And Naturally Occurring Generalist Predators

BIOLOGICAL CONTROLÐPARASITOIDS AND PREDATORS Interactions Between the Augmentatively Released Predaceous Mite Galendromus occidentalis (Acari: Phytoseiidae) and Naturally Occurring Generalist Predators

R. G. COLFER, J. A. ROSENHEIM, L. D. GODFREY, AND C. L. HSU

Department of Entomology, University of California, One Shields Avenue, Davis, California, 95616

Environ. Entomol. 32(4): 840Ð852 (2003) ABSTRACT Predatory mite releasescanbe an effective meansof managing spidermitesin many perennial cropping systems, yet little research has been performed in annual cropping systems. Herein we evaluate the compatibility of predaceous mite releases with the conservation of resident natural enemies in an annual agroecosystem. We quantify the impact of naturally occurring generalist predators, Geocoris spp. and Orius tristicolor White, and the omnivore Frankliniella occidentalis (Pergande), on the establishment of the western predatory mite Galendromus occidentalis (Nesbitt) and how these predator-predator interactions inßuence spider mite control. Field experiments showed that in the absence of generalist predators, released predatory mites can establish populations on cotton, increase in abundance through reproductive recruitment, and suppress spider mite populations.Hemipteran predatorshad a negative impact on predatory mite populationsbut generally improved spider mite suppression. The presence of F. occidentalis had no impact on predatory mite performance.

KEY WORDS augmentative biological control, generalist predator, intraguild predation, Galendro- mus occidentalis, Gossypium hirsutum

BIOLOGICAL CONTROL IS ONE of the most important al- man and Zohdi 1976, Tijerina-Chavez 1991) and Þeld ternatives to conventional pesticide use in pest man- corn (Pickett and Gilstrap 1986, Pickett et al. 1987). agement today. Biological control isfree of many of Attemptsto improve augmentative biological control the problems associated with pesticide use, such as have focused primarily on factors such as selection of pest resistance, environmental pollution, and worker control agents, quality control, mass rearing tech- health impacts. Recently, however, there has been a niques, release methods, and efÞcacy of target pest critical reevaluation of the compatibility of classical suppression (Van Driesche and Bellows 1996). Al- biological control (the introduction of exotic natural though all of these factors are critical to the successful enemies) with the conservation of endemic arthro- use of augmentative biological control, there has been pods, as a result of the potential impact of imported very little examination of how augmented natural en- control agentson nontarget native fauna (Howarth emiesmay interact with naturally occurring natural 1991, Simberloff and Stiling 1996, Louda et al. 1998). enemies. There is increasing recognition that complex These conservation concerns will likely place some multispeciesinteractionsamongnatural enemiescan limits on the future implementation of classical bio- be important in biological control (Rosenheim et al. logical control. Therefore, there isa need to put 1995, Sunderland et al. 1997, Rosenheim 1998, see greater emphasisonother areasof biological control, ÔInvited FeatureÕ Ecological Applications9:363Ð429). such as natural enemy conservation and augmentation In an annual crop such as cotton, generalist predators (Parrella et al. 1992, Barbosa 1998, Pickett and Bugg are often abundant (van den Bosch and Hagen 1966). 1998). In thisstudy,we evaluate whether or not the Asa result,predator-predator interactionsmay be conservation of naturally occurring generalist preda- common and could be a key factor inßuencing the torsiscompatible with augmentation of predatory establishment of augmentatively released natural en- phytoseiid mites. emies. Augmentative releasesofpredaceousphytoseiid There is substantial evidence that generalist pred- mites have been shown experimentally to reduce spi- ators can have signiÞcant impacts on phytoseiid mites. der mite densitiesinmany perennial crops(McMurtry Predation upon predatory phytoseiidmiteshasbeen 1982, Hoy et al. 1982, Flaherty et al. 1985, Helle and studied (either directly or indirectly) in experiments Sabelis1985, Croft and MacRae 1992b, Nyrop et al. conducted in the laboratory (Gillespie and Quiring 1998) and some annual row crops such as cotton (Os- 1992, Cloutier and Johnson 1993, Croft and Croft 1996,

0046-225X/03/0840Ð0852$04.00/0 ᭧ 2003 Entomological Society of America August 2003 COLFER ET AL.: G. occidentalis-GENERALIST PREDATOR INTERACTIONS 841

