82% of the cases, while the null model, tively (Table 1). The numbers of adults recovered assuming all releases were successful, only correctly increased over time (i.e., F1 to F3 generation) at nine classiÞed 59% of the cases. Tolerance statistics for the (75%) of the 12 total sites with recoveries. Three independent variables were Ͼ0.1 and multicollinear- (25%) sites had stable or declining adult recoveries, ity was not a problem. No outliers were detected. and four (33%) sites had no recoveries until the third The probability of establishment was positively re- year after release. lated to minimum winter temperature and release size

Both F2 and F3 L. nigrinus larvae were recovered by as indicated by their positive coefÞcients (Table 3). branch clipping at 13 (59%) of the 22 sites (Table 1). The models response across a range of minimum tem- The numbers of larvae recovered increased over time peratures for our release sizes shows the inßuence of at nine (64%) of the 14 total sites with recoveries (i.e., both variables in the probability of observing L. nig- April 2010 MAUSEL ET AL.: ESTABLISHMENT OF Laricobius nigrinus 445

Table 3. Variables, their corresponding parameter estimates, indicate that L. nigrinus has become the Þrst potential and likelihood-ratio tests for elimination of variables from the full biological control agent of A. tsugae in the eastern model and simpler models from the stepwise (backwards likelihood- ratio method) logistic regression predicting L. nigrinus establish- United States to become widely established and re- ment (i.e., F3 adult or larval presence/absence) in the eastern covered in substantial numbers. Sites where numbers -declined likely represent populations that are dispers (22 ؍ United States (n ing from the release trees because of declining tree Change in health and lower A. tsugae density as noted during Predictor variable ␤ SE Ϫ2 log df P value likelihood sampling. For sites where we report L. nigrinus as not established the predators might be recovered in the Step 1 Minimum temp 0.494 0.242 6.668 1 0.01 future. Nonrecovery does not necessarily mean ab- A. tsugae density Ϫ0.035 0.057 0.385 1 0.53 sence of L. nigrinus in the locality, but may indicate Release size 0.004 0.004 2.431 1 0.12 that it was not in sufÞcient numbers for detection. For Release season 0.675 3 0.88 example, several predators released for control of bal- Release season (1) Ϫ1.214 2.222 Release season (2) Ϫ0.966 2.398 sam woolly adelgid, Adelges piceae (Ratzeburg), that Release season (3) 0.131 2.186 were considered not to have established, appeared Constant 10.198 6.236 many years after release (Humble 1994). Although Step 2 possible, this is improbable for L. nigrinus because the Minimum temp 0.489 0.233 7.292 1 0.01 A. tsugae density Ϫ0.039 0.048 0.703 1 0.40 branch clipping sampling used in our study appears to Release size 0.006 0.004 4.930 1 0.03 be a reliable sampling method. Branch clipping unex- Constant 9.182 4.968 pectedly collected numerous L. rubidus, a predator Step 3 previously collected only in low numbers from A. Minimum temp 0.428 0.210 6.596 1 0.01 Release size 0.006 0.004 4.421 1 0.03 tsugae-infested hemlock (Montgomery and Lyon Constant 6.709 3.633 1996, Wallace and Hain 2000). This inconsistency is likely because of different sampling methods and per- haps L. rubidus is adapting to A. tsugae (Mausel et al. rinus establishment (Fig. 2). At a given minimum tem- 2008). perature, the probability of establishment increased Variables Affecting Recovery Numbers and Estab- with larger release sizes. This was most apparent at lishment. The numbers of adult L. nigrinus recovered colder temperatures, while at warmer temperatures were not correlated with any release variables likely even the smallest release sizes had high probabilities because of the inherent difÞculty in collecting accu- of establishment. Within each release size, the prob- rate numbers with beat sheets. Adults are small, their ability of establishment decreased with colder mini- behavior is not well understood, and beat sheet sam- mum temperatures. This was most apparent for the pling was limited to the lower canopy of trees. The 75Ð600 release sizes, while the 1,200 release size had numbers of F3 larvae recovered were related to min- a high probability of establishment regardless of tem- imum winter temperature, but not to A. tsugae density, perature. release size, or release season, which suggests that these three variables do not affect the initial increase Discussion in L. nigrinus numbers. Establishment was affected by both minimum temperature (i.e., climate) and release Determination of Establishment. Recovery of the size, two important factors in the colonization of clas- F3 generation and increasing numbers at most sites sical biological control agents. Because the release sites spanned a large geograph- ical area in this study, the inßuence of climate on L. nigrinus was apparent. Climate matching of collection and release areas has long been recognized as a re- quirement in the establishment and impact of natural enemies in classical biological control projects (Pschorn-Walcher 1977, Clausen 1978). One aspect of climate, minimum winter temperature, is likely an important abiotic mortality factor for L. nigrinus be- cause it is released and active during fall, winter, and early spring. L. nigrinus is intolerant of freezing and adult super-cooling points range from Ϫ16 to Ϫ19ЊC (Humble and Mavin 2005). Eggs have supercooling points of Ϫ27.5 to Ϫ26.9ЊC and the larval supercooling points diminish from Ϫ22.1, Ϫ17.0, Ϫ15.0, to Ϫ13.0ЊC for each consecutive instar as they develop later in the spring. Many of these physiological limits for L. nig- Fig. 2. Logistic model response for estimating probability rinus had been exceeded at sites where establishment of observing adult L. nigrinus establishment in A. tsugae- failed. Even if L. nigrinus was not killed by freezing at infested forests as a function of minimum winter tempera- these failed sites, cold stress above the supercooling ture and release size (75Ð1200 per site). point could have played a role in mortality (Humble 446 ENVIRONMENTAL ENTOMOLOGY Vol. 39, no. 2 and Mavin 2005). The L. nigrinus source location in each release would have a 0.68 probability of estab- Victoria, BC, Canada was in hardiness zone 8b (Ϫ6.7 lishment, which would produce more established pop- to Ϫ9.4ЊC), and this population appears to not be well ulations than a single release of 1,200. The probability adapted to zones 5b (Ϫ23.4 to Ϫ26.1ЊC) and zone 5a of L. nigrinus establishment is low in cold areas (i.e., (Ϫ26.2 to Ϫ28.8ЊC) in the eastern United States. Sim- zones 5b and 5a) and larger release sizes are recom- ilarly, A. tsugae survival is positively related to mini- mended to overcome the less favorable climate. For mum winter temperature (Trotter and Shields 2009) example, in an area where minimum winter temper- and its rate of spread has been shown to be reduced ature reaches Ϫ23ЊC, a release of 75 L. nigrinus has a in zone 5b (Evans and Gregoire 2007). 0.06 probability of establishment (Fig. 2). Increasing In this study, release size affected the probability of the release size to 600 increases the probability of establishment and larger release sizes have been establishment to 0.61. shown to increase the likelihood of establishment of Releases of L. nigrinus at newly infested sites with biological control agents in other Þeld studies (Camp- low A. tsugae densities (i.e., Ͻ25% shoots infested) bell 1976, Memmott et al. 1998, Grevstad 1999). De- should be preferred, when possible. Although L. nig- spite this relationship, small releases often result in rinus established equally well across the range of A. establishment and releasing more than the minimum tsugae densities tested, releases during the early stages required could be an inefÞcient strategy (Campbell of A. tsugae infestation may give sufÞcient time for 1976, Memmott et al. 1998). For example, in the Þeld effective predator: prey ratios to develop before se- of conservation biology a release of Ͼ80Ð120 endan- vere decline in eastern hemlock health or death. In gered bird or mammal species does not increase the terms of release timing, because L. nigrinus established probability of establishment (GrifÞth et al. 1989). Fur- equally well when released in fall, winter, and early thermore, small releases of a root weevil have been spring the only recommendation is to release them as shown to be as effective as