Croft et al. 1996, MacRae and Croft 1996, Wittmann Materials and Methods and Leather 1997, Schausberger and Croft 2000), High Spider Mite Availability and High Predatory greenhouse (Ramakers 1993, Brodsgaard and Enkeg- Mite Release Rate—Cage Experiment. Thisexperi- aard 1997, Schausberger and Walzer 2001) and Þeld ment wasdesignedto quantify: (1) the ability of (Croft and MacRae, 1992a, 1992b, 1993; Croft 1994; G. occidentalis to establish and build populations on Walde et al. 1997). Croft and MacRae (1992a) showed cotton under conditionsof high spidermite availabil- that predation by the generalist predatory mite Zetzel- ity and low abundance of resident generalist preda- lia mali usually displaced western predatory mite Ga- tors; and (2) the impact of G. occidentalis releases on lendromus occidentalis (Nesbitt) populations, which T. urticae abundance. The experiment wasconducted sometimesledto increasesinphytophagousmite pop- from 31 May to 25 June 1997 in a 0.4-ha experimental ulations. Phytoseiid mites may be especially suscep- planting of cotton, Gossypium hirsutum cultivar tible to predation by predatory insects because they “Maxxa,” at the UC DavisPlant Pathology Fieldhouse, are relatively small compared with insects (Polis et al. Davis, CA (a Þeld station with both Þeld and green- 1989), but there have been no Þeld studies examining house facilities). Plants were grown on rows separated the impact of predatory insects on phytoseiid mites. by 76 cm following standard commercial practices, Our research was also motivated by earlier work in except that no acaricidesor insecticideswereused. which releases of G. occidentalis into cotton were Plantswere small( Ϸ8 mainstem nodes) and not yet unsuccessful in establishing populations of this mite ßowering when the experiment wasinitiated. (R. G. Colfer, J. A. Rosenheim, L. D. Godfrey, C. L. The experimental unit wasa singleplant. On Hsu, unpublished data). We were unable to establish 31 May, plantswere randomly selectedand thor- G. occidentalis populations in the cotton agroecosys- oughly sprayed with Safer Brand insecticidal soap tem even though the majority of the releases were (Woodstream Corp., Lititz, PA) at the labeled rate made in Þeldsthat had abundant spidermite prey and (20 ml soap/ L H2O) to reduce resident populations were organically farmed (no pesticides were used). of predators. Plants were then enclosed in cylindrical In thisstudy,we evaluate the importance of inter- cages composed of a plastic polyvinyl chloride base actions between augmentatively released western (30 cm diameter) and No-Thripsmesh(Greentek Inc; predatory mites, G. occidentalis, and a group of im- pore size Ϸ150 ␮m; cage dimensions: height 45 cm, portant naturally occurring generalist predators. diameter 30 cm). Cage bases were imbedded in the These predators include the minute pirate bug, Orius ground to provide a tight seal at the base. tristicolor (White) (Hemiptera: Anthocoridae), the On 1 June, plants were randomly assigned to one of bigeyed bugs, Geocoris pallens Sta¨l, G. punctipes (Say), two treatments, each replicated twenty-eight times: and G. atricolor Montandon (Hemiptera: Lygaeidae), (1) spider mites alone (T. urticae) or (2) spider mites and the western ßower thrips, Frankliniella occiden- pluswesternpredatory mites( G. occidentalis). Spider talis (Pergande) (Thysanoptera: Thripidae). These miteswere added to all replicatesby placing two predators are commonly associated with the spider spider mite-infested cotton seedlings from a labora- mite complex found in cotton grown in the San Joa- tory culture onto each plant; thisdelivered 471 Ϯ 45 quin Valley of California (Wilson et al. 1991a). Species (mean Ϯ 1 SE) mobile spider mites to each replicate. in thisspidermite complex include Tetranychus paciﬁ- Western predatory mites were purchased from Bio- cus McGregor, Tetranychus turkestani Ugarov and Ni- tactics Inc., Riverside, CA, and were released within kolski, and Tetranychus urticae Koch. 2 d of receiving the product. On 1 June, Ϸ10 adult We address four questions in this study. First, can predatory miteswere added to each replicate of the western predatory mites establish and build up pop- predatory mite treatment. On 7 June, a second release ulationsthrough reproductive recruitment on cotton of 68 Ϯ 16 predatory mitesin a corn-cob grit carrier when spider mite availability is high and in the ab- wasmade to each replicate of the predatory mite sence of generalist predators? Second, what impact do treatment. The second predatory mite release was generalist predators have on western predatory mite done to ensure that all replicates of the predatory mite population dynamicsin cotton? Third, what impact do release treatment contained predatory mite popula- naturally occurring generalist predators have on spi- tions. A small number of spider mite eggs (Ͻ50 per der mite population dynamics? Fourth, do the inter- release, T. paciﬁcus McGregor) were included in the actionsbetween westernpredatory mitesand gener- corn-cob grit carrier by BiotacticsInc. to feed pred- alist predators inßuence the overall level of spider atory miteswhile in transit. mite suppression? We evaluated interactions between On 15 June, 18 out of the 28 replicatesof each western predatory mites and generalist predators us- treatment were terminated (8 d after the second pred- ing two approaches: we examined short-term interac- atory mite release). The remaining 10 replicates of tionsunder controlled conditionsusingsmallÞeld each treatment were collected on 25 June (18 d after cagesin which we testedpredatorssinglyand in de- the second predatory mite release). All leaves from Þned combinations; and, we examined longer-term these replicates were collected into plastic bags, pre- interactions at a larger spatial scale using insecticide served with 70% ethanol, and stored at 4ЊC. All ar- manipulations. In this later approach, we tested the thropodswere later removed from the leaf material impact of the whole generalist predator complex on using a leaf washing method developed by Leigh et al. western predatory mite and spider mite populations. (1984). Note that we did not use a repeated-measures 842 ENVIRONMENTAL ENTOMOLOGY Vol. 32, no. 4 design, replicates were destructively sampled during (G. occidentalis, 10.6 Ϯ 0.7 motile stages per leaf), (3) the 8-d and 18-d sampling dates. To reduce the time spider mites, predaceous mites, and O. tristicolor (four necessary to quantify samples, we counted only the Þrst to third instar nymphs per leaf), (4) spider mites, larger stages of mites and used a previously deter- predaceousmites,and Geocoris spp. (one Þrst- to mined linear regression relationship to estimate the third-instar nymph per leaf), and (5) spider mites, total number of motile mites. The linear regression predaceousmites,and F. occidentalis (Ϸ12 adultsper relationship was determined by separating mites by leaf). Densities were chosen to reßect natural densi- size using two mesh sieves: one with 260-␮m diameter tiesof predatorsin cotton when spidermite densities poresto collect adultsand larger immature mite stages are high (R.G. Colfer, unpublished data). Tetranychus (i.e., deutonymphs), and a second with 100-␮m diam- urticae and F. occidentalis were collected from labo- eter pores to collect all other smaller stages (i.e., eggs, ratory cultures, G. occidentalis waspurchasedfrom larvae, protonymphs). By quantifying all motile stages BiotacticsInc., and Geocoris and O. tristicolor were (eggswere not counted) in both sievesandusing hand collected at or near the cotton Þeld where this linear regression through the origin, we developed a experiment wasconducted. Spider miteswere added relationship between the proportion of spider mites to all replicates by placing one spider mite-infested and phytoseiid mites found in the two sieves (spider cotton cotyledon from a laboratory culture onto each mites: bottom sieve ϭ top sieve ϫ 0.937, r2 ϭ 0.927, P Ͻ leaf; thisdelivered 147 Ϯ 15 (mean Ϯ 1 SE) mobile 0.0001; phytoseiid mites: bottom sieve ϭ top sieve ϫ spidermitesto each replicate. Predatory miteswere 0.358, r2 ϭ 0.853, P Ͻ 0.0001). The regression analysis delivered to cageswithin corn-cob grit carrier. allowed usto quantify only the larger stagesofmites The duration of thisexperiment was7 d (approxi- but obtain an estimate of the total number of motile mately the generation time for the spider mites, pre- mites. The relationship between large spider mites and daceous mites, and thrips). From 28Ð30 August, rep- total spidermiteswassimilarfor different treatments licateswere collected and all herbivorousand (R.G. Colfer, unpublished data). predatory arthropodswere quantiÞed in the labora- Compatibility of Spider Mite Predators—Cage Ex- tory with the aid of a dissecting stereomicroscope. periment. Thisexperiment wasdesignedto quantify Both the motile and egg stages of spider mites and the impact of generalist predators Geocoris spp., O. predatory miteswere quantiÞed. tristicolor, and F. occidentalis, on spider mite and pred- Compatibility of Spider Mite Predators—Open Plot atory mite abundance. The experiment wasconducted Experiment. Thisexperiment wasdesignedto evalu- from 14 to 30 August 1997 in a 0.2-ha experimental ate the compatibility of predaceousmiteswith the planting of G. hirsutum cultivar “Maxxa” at the UC unmanipulated naturally occurring generalist preda- Davis Agronomy Field Plots, Davis, CA. Plants were tor community and to quantify the impact of the gen- grown on rows separated by 76 cm, following standard eralist predator community on spider mite popula- commercial practices, except that no acaricides or tions. The experiment was conducted from 21 May to insecticides were used. Plants were medium sized 15 July 1997 in a two ha planting of G. hirsutum cultivar (Ϸ20 mainstem nodes) and setting squares and bolls “Maxxa” at the UC Cotton Research Station, Shafter, when the experiment began. CA. Experimental unitswere cotton plots(28 m ϫ The experimental unit was a single mainstem leaf 12 m) surrounded by 3.5 m of bare soil. Plots were located at the Þfth node from the plant terminal. From randomly allocated to one of four treatments, each 14Ð15 August, plants were randomly selected and the replicated seven times: (1) prerelease application of Þfth node mainstem leaf was thoroughly cleaned with acephate (Orthene) and release of predatory mites, a paint brush to reduce resident populations of west- (2) prerelease application of acephate and no release, ern ßower thrips and other insects. Leaves were then (3) release of predatory mites only, and (4) no ma- enclosed in square cages composed of No-Thrips mesh nipulation control. Acephate wassprayedat 4.0 oz (Greentek Inc.; pore size Ϸ0.15 mm; cage dimensions: AI/ac on 21 May in an attempt to reduce naturally length and width 22.7 cm). Two seams were closed occurring populationsof arthropod predators.On using plastic folder bindings to facilitate easy entry 30 May, a leaf-disk bioassay using leaves from sprayed into cages; the petiole-side seam was closed using a and unsprayed plots showed that mortality of preda- combination of double-sided mounting tape, Duck tory mites on leaf disks from sprayed and unsprayed tape, and rope caulk weather-stripping (Ace Hard- plotsdid not differ signiÞcantly( P Ͼ 0.3). Twelve ware Corp., Oak Brook, IL). hundred predatory miteswere manually releasedper Seven dayslater, between 21Ð22 August,cageswere plot (release rate equivalent to 38,200 mites per ha) on reopened and brushed a second time to remove newly 5 June. The manual released involved evenly distrib- emerged insects that have egg stages embedded in the uting a mixture of predatory mitesand corn-cob grit leaf tissue (F. occidentalis and O. tristicolor). Thiswas carrier onto each plant within the release plots. Re- considered sufÞcient time for all eggs to hatch. This leases were performed between 0600 and 0800 hours removal technique waseffective for O. tristicolor but to minimize predatory mite mortality related to high only partially effective for F. occidentalis. Once midday temperatures. Plots were monitored every brushed, caged leaves were randomly allocated to one 2 wk using leaf and sweep (20 sweeps/plot) sampling of Þve treatments, each replicated eighteen times: (1) techniquesfrom 21 May to 15 July. Leaf sampling spider mites alone (T. urticae, 147 Ϯ 15 motile stages involved randomly collecting 25 mainstem leaves lo- per leaf), (2) spidermitespluspredaceousmites cated Þve nodesbelow the apex of the plant. All leaf August 2003 COLFER ET AL.: G. occidentalis-GENERALIST PREDATOR INTERACTIONS 843 samples were collected into plastic bags, preserved Table 1. Density (mean ؎ SE) of motile spider mites and with 70% ethanol, and stored at 4ЊC. All arthropods predatory mites per plant in the high spider mite availability/high predatory mite release rate experiment where generalist predators were later removed from the leaf material using a leaf were suppressed washing method developed by Leigh et al. (1984) and quantiÞed using the methods described earlier (see Experimental treatment Þrst experiment). Mite populations were estimated Spider mites ϩ Spider mitesalone usingcountsof the larger stagesofmitesand the G. occidentalis previously described regression relationship. All adult Arthropod Day 8 Day 18 Day 8 Day 18 phytoseiid mites recovered from the leaf samples were abundance slide mounted and identiÞed to species. Spider mites782 Ϯ 147 6931 Ϯ 1354 296 Ϯ 58** 2639 Ϯ 1021* Statistical Analyses. For the Þrst experiment, we Predatory 0.7 Ϯ 0.3 1.7 Ϯ 0.8 52.4 Ϯ 6.2*** 126.7 Ϯ 33.2*** analyzed the inßuence of predatory mite releases on mites Þnal spider mite and predatory mite abundance using Statistical comparisons are made between treatments for the same Kruskal-Wallis rank-sum tests for the 8-d and 18-d sample date. Two-tailed Wilcoxon rank-sum tests: *, P Ͻ 0.05; **, P Ͻ samples (Sokal and Rohlf 1995). For the second ex- 0.01; ***, P Ͻ 0.001; NS, not signiÞcant. periment, we analyzed the inßuence of different predatorson spidermite and predatory mite abundance using Kruskal-Wallis rank-sum tests and planned within cageswasfairly effective in keeping replicates paired comparisons using two-tailed Wilcoxon rank- free of naturally occurring predators of spider mites. sum tests (Sokal and Rohlf 1995). P valueswere ad- Only 14% of cagescontained naturally occurring pred- justed for the number of pairwise comparisons to ators, and within these replicates the densities of these maintain an overall ␣ value equal to 0.05. To determine predatorswere below densitiesseenon unmanipu- whether competition and predation were important in lated plants. On plants on which the generalist pred- the interactionsbetween hemipteran predatorsand ator community was experimentally suppressed, the predatory mites, we used three-factor analysis of co- release of predatory mites greatly enhanced the num- variance (ANCOVA) where Þnal predatory mite ber of predatory mitesrecovered a t 8 d post release abundance was the response variable, Þnal spider mite compared with the control (␹2 ϭ 27.6, df ϭ 1, P Ͻ abundance wasthe covariate, and addition of O. tris- 0.001; Table 1). However, the predaceousmite per ticolor, Geocoris, and F. occidentalis were main factors capita population growth rate from initial release, (Neter et al. 1990). To meet the assumptions of AN- when 78 mites were released, until the census on day COVA, Þnal spider mite and predatory mite abun- 8 wasslightlynegative (per capita growth ϭ [Þnal dance were log-transformed (ln[x ϩ1]). For the third density Ð release rate]/ [release rate] ϭϪ0.33). In the experiment, we analyzed the inßuence of predatory replicates sampled at 18 d post release, the difference mite releasesandacephate applicationson spidermite in predatory mite abundance between the release and and predatory mite populationsusinga two-factor, control treatmentsremained very large ( ␹2 ϭ 13.6, repeated-measures analysis of variance (ANOVA) df ϭ 1, P Ͻ 0.001), and the predaceousmite per capita (Neter et al. 1990, von Ende 1993). The impact of the growth rate from the time of the initial release rate was acephate on the abundance of immature predatory positive (per capita growth ϭ 0.62), indicating the insects2wk after the application wasanalyzed with a predatory mite population had established and grown one-factor ANOVA. Spider mite and predatory mite by Ͼ60%. Regression analysis of the predaceous mite abundanceswere log-transformedto meet the re- counts at both the 8- and 18-d censuses showed that quirements of the ANOVA tests. To determine predatory mite numberswere stronglycorrelated with whether competition and predation are important in spider mite densities (simple linear regression, r2 ϭ the interactionsbetween hemipteran predatorsand 0.71, F ϭ 64.7, df ϭ 1, P Ͻ 0.001; Fig. 1), suggesting that predatory mitesin the large-scale,acephate experi- predatory miteshad greater population recruitment ment, we again used ANCOVA, where the acephate under conditions of high spider mite densities. application wasthe main factor, spidermite abun- Spider mite population abundance wassigniÞcantly dance (summed over four weekly samples), and im- reduced by the predatory mite releases at both8dpost mature hemipteran abundance (from the sweep sam- release (␹2 ϭ 9.6, df ϭ 1, P ϭ 0.002) and 18 d post plestaken 2 wk after the acephate application) were release (␹2 ϭ 5.6, df ϭ 1, P ϭ 0.018, Table 1). At both covariates, and predatory mite abundance (summed censuses, predatory mites reduced spider mite densi- over four weekly samples) was the response variable. tiesby Ϸ60% compared with the spider mites only Predatory mite and spider mite abundances were log- treatment. Despite this suppressive effect, spider mite transformed to meet the assumptions of ANCOVA. All populationsgreatly increasedin both treatmentsover statistical analysis was performed using JMP statistical the duration of the experiment. Spider mite popula- package (SAS Institute 1995). tionswere Ϸ9 timeslarger at day 18 than at day 8 in both treatments. Compatibility of Spider Mite Predators—Cage Ex- Results periment. The manipulations of the generalist insect High Spider Mite Availability and High Predatory predators were successful, with much higher predator Mite Release Rate—Cage Experiment. The combina- densitiesintreatmentsin which predatorswere added tion of insecticidal soap and the containment of plants in comparison to the controls (Table 2). Frankliniella 844 ENVIRONMENTAL ENTOMOLOGY Vol. 32, no. 4

a signiÞcant effect on predatory mite abundance (␹2 ϭ 0.4, df ϭ 1, P ϭ 0.53). The negative impact that O. tristicolor and Geocoris had on the predatory miteswascausedby either predation, competition, or both. In an attempt to distin- guish between these effects, we performed an ANCOVA with Þnal spider mite abundance as the covariate and main effectsfor Orius and Geocoris presence. This analysis showed that Þnal spider mite abundance did not affect Þnal predatory mite abundance (F ϭ 1.5, df ϭ 1, P ϭ 0.22) but that the addition of both Geocoris and O. tristicolor signiÞcantly depressed Þnal predatory mite abundance (F ϭ 11.4, df ϭ 1, P ϭ 0.001; F ϭ 9.2, df ϭ 1, P ϭ 0.003, respectively). A preliminary ANCOVA test veriÞed that there was no signiÞcant Fig. 1. Predatory mite and spider mite abundance per interaction between the covariate and the main fac- plant in the predatory mite release treatment in high spider tors. Therefore, this analysis suggests that predation mite availability/ high predatory mite release rate experi- on predatory mitesby Geocoris and O. tristicolor ment. Numbersof spiderand predatory miteswere strongly 2 ϭ ϭ ϭ Ͻ played a more important role than did competition for correlated (r 0.705, F 64.7, df 1, P 0.0001). The line spider mite prey. represents the Ôline-of-best-ÞtÕ from simple linear regression. Note that spider mite abundance is plotted on a log scale. The addition of predatory mitesreduced spider 10 mite densities to 47% of the density reached in the control (␹2 ϭ 7.6, df ϭ 1, P ϭ 0.006; Fig. 2B). The occidentalis was somewhat of an exception, because simultaneous addition of predatory mites ϩ thrips, or thrips were present in nearly all the treatments. De- predatory mites ϩ Geocoris produced levelsof spider spitethisproblem, the thripsaddition treatment had mite suppression that were signiÞcant in comparison signiÞcantly higher thrips densities compared with the to the control (␹2 ϭ 12.9, df ϭ 1, P Ͻ 0.001; ␹2 ϭ 15.4, 2 control (␹ ϭ 6.7, df ϭ 1, P ϭ 0.01). df ϭ 1, P Ͻ 0.001, respectively), but not signiÞcantly Predatory mite releases enhanced the predaceous different from the predatory mitesalone treatment mite density (Table 2). Final densities of predatory (␹2 ϭ 0.4, df ϭ 1, P ϭ 0.53; ␹2 ϭ 2.5, df ϭ 1, P ϭ 0.12, miteswere, however, lower than the releaserate (re- respectively). However, the addition of predatory lease rate, 10.6 Ϯ 0.7 mitesper plant; Þnal predator mites ϩ O. tristicolor lowered spider mite abundance density; 3.7 Ϯ 0.8), resulting in a negative per capita below both the control and the predatory mite alone population growth rate (per capital growth ϭ [(Þnal treatment levels( ␹2 ϭ 26.9, df ϭ 1, P Ͻ 0.001; ␹2 ϭ 21.7, abundance Ð initial abundance)/ initial abundance] ϭ df ϭ 1, P Ͻ 0.001, respectively). Ϫ0.65). This result is similar to that observed at the 8-d Although we attempted to exclude thripsfrom all censusinexperiment 1, in which there wasalsoa the treatmentsexcept one, we observedthripsin all period of negative per capita population growth after treatments; however, the densities of these naturally the initial release. present thrips were substantially reduced in the The addition of hemipteran predatorshad a strong hemipteran predator treatments(Fig. 2C). Final den- negative impact on predatory mite abundance (Fig. sitiesofthripswere signiÞcantlylower in treatments 2A). Predatory mite abundance wasreduced from containing Orius and Geocoris compared with the spi- 3.67 Ϯ 0.78 in the predatory mite only treatment to 0.0 der mite alone treatment (␹2 ϭ 23.5, df ϭ 1, P Ͻ 0.001; 2 in the predatory mite ϩ O. tristicolor treatment (␹ ϭ ␹2 ϭ 9.8, df ϭ 1, P ϭ 0.002, respectively). However, the 22.8, df ϭ 1, P Ͻ 0.001) and 0.83 Ϯ 0.35 in the predatory presence of western predatory mites alone did not 2 mite ϩ Geocoris treatment (␹ ϭ 9.8, df ϭ 1, P ϭ 0.002). affect thripsabundance ( ␹2 ϭ 0.3, df ϭ 1, P ϭ 0.57). In contrast, the addition of F. occidentalis did not have Compatibility of Spider Mite Predators—Open Plot Experiment. In thisexperiment, acephate wasapplied Table 2. Arthropod treatment manipulations in the predator to plotsto reduce the densitiesofnaturally occurring -compatibility cage experiment. Shown are the mean ؎ SE number generalist predators while minimizing the pesticide of motile predators per cage at day 7. Means represent densities of induced mortality of spider mites. The application of predators in the four different predator addition treatments acephate waseffective at reducing predators;2 wk Predator after the acephate application immature predator Treatment Control addition abundance (excluding western ßower thrips) was reduced by 80% (␹2 ϭ 13.6, df ϭ 1, P Ͻ 0.001, Fig. 3A) Western predatory mite only 0 3.7 Ϯ 0.8*** Thripsa 7.7 Ϯ 1.7 13.0 Ϯ 1.5* and westernßower thripsabundance wasreduced by 2 Geocoris spp. 0 0.5 Ϯ 0.1*** 48% (␹ ϭ 4.8, df ϭ 1, P ϭ 0.026, Table 3). Some Orius tristicolor 0.05 Ϯ 0.03 1.1 Ϯ 0.3** generalist predators, such as Geocoris spp., remained at suppressed densities in the acephate-treated plots Two-tailed Wilcoxon rank-sum tests: *, P Ͻ 0.05; **, P Ͻ 0.01; ***, ϭ ϭ Ͻ P Ͻ 0.001; NS, not signiÞcant. throughout the 8-wk experiment (F 35.8, df 1, P a Dominant species was Frankliniella occidentalis; minor species 0.001, Table 3). Other predators, such as O. tristicolor included Caliothrips fasciatus and Scolothrips sexmaculatus. and F. occidentalis, actually became more abundant in August 2003 COLFER ET AL.: G. occidentalis-GENERALIST PREDATOR INTERACTIONS 845