large releases in con- soon as they are available, weather permitting. trolling the weed, houndstongue (De Clerck-Floate The relative ease of L. nigrinus establishment over and Wikeem 2009). Releases of just 80 Leucopis ob- a wide geographic area is encouraging and might be scura Haliday (Diptera: Chamaemyiidae), a predator indicative of its future success for biological control of of the Eurasian pine adelgid, Pineus pini (Macquart) A. tsugae. Whether L. nigrinus can succeed in the (Hemiptera: Adelgidae), led to successful biological ultimate goal of suppressing A. tsugae density below control of the pest in Hawaii (Culliney et al. 1988). damaging densities has yet to be determined. The Prey density and the season of release had no in- effect of minimum winter temperature, A. tsugae den- ßuence on L. nigrinus numbers or establishment in our sity at the time of release, release size, and release study. Releases across the range of A. tsugae densities season on the ability of L. nigrinus to reduce A. tsugae tested (17Ð62% of shoots infested) and during fall, density is also not clear at this time. At the 22 sites winter, and spring were successful. We expected that reported here, we have initiated long-term monitoring releases on trees with low A. tsugae density would of eastern hemlock health, A. tsugae density, and L. decrease the likelihood of establishment. We also ex- nigrinus populations. Through periodic remeasure- pected early spring releases to have better establish- ment and other intensive studies, we anticipate an- ment rates because female predators could oviposit swers to these key questions. immediately after release, while those released in the fall or winter would have to survive adverse winter conditions before ovipositing. Furthermore, predators Acknowledgments released in the fall might be unmated and subject to We thank A. Lamb, M. Cornwell, H. Benero, and R. Allee effects if they could not locate mates. Hughes for rearing L. nigrinus; L. Solter, A. Lawrence, and J. Implications for L. nigrinus Releases. The USDA Becnel for pathogen testing, N. Havill for PCR-restriction plant hardiness zone map (Cathey 1990), based on fragment-length polymorphism analyses; E. Day for insect minimum winter temperatures, is a useful tool for identiÞcation conÞrmation; and H. Cowart, L. Barnes, M. selecting L. nigrinus release sites and sizes in the east- Dalusky, L. Randolth, J. Compton, G. Taylor, J. Dimeglio, D. ern United States for biological control of A. tsugae. Hamilton, S. Skeate, R. McDonald, C. Thomas, P. Gilliland, R. Adult L. nigrinus from the marine west coast climate Turcotte, T. Elliott, S. Hutchinson, K. Kish, M. Fugate, M. of northwestern North America have a high establish- Bing, R. Rabaglia, S. Tilley, B. Handley, T. Lupp, B. Thomp- son, B. Onken, A. Soir, M. Blumenthal, J. Unger, S. Werner, ment probability in warm areas (i.e., zones 7a, 6b, and and A. Paradis for beat sheet sampling in their regions and 6a). If the aim is to establish as many populations as helping locate release sites; and R. Van Driesche, J. Birch, M. possible throughout A. tsugaeÕs large invasive range, Lavine, and two anonymous reviewers for their comments. small release sizes should be considered in warm areas. Funding was provided by USDA Forest Service Southern This would allow for more established populations, Region Grant No. 04-DG-11083150Ð030, USDA APHIS Co- reproduction, dispersal, and potential impact on A. operative Agreement No. 05Ð8351-0305-CA, and Friends of tsugae across the landscape given a limited supply of the Blue Ridge Parkway. predators. For example, in an area where minimum winter temperature reaches Ϫ15ЊC, a release of 1,200 L. nigrinus has a 0.99 probability of establishment (Fig. References Cited 2). One alternative approach to this scenario is to Anonymous. 2009. Hemlock woolly adelgid distribution make 16 releases of 75 L. nigrinus/site. In this case, maps. (http://www.na.fs.fed.us/fhp/hwa/). April 2010 MAUSEL ET AL.: ESTABLISHMENT OF Laricobius nigrinus 447

Campbell, M. M. 1976. Colonization of Aphytis melinus De- Humble, L. M., and L. Mavin. 2005. Preliminary assessment Bach (Hymenoptera: Aphelinidae) in Aonidiella aurantii of the cold tolerance of Laricobius nigrinus, a winter- (Mask.) (Hemiptera: Coccidae) on citrus in South Aus- active predator of the hemlock woolly adelgid from west- tralia. Bull. Entomol. Res. 65: 639Ð668. ern Canada. pp. 304. In B. Onken and R. C. Reardon Cathey, H. M. 1990. U.S. Department of Agriculture plant [eds.], Third Symposium on Hemlock Woolly Adelgid in hardiness zone map. U.S. Department of Agriculture. the Eastern United States. U.S. Department of Agricul- Washington, DC. ture. Forest Service, Washington, DC. Cheah, C., M. E. Montgomery, S. M. Salom, B. L. Parker, S. Kohler, G. R., V. L. Stiefel, K. F. Wallin, and D. W. Ross. Costa, and M. Skinner. 2004. Biological control of hem- 2008. Predators associated with the hemlock woolly ad- lock woolly adelgid. U.S. Department of Agriculture. For- elgid (Hemiptera: Adelgidae) in the PaciÞc Northwest. est Service, Washington, DC. Environ. Entomol. 37: 494Ð505. Clausen, C. P. 1951. The time factor in biological control. J. Lamb, A. B., S. M. Salom, and L. T. Kok. 2005a. Survival and Econ. Entomol. 44: 1Ð8. reproduction of Laricobius nigrinus Fender (Coleoptera: Clausen, C. P. 1978. Introduced parasites and predators of Derodontidae), a predator of hemlock woolly adelgid, pests and weeds: a world review. U.S. Depart- Adelges tsugae Annand (Homoptera: Adelgidae) in Þeld ment of Agriculture. Washington, DC. cages. Biol. Control. 32: 200Ð207. Costa, S., and B. Onken. 2006. Standardizing sampling for Lamb, A. B., S. M. Salom, and L. T. Kok. 2005b. Guidelines detection and monitoring of hemlock woolly adelgid in for rearing Laricobius nigrinus Fender, pp. 309Ð318. In B. eastern hemlock forests. U.S. Department of Agriculture. Onken and R. C. Reardon [eds.], Third Symposium on Forest Service, Washington DC. Hemlock Woolly Adelgid in the Eastern United States. Culliney, T. W., J. W. Beardsley, Jr., and J. J. Drea. 1988. U.S. Department of Agriculture. Forest Service, Wash- Population regulation of the Eurasian pine adelgid (Ho- ington, DC. moptera: Adelgidae) in Hawaii. J. Econ. Entomol. 81: Lamb, A. B., S. M. Salom, L. T. Kok, and D. L. Mausel. 2006. 142Ð147. ConÞned Þeld release of Laricobius nigrinus (Coleoptera: Debach, P., and D. Rosen. 1991. Biological control by nat- Derodontidae), a predator of the hemlock woolly adel- ural enemies. Cambridge University Press, Cambridge, gid, Adelges tsugae (Hemiptera: Adelgidae), in Virginia. United Kingdom. Can. J. For. Res. 36: 369Ð375. De Clerck-Floate, R., and B. Wikeem. 2009. Inßuence of Lawrence, J. F., and T. F. Hlavac. 1979. Review of the Der- release size on establishment and impact of a root weevil odontidae (Coleoptera: ) with new species from North America and Chile. Coleop. Bull. 33: 369Ð414. for the biocontrol of houndstongue (Cynoglossum offici- Mausel, D. L., S. M. Salom, L. T. Kok, and J. G. Fidgen. 2008. nale). Biocontrol Sci. Technol. 19: 169Ð183. Propagation, synchrony, and impact of introduced and Eschtruth, A. K., N. L. Cleavitt, J. J. Battles, R. A. Evans, and native Laricobius spp. (Coleoptera: Derodontidae) on T. J. Fahey. 2006. Vegetation dynamics in declining east- hemlock woolly adelgid in Virginia. Environ. Entomol. 37: ern hemlock stands: 9 years of forest response to hemlock 1498Ð1507. woolly adelgid infestation. Can. J. For. Res. 36: 1435Ð1450. Memmott, J., S. V. Fowler, and R. L. Hill. 1998. The effect Evans, A. M., and T. G. Gregoire. 2007. A geographically of release size on the probability of establishment of variable model of hemlock woolly adelgid spread. Biol. biological control agents: gorse thrips (Sericothrips Invasions 9: 369Ð382. staphylinus) released against gorse (Ulex europaeus)in Fidgen, J. G., D. E. Legg, and S. M. Salom. 2006. Binomial New Zealand. Biocontrol Sci. Technol. 8: 103Ð115. sequential sampling plan for hemlock woolly adelgid Montgomery, M. E., and S. M. Lyon. 1996. Natural enemies (Hemiptera: Adelgidae) sistens infesting individual east- of adelgids in North America: their prospect for biological ern hemlock trees. J. Econ. Entomol. 99: 1500Ð1508. control of Adelges tsugae (Homoptera: Adelgidae), pp. Grevstad, F. S. 1999. Experimental invasions using biologi- 89Ð102. In S. M. Salom, T. C. Tigner, and R. C. Reardon cal control introductions: the inßuence of release size on [eds.], Proceedings of the First Hemlock Woolly Adelgid the chance of population establishment. Biol. Invasions 1: Review. U.S. Department of Agriculture. Forest Service, 313Ð323. Washington, DC. Griffith, B., J. M. Scott, J. W. Carpenter, and C. Reed. 1989. Orwig, D. A., and D. R. Foster. 