Fig. 2. Mean (ϩSE) arthropod abundance per leaf in predator compatibility cage experiment. (A) Predatory mite abundance in treatmentswith predatory mitesalone and in treatmentswhere predatory miteswere combined with the generalist predators: minute pirate bug (Orius tristicolor), big-eyed bug (Geocoris spp.), and western ßower thrips (Fran- kliniella occidentalis). (B) Spider mite abundance in treatmentswith spidermitesalone and spidermitescombined with different combinationsof predatory mitesand generalistpredators.(C) Thripsabundance in all treatments.Symbolsabove bars display results of statistical tests comparing arthropod species abundance in each predator treatment with the species alone treatments. the acephate plotsthan in the control plots6 wk after the release and control plots in the postrelease samples the acephate applications( O. tristicolor: ␹2 ϭ 16.5, (release: 96% G. occidentalis, no release: 92% G. occi- df ϭ 1, P Ͻ 0.001; F. occidentalis: ␹2 ϭ 7.9, df ϭ 1, P ϭ dentalis). 0.005; Table 3). With the decline in predatory insect abundance, we Predatory mite numbers increased signiÞcantly in observed higher spider mite densities in acephate plotssprayedwith acephate over the duration of the treated plotscompared with untreated plotsover the experiment (F ϭ 10.6, df ϭ 1, P ϭ 0.003, Fig. 4). This duration of the experiment (F ϭ 29.0, df ϭ 1, P Ͻ 0.001, increase was especially conspicuous when viewed as Fig. 5). Spider mite densitiesintreated plotswere 4.2 cumulative predatory mite abundance across the four timesgreater and 9.8 timesgreater than the untreated sampling dates after the acephate application (␹2 ϭ plots2 and 4 weeksafter the acephate application, 11.4, df ϭ 1, P Ͻ 0.001, Fig. 3B). Predatory mite re- respectively. Spider mites declined in both sprayed leases, however, did not have any detectable effect on and unsprayed plots after 2 June, and sprayed and total predatory mite densities over the duration of the unsprayed plots had similarly low densities of spider experiment follow the releases (F ϭ 0.0, df ϭ 1, P ϭ mitesby 15 July ( F ϭ 2.2, df ϭ 1, P ϭ 0.14); the 0.91, Fig. 4). Predatory mitesrecovered from the Þeld decrease in spider mite populations was correlated included G. occidentalis (94/114, 82%), Neoseiulus au- with an increase in insect predator abundance across rescens Athias-Henriot (15%), and Neoseiulus fallacis sampling dates (Spearmans ␳ ϭϪ0.3, P Ͻ 0.001). (Garman) (3%). On 2 June, before the release of Given that the releasesofpredatory miteshad no G. occidentalis, the most abundant species was N. au- effect on overall predatory mite densities (Fig. 4), it rescens (14/25, 56%) followed by G. occidentalis was not surprising that these releases also had no (44%). On 15 and 29 June, 2 and 4 wk after the pred- detectable effect on spider mite densities (F ϭ 0.3, atory mite release, the predatory mite community was df ϭ 1, P ϭ 0.60; Fig. 5). dominated by G. occidentalis (Table 4). Galendromus Acephate sprays led to decreases in the populations occidentalis was, however, equally common in both of generalist insect predators and increases in the 846 ENVIRONMENTAL ENTOMOLOGY Vol. 32, no. 4

dation pressure or because they experienced less competition for spider mite prey. To determine which of thesefactorswasmore important for the variation in predatory mite abundance between the sprayed and unsprayed plots, we used ANCOVA on the census data after the acephate spray. In this analysis, predatory mite abundance (population abundance summed across four sampling dates) was the response variable, generalist predator abundance (from sweep samples 2 wk after the acephate application) and spider mite abundance (population abundance summed across four sampling dates) were the covariates, and the acephate spray was the main effect. A preliminary ANCOVA test veriÞed that there was no signiÞcant interaction between the covariatesand the main factor. Spider mite abundance had a signiÞcant effect on predatory mite abundance (F ϭ 4.4, df ϭ 1, P ϭ 0.043); generalist predator abundance and the acephate spray did not affect predatory mite abundance (F ϭ 1.0, df ϭ 1, P ϭ 0.3; F ϭ 2.3, df ϭ 1, P ϭ 0.14, respectively). This analysisindicatesthatcompetition with generalist predatorsfor spidermite prey may have had a greater inßuence on predatory mite populationsthan did direct predation. This result was in contrast to the small cage experiment where generalist predator additions had an impact on predatory mite abundance but spider mite abundance did not.

Fig. 3. Generalist predator and predatory mite abun- Discussion dance in the predator compatibility, acephate, open-plot ϩ In this study, our Þrst objective was to determine if experiment. (A) Mean ( SE) immature predator abundance commercially reared western predatory mites could in plots that were sprayed with acephate and in unsprayed plots.Predatorsincluded big-eyed bugs( Geocoris spp.), establish and increase their population size on cotton minute pirate bugs( Orius tristicolor), damsel bugs (Nabis if generalist predators were removed. We found in our spp.), lacewings (chrysopid spp.), and coccinellid spp. (listed Þrst experiment that G. occidentalis could establish in order of abundance). Samples consisted of 20 sweeps of and increase its population size by 60% over 18 d on the plant canopy using a sweep net. Samples were collected cotton under conditionsof high spidermite availabil- 2 wk follow the acephate spray. (B) Mean (ϩ SE) cumulative ity and low generalist predator abundance. This result predatory mite abundance in acephate sprayed and un- is of interest because earlier Þeld releases of G. occi- sprayed plots. Mite abundance was summed from the four dentalis at low release rates were not successful in samplingdates(25 leavesper sample)after the acephate establishing predatory mite populations (R. G. Colfer, application. J. A. Rosenheim, L. D. Godfrey, C. L. Hsu, unpublished data). For our second objective, we evaluated the populationsof predatory mitesand spidermites.The impact that naturally occurring generalist predators, most likely reason that spider mite populations in- including Geocoris spp., O. tristicolor, and F. occiden- creased after the acephate spray is that they experi- talis, have on western predatory mite establishment enced less predation pressure from generalist preda- and on spider mite control. We examined the impact tors. Predatory mite populations might have increased of these generalist predators at two scales: in a well- after sprays either because they experienced less pre- controlled small-scale manipulative experiment em-

Table 3. Density (mean ؎ SE) of generalist predators associated with spider mites over the duration of the acephate open-plot experiment

2 wk after spray 4 wk after spray 6 wk after spray Treatments Insecticide Control Insecticide Control Insecticide Control Geocoris spp.a 1.8 Ϯ 0.3 5.0 Ϯ 0.5*** Ð Ð 5.1 Ϯ 1.0 10.9 Ϯ 0.7*** Orius tristicolora 1.2 Ϯ 0.3 0.7 Ϯ 0.2 NS Ð Ð 14.0 Ϯ 1.6 6.2 Ϯ 0.7*** F. occidentalisb 20.4 Ϯ 2.2 38.9 Ϯ 7.1* 135.2 Ϯ 16.0 73.9 Ϯ 7.6*** 65.1 Ϯ 7.7 45.5 Ϯ 2.6**

Two-tailed Wilcoxon rank-sum tests: *, P Ͻ 0.05; **, P Ͻ 0.01; ***, P Ͻ 0.001; NS, not signiÞcant. a Geocoris and Orius abundancesare from sweepsamples(reportedasnumber of predatorsper 20 sweeps). b F. occidentalis abundance isfrom leaf samples(reportedasthripsper 25 leaves). August 2003 COLFER ET AL.: G. occidentalis-GENERALIST PREDATOR INTERACTIONS 847

Fig. 4. Population dynamicsof predatory mites(mean ϩ SE) in plots in which generalist predators were suppressed with an acephate application and in which western predatory mites were released in the acephate, open-plot experiment. Arrows indicate when plotsreceived an acephate application and/or a predatory mite release. ploying cages, and in a less-controlled larger-scale Geocoris spp. and O. tristicolor, can have negative manipulative experiment done with open plots. We impactson G. occidentalis populations. In our cage found that when western predatory mites were com- experiment, the addition of Geocoris spp. and O. tris- bined with the hemipteran predators Geocoris spp. or ticolor reduced predatory mite densities to 23% and O. tristicolor, their densities were greatly reduced. The 0%, respectively, compared with the density observed outcome was similar in the insecticide manipulation in the predatory mite alone treatment. In our insec- experiment; western predatory mites nearly tripled ticide manipulation experiment, predatory mite abun- their abundance when generalist predators were dance wasnearly three timesgreater in plotsin which chemically suppressed. In the cage experiments, in generalist predators were chemically suppressed com- which insect predator exclusion was nearly complete, pared with control plots. The negative effects that predatory mites could suppress spider mites when Orius spp. can have on predaceous phytoseiid mites release rates were high (Table 1, Fig. 2B). However, has been studied in the laboratory (Gillespie and Quir- in the open-plot experiment, in which predatory in- ing 1992, Cloutier and Johnson 1993, Wittmann and sect exclusion was only partial, predatory mite releases Leather 1997), and in greenhouse nurseries (Ramak- had no impact on spider mite densities (Fig. 5). The ers1993, Brodsgaardand Enkegaard 1997) where both strong negative effect that generalist predators had on groupsare usedasbiological control agents.However, western predatory mites did not, however, disrupt we know of no studies that have documented preda- spider mite biological control; the hemipteran pred- tory interactionsbetween Orius and phytoseiid mites atorscompensatedfor their impact on G. occidentalis under Þeld conditionsor predatory interactionsbe- by consuming spider mites themselves (Fig. 2B). Gen- tween Geocoris and phytoseiids under any conditions. eralist predators effectively suppressed spider mites Both of these hemipteran predators are very common compared with treatmentswhere they were excluded. in cotton and other row cropsand are important nat- Impact of Hemipteran Predation on Western Pred- urally occurring spider mite predators (Wilson et al. atory Mite Populations. In thisstudy,we demon- 1991a). strated that generalist hemipteran predators, such as In contrast to the hemipteran predators, western ßower thrips( F. occidentalis) did not appear to have negative effects on western predatory mites. This re- Table 4. Predatory phytoseiid mite species composition over sult is in contrast to the impact that F. occidentalis has the duration of the acephate open-plot experiment for all treatments on spider mite population via egg predation (Trichilo and Leigh 1986, Agrawal et al. 1999, Agrawal and Klein Galendromus Neoseiulus Neoseiulus Sample 2000, R. G. Colfer and A. A. Agrawal, unpublished occidentalis aurescens fallacis size data). Western ßower thrips have been shown to con- 2 wk after spray 44% 56% 0% n ϭ 25 sume predatory phytoseiid mite eggs under laboratory 4 wk after spray 95% 3% 2% n ϭ 61 conditions(Colfer et al. 1998, Roda et al. 2000). Per- 6 wk after spray 89% 4% 7% n ϭ 28 hapspredatory miteswere able to avoid egg predation Total 82% 15% 3% n ϭ 114 by thripsby laying their eggsin closeproximity to 848 ENVIRONMENTAL ENTOMOLOGY Vol. 32, no. 4

Fig. 5. Population dynamicsof spidermites(mean ϩ SE) in plots in which generalist predators were suppressed with an acephate application and in which western predatory mites were released in the acephate, open-plot experiment. Arrows indicate when plotsreceived an acephate application and/or a predatory mite release. spider mite webbing and cotton leaf trichomes that are second species, known as intraguild predation (Polis common on Acala cotton (the type of cotton most et al. 1989, Rosenheim et al. 1995). Second, one pred- grown in the San Joaquin Valley). Roda et al. (2000) ator species can have a detrimental effect on a second found that predatory mite egg predation asa resultof predator species by depleting a resource that the sec- westernßower thripswassigniÞcantlyreduced when ond predator requires. This type of interaction is eggswere located on leaveswith artiÞcial trichromes known asexploitative competition (Schoener 1983). or spider mite webbing versus on trichome-free Hemipteran predatorscould have negatively inßu- leaves. Another explanation is that predatory mite enced predatory mite populationsby meansof intra- eggs may be a suboptimal food item for western ßower guild predation, exploitative competition, or both of thripsthat they chosenot to feed on when other food these mechanisms. In the cage experiment, our anal- sources are available such as spider mite eggs and ysisindicatedthat intraguild predation wasthe more pollen. Also, our experimental approaches did not important mechanism by which western predatory completely exclude thripsfrom experimental units mite populations were suppressed. However, in the (cages, open-plots). Thrips contamination limits our large-scale insecticide experiment, our analysis sug- ability to evaluate the impact of thripson western gested that exploitative competition for spider mites predatory mites. may have been the more important mechanism caus- Each of the experimental approachesthat we used ing the reduction in western predatory mite popula- to evaluate the compatibility of predatorshaslimita- tions. As a caveat, the use of ANCOVA to determine tions. The cage experiments contained fairly small whether predation or competition wasmore impor- populationsof arthropodsthat were conÞned within tant to predatory mite populationshaslimitations.In the enclosures. In the open-plot experiment, in addi- systems with intraguild predation, it is difÞcult to tease tion to our desired effect of reducing generalist pred- apart statistically the effects of predation and compe- ator populations, the acephate application could have tition, because the intensity of both of these forces inßuenced the arthropod communitiesin other ways. may be mediated by the availability of the shared Because we observed similar outcomes using these resource (Schoener 1983, Polis et al. 1989). very different experimental approaches, we believe There are four major differencesbetween the cage the negative impact of hemipteran predatorson the experiment and large-scale insecticide experiment westernpredatory mite populationsisreal. Generalist that may explain why intraguild predation played the hemipteran predatorsare common in many annual more important role in the cage experiment and ex- cropping systems, and our Þndings suggest that the ploitative competition played a more important role in augmentation of predatory mitesor other biological the large-scale acephate experiment. First, the differ- control agentscould be generally limited in annual ence in predator communitiesin the cage experiment agroecosystems by these predators. wasdiscrete;hemipteran predatorswere either When one predator hasa negative impact on an- present or absent from cages, depending on the treat- other predator, there are two major mechanisms that ment. In contrast, the exclusion of hemipteran pred- may be operating. First, one predator can consume the atorsin the large-scaleacephate experiment waspar- August 2003 COLFER ET AL.: G. occidentalis-GENERALIST PREDATOR INTERACTIONS 849 tial and occurred during a single event (one acephate ator interactionscausereductionsin the efÞcacy of application). Bare soil surrounding the experimental biological control (Croft and MacRae 1993, Rosen- plotswasthe only barrier againstgeneralistpredators heim et al. 1993, Cisneros and Rosenheim 1997, Rosen- reestablishing in sprayed plots. Thus, the difference in heim 2001, Snyder and Wise 2001, Snyder and Ives potential predation pressure on western predatory 2001). Indeed, in the apple agroecosystem, predation mitesbetween treatmentswhere hemipteran preda- by Zetzellia mali on G. occidentalis can disrupt control tors were present or absent was more distinct in the of Panonychus ulmi (Croft and MacRae 1992a). How- cage experiment than the acephate experiment. The ever our results are in agreement with other systems cage experiment was, therefore, more appropriately with high levelsof intraguild predation. In both early- designed for evaluating predation as a factor limiting season cotton and alfalfa, aphid suppression was im- western predatory mite populations. Second, densities proved by combining predatorswith aphid parasitoids of spider mites used in the cage experiment were compared with systems with aphid parasitoids alone, representative of high density Þeld conditions (R.G. even though predatorsreadily fed on immature para- Colfer and J.A. Rosenheim, unpublished data). In the sitoids (Colfer and Rosenheim 2001, Snyder and Ives large-scale acephate experiment, initial spider mite 2002). Thus, in systems with multiple predators, in- densities were low, compared with their densities in traguild predation may have a variety of impactson the cage experiment (acephate experiment: 6.4 per herbivore biological control: herbivore suppression leaf; cage experiment: 147 mitesper leaf). The lower may be disrupted, unchanged, or improved by com- densities of spider mites in the acephate experiment bining predator species. In the case of augmentative likely increased the importance of exploitative com- biological control, releases may not be economically petition (Schoener 1983). Predatory mite abundance feasible if the released natural enemies do not estab- may have been more limited by spider mite prey lish populations for the duration of the growing sea- availability than by predation. Third, the duration of son, even if they improve herbivore suppression over the large-scale acephate experiment was much longer the short term. than the cage experiment (56 d versus 7 d). Theory In our open-plot experiment, we found that predictsthat effectsfrom indirect interactionssuchas acephate substantially reduced generalist predator exploitative competition should take a longer time to densities and caused spider mite densities to increase observe than direct effects (Bender et al. 1984, Yodzis 4.2Ð9.8 times above mite densities in unsprayed plots. 1988). The longer duration of the acephate experi- There are several potential causes of such a secondary ment may have made thisexperiment more conducive outbreak of spider mites. One of the best supported for detecting exploitative competition compared with explanationsfor thisobservationisthat insecticides the cage experiment. It should be noted, however, that decimate naturally occurring generalist predators, yet reviewsof experimental manipulationsof communi- cause little mortality to spider mites (Gonzalez et al. tieshave indicated that indirect effectstend to take 1982, Trichilo and Leigh 1986, Wilson et al. 1991a, Sclar equal or only slightly more time to detect than do et al. 1998, this study). When spider mite populations direct effects(Schoener 1993; Menge 1995, 1997). are no longer limited by predation they can expand Fourth, the conÞnement of the cagescould potentially and cause severe foliar damage to cotton plants (Wil- have increased the likelihood that organisms would son et al. 1991b). Another potential reason for sec- interact with each other. However, more recent ex- ondary outbreaksof mitesisthat broad spectrumin- perimentsthat allowed predator movement and insecticides may cause spider mite population growth creased cage size produced very similar results as the ratesto increaseeither directly or indirectly by mod- predator compatibility cage experiment described ifying plant quality (Leigh and Wynholds1980). Al- above (R. G. Colfer and J. A. Rosenheim, unpublished though both of these factors may contribute to spider data), indicating that thisfactor islessimportantthan mite outbreaks, simulation analysis of this system has the others. shown that the reduction in predator abundance is Inﬂuence of Predator-Predator Interactions on Spi- more important (Trichilo and Wilson 1993). In this der Mite Control. Although the simultaneous addition study we employed several techniques to evaluate of predatory mitesand hemipteran predatorshad neg- natural enemy impacts on spider mites. Because the ative effects on predatory mite establishment, it did magnitude of the impact of generalist predators on not interfere with spider mite suppression. Indeed, the spidermite populationswassimilarwhether predators best spider mite suppression in the small-scale cage were excluded with cages or with insecticides, we experiment wasin the O. tristicolor ϩ predatory mite conclude that the accelerated spider mite population treatment. Resultsfromthisstudyaswell asfrom growth after the acephate treatment wasprimarily a larger-scale, longer-term exclusion cage experiments result of the suppression of generalist predators. (R. G. Colfer and J. A. Rosenheim, unpublished data), Factors Inﬂuencing Western Predatory Mite Estab- indicate that although intraguild predation by lishment in the Cotton Agroecosystem. The outcome hemipteran predatorson westernpredatory mitespre- of the experiment evaluating predatory mite estab- vented predatory mite populationsfrom reaching high lishment under conditions of high spider mite avail- abundance, it does not appear to disrupt spider mite ability and low generalist predator abundance differed biological control. in important waysfrom the resultsofan earlier study These results are in contrast to some other arthro- of predatory mite releases in grower Þelds, in which pod systems where the occurrence of predator-pred- we failed to obtain establishment of the released mites 850 ENVIRONMENTAL ENTOMOLOGY Vol. 32, no. 4

(R. G. Colfer, J. A. Rosenheim, L. D. Godfrey, C. L. substantially reduced the abundance of an augmen- Hsu, unpublished data). First, in this study, using high tatively released predatorÑthe western predatory release rates, we were able to greatly increase the mite. Despite the negative effect of these hemipteran predatory mite population compared with the natu- predatorson G. occidentialis, the presence of these rally-occurring background population. Thiseffect predatorsimproved spidermite control. No effect of wasprobably causedby our experimental preparation F. occidentalis was observed on western predatory of replicates (spraying of insecticidal soap and brush- mite populations. Under conditions of high spider mite ing of plant material), which may have reduced the availability, low generalist predator abundance, and background level of predatory mitesin theserepli- high release rates, western predatory mite releases cates. Second, the released predatory mite popula- increased predatory mite populations; these populations increased at least 60% in size over 18 d when tions subsequently increased their abundance through afforded high spider mite availability and low preda- reproductive recruitment and suppressed spider mite tion risk(thiscalculation excludespredatory mite populations. However, a lower release rate in an open- eggs, so the actual population increase was likely to be plot experiment did not increase predatory mite abun- well above 60%). Thisresultisimportant in that it dance. Thisstudyindicatesthat augmentative releases demonstrates that G. occidentalis can build up popu- of G. occidentalis may be limited by naturally occur- lationsin cotton. Predatory mite densitieswerehighly ring generalist predators in the cotton agroecosystem. correlated with spider mite densities, suggesting that Also, our results suggest that there can be intraguild predatory mites showed the greatest recruitment on predation between natural enemieswithout disrupt- plantsthat had greater availability of prey. Whether G. ing biological control. occidentalis mites need these high spider mite densi- tiesto reproduce in cotton remainsunknown at this time. Acknowledgments It isunclear why releasesofpredatory mitesin the We thank TobiasGlik, Paola Santilla´n, Gail Siu, Suset- open-plot, acephate experiment did not boost preda- teVillanueva, and Elmer Yee for research assistance. We tory mite abundance in the sprayed plots, in which thank John Peterson and the staff at the UC Shafter Research generalist predator abundance had been reduced. Extension Center, the staff at the UCD Plant Pathology Field- There are three potential bases for this result. First, house, and the staff at the UCD Agronomy Field Shop. This release rates in the open-plot experiment were rela- researchwaspartially funded by grantsfrom the California tively modest (37,500 mites per ha or Ϸ0.5 per plant) State Support Board of Cotton Incorporated, the Department compared with the cage experiment (78 mitesper of Pesticide Regulation (Cal EPA), the California Cotton plant). The combination of post release mortality of Pest Control Board, and the U.S. Environmental Protection Agency (STAR Fellowship Program). predatory mitesand releaseratesthat were slightly less than the resident population of predator mites could help explain why the releases did not increase References Cited predatory mite populations(residentpredatory mite density at the time of the release: 0.74 mites per plant). Agrawal, A. A., and C. N. Klein. 2000. What omnivoreseat: Second, whereasthe exclusionof predatorsfrom the direct effects of induced plant resistance on herbivores cage experiment wasnearly complete, the acephate and indirect consequences for diet selection by omni- vores. J. Anim. Ecol. 69: 525Ð535. treatment produced only a partial and temporary sup- Agrawal, A. A., C. Kobayashi, and J. S. Thaler. 1999. Inßu- pression of hemipteran predators. Thus, hemipteran ence of prey availability and induced host-plant resis- predatorsmay have continued to imposepredation tance on omnivory by western ßower thrips. Ecology 80: risks on G. occidentalis. We released western preda- 518Ð523. tory mites2 wk after the acephate application, after Barbosa, P. 1998. Conservation biological control. Aca- our leaf assay showed that the acephate was no longer demic, San Diego, CA. toxic to western predatory mites. At the time of the Bender, E. A., T. J. Case, and M. E. Gilpin. 1984. Perturba- release,it islikely that the acephate wasno longer tion experimentsin community ecology: Theory and toxic to hemipteran predatorsand their populations practice. Ecology 65: 1Ð13. Brodsgaard, H. F., and A. Enkegaard. 1997. Interactions were reestablishing in these plots. Third, our Þeld site among polyphagousanthocorid bugsusedfor thripscon- may have had greater densities of G. occidentalis than trol and other beneÞcialsin multi-speciesbiologicalpest generally observed in cotton, because it was down- management systems. Rec. Res. Devel. Entomol. 1: 153Ð wind from an almond orchard that could have been a 160. source of G. occidentalis. Western predatory mites are Cisneros, J. J., and J. A. Rosenheim. 1997. Ontogenetic known to develop high densities and to disperse ae- change of prey preference in a generalist predator, Zelus rially over large distancesinalmond orchards(Hoy renardii, and itsinßuence on the intensityof predator- 1982, Grafton-Cardwell et al. 1991). Indeed, predatory predator interactions. Ecol. Entomol. 22: 399Ð407. mite abundance declined with increasing distance Cloutier, C., and S. G. Johnson. 1993. Predation by Orius tristicolor (Hemipteran: Anthocoridae) on Phytoseiulus from the almond orchard (cumulative predatory mites 2 ϭ ϭ ϭ ϭ persimilis (Acarina: Phytoseiidae): testing for compati- for June sample dates: r 0.16, F 7.4, df 1, P bility between biocontrol agents. Environ. Entomol. 22: 0.01). 477Ð482. In summary, we found that the naturally occurring Colfer, R. G., J. A. Rosenheim, L. D. Godfrey, and C. L. Hsu. generalist predators Geocoris spp. and O. tristicolor 1998. Evaluation of predaceousmite releasesforspider August 2003 COLFER ET AL.: G. occidentalis-GENERALIST PREDATOR INTERACTIONS 851

mite management. Proc. Beltwide Cotton Conf. 2: 976Ð Louda, S. M., D. Kendall, J. Connor, and D. Simberloff. 1998. 982. Ecological effectsof an insectintroduced for the biolog- Colfer, R. G., and J. A. Rosenheim. 2001. Predation on im- ical control of weeds. Science 277: 1088Ð1090. mature parasitoids and its impact on aphid suppression. MacRae, I. V., and B. A. Croft. 1996. Differential impact of Oecologia (Berl.). 126: 292Ð304. egg predation by Zetzellia mali (Acari: Stigmaeidae) on Croft, B. A. 1994. Biological control of apple mitesby a Metaseiulus occidentalis and Typhlodromus pyri (Acari: phytoseiid mite complex and Zetzellia mali (Acari: Stig- Phytoseiidae). Exp. Appl. Acarol. 20: 143Ð154. maeidae): long term effectsand impact of azinphos- McMurtry, J. A. 1982. The use of phytoseiids for biological methyl on colonization by Amblyseius andersoni (Acari: control: progress and future prospects, pp. 23Ð48. In M. A. Phytoseiidae). Environ. Entomol. 23: 1317Ð1325. Hoy [ed.], Recent advancesin knowledge of the phyto- Croft, B. A., and M. B. Croft. 1996. Intra- and interspeciÞc seiidae. University of California Press, Berkeley, CA. predation among adult female phytoseiid mites (Acari: Menge, B. A. 1995. Indirect effectsin marine rocky inter- Phytoseiidae): effects on survival and reproduction. En- tidal interaction webs: patterns and importance. Ecol. viron. Entomol. 25: 853Ð858. Monogr. 65: 21Ð74. Croft, B. A., S. S. Kim, and D. I. Kim. 1996. Intra- and Menge, B. A. 1997. Detection of direct versus indirect ef- interspeciÞc predation on four life stage groups by the fects: were experiments long enough? Am. Nat. 149: 801Ð adult femalesof Metaseiulus occidentalis, Typhlodromus 823. pyri, Neoseiulus fallacis and Amblyseius andersoni. Exp. Neter, J., W. Wasserman, and M. H. Kutner. 1990. Applied Appl. Acarol. 20: 435Ð444. linear statistical models, 3rd ed. Richard D. Irwin, Inc., Croft, B. A., and I. V. MacRae. 1992a. Persistence of Typhlo- Homewood, IL dromus pyri and Metaseiulus occidentalis (Acari: Phyto- Nyrop, J., G. English-Loeb, and A. Roda. 1998. Conservation seiidae) on apple after inoculative release and competi- biological control of spider mites in perennial cropping tion with Zetzellia mali (Acari: Stigmaeidae). Environ. systems, pp. 307Ð333. In P. Barbosa [ed.], Conservation of Entomol. 21: 1168Ð1177. natural enemies of pest species. Academic, San Diego, Croft, B. A., and I. V. MacRae. 1992b. Biological control of CA. apple mitesby mixed populationsof Metaseiulus occiden- Osman, A. A., and G. Zohdi. 1976. Suppression of the spider talis (Nesbitt) and Typholodromus pyri Scheuten (Acari: miteson cotton with massreleasesof Amblyseius gossipi Phtyoseiidae). Environ. Entomol. 21: 202Ð209. (El Badry). Z. Ang. Entomol. 81: 245Ð8. Croft, B. A., and I. V. MacRae. 1993. Biological control of Parrella, M. P., K. M. Heinz, and L. Nunney. 1992. Biological apple mites: impact of Zetzellia mali (Acari: Stigmaeidae) control through augmentative releases of natural ene- on Typhlodromus pyri and Metaseiulus occidentalis (Acari: mies: a strategy whose time has come. Am. Entomol. 38: Phytoseiidae). Environ. Entomol. 22: 865Ð873. 172Ð179. Flaherty, D. L., L. T. Wilson, V. M. Stern, and H. Kido. 1985. Pickett, C. H. and R. L. Bugg [eds.]. 1998. Enhancing bio- Biological control in San Joaquin Valley vineyards, pp. logical control: habitat management to promote natural 501Ð520. In M. A. Hoy and D. C. Herzog [eds.], Biological enemies of agricultural pests. University of California control in agricultural IPM systems. Academic, Orlando, Press, Berkeley, CA. FL. Pickett, C. H., and F. E. Gilstrap. 1986. Inoculative releases Gillespie, D. R., and D.J.M. Quiring. 1992. Competition be- of phytoseiids (Acari) for the biological control of spider tween Orius tristicolor (White) (Hemipteran:Anthroco- mites(Acari: Tetranychidae) in corn. Environ. Entomol. ridae) and Amblyseius cucumeris (Oudemans) (Acari: 15: 790Ð794. Phytoseiidae) feeding on