1998. Forest response to the Translocation as a species conservation tool: status and introduced hemlock woolly adelgid in southern New En- strategy. Science 245: 477Ð480. gland, U.S.A. J. Torrey Bot. Soc. 125: 60Ð73. Havill, N. P., M. E. Montgomery, G. Y. Yu, S. Shiyake, and A. Pschorn-Walcher, H. 1977. Biological control of forest in- Caccone. 2006. Mitochondrial DNA from hemlock sects. Ann. Rev. Entomol. 22: 1Ð22. woolly adelgid (Hemiptera: Adelgidae) suggests cryptic Sasaji, H., and M. S. McClure. 1997. Description and distri- speciation and pinpoints the source of the introduction to bution of Pseudoscymnus tsugae sp. nov. (Coleoptera: eastern North America. Ann. Entomol. Soc. Am. 99: 195Ð Coccinellidae), an important predator of hemlock woolly 203. adelgid in Japan. Ann. Entomol. Soc. Am. 90: 563Ð568. Havill, N. P., G. A. Davis, J. Klein, A. Caccone, and S. M. Shea, K., and H. P. Possingham. 2000. Optimal release strat- Salom. 2010. Hemlock woolly adelgid biological con- egies for biological control agents: an application of sto- trol: molecular methods to distinguish Laricobius nig- chastic dynamic programming to population manage- rinus, L. rubidus, and their hybrids. In K. A. McManus ment. J. Appl. Ecol. 37: 77Ð86. and K. W. Gottschalk [eds.], Proceedings, 21st U.S. Steward, V. B., and T. A. Horner. 1994. Control of hemlock Department of Agriculture Interagency Research Fo- woolly adlegid using soil injections of systemic insecti- rum on Invasive Species. USDA, Forest Service, Wash- cides. J. Aboric. 20: 287Ð288. ington, DC (in press). Sutherst, R. W., G. F. Maywald, W. Bottomley, and A. Humble, L. M. 1994. Recovery of additional exotic preda- Bourne. 2004. CLIMEX v. 2 UserÕs Guide. Hearne Sci- tors of balsam woolly adelgid, Adelges piceae (Ratzeburg) entiÞc Software, Melbourne, Australia. (Homoptera: Adelgidae), in British Columbia. Can. En- Trotter, R. T., and K. S. Shields. 2009. Variation in winter tomol. 126: 1101Ð1103. survival of the invasive hemlock woolly adelgid 448 ENVIRONMENTAL ENTOMOLOGY Vol. 39, no. 2

(Hemiptera: Adelgidae) across the eastern United States. Zilahi-Balogh, G.M.G., L. T. Kok, and S. M. Salom. 2002. Environ. Entomol. 38: 577Ð587. Host speciÞcity of Laricobius nigrinus Fender (Co- Wallace, M. S., and F. P. Hain. 2000. Field surveys and leoptera: Derodontidae), a potential biological control evaluation of native and established predators of the agent of the hemlock woolly adelgid, Adelges tsugae An- hemlock woolly adelgid (Homoptera: Adelgidae) in the nand (Homoptera: Adelgidae). Biol. Control. 24: 192Ð198. southeastern United States. Environ. Entomol. 29: 638Ð Zilahi-Balogh, G.M.G., C. D. Broeckling, L. T. Kok, and S. M. 644. Salom. 2005. Comparison between a native and exotic Ward, J. S., M. E. Montgomery, C.A.S.-J. Cheah, and R. S. adelgid as hosts for Laricobius rubidus (Coleoptera: Der- Cowles. 2004. Eastern hemlock forests: guidelines to odontidae). Biocontrol Sci. Technol. 15: 165Ð171. Zilahi-Balogh, G.M.G., L. M. Humble, A. B. Lamb, S. M. minimize the impacts of hemlock woolly adelgid. U.S. Salom, and L. T. Kok. 2003. Seasonal abundance and Department of Agriculture. Forest Service, Washington, synchrony between Laricobius nigrinus (Coleoptera: DC. Derodontidae) and its prey, the hemlock woolly adelgid Yu, G., and M. E. Montgomery. 2007. A new species of (Hemiptera: Adelgidae). Can. Entomol. 135: 103Ð115. Laricobius (Coleoptera: Derodontidae) from Taiwan. Zilahi-Balogh, G.M.G., J. Jelı´nek, T. J. McAvoy, S. M. Salom, Formosan Entomol. 27: 341Ð347. and L. T. Kok. 2007. Two new species of Laricobius (Co- Yu, G. Y., M. E. Montgomery, and D. F. Yao. 2000. Lady leoptera: Derodontidae) from China, and a key to Lari- (Coleoptera: Coccinellidae) from Chinese hem- cobius in the southeastern palaearctic. Proc. Entomol. locks infested with the hemlock woolly adelgid, Adelges Soc. Wash. 109: 377Ð384. tsugae Annand (Homoptera: Adelgidae). Coleop. Bull. 54: 154Ð199. Received 18 March 2009; accepted 29 October 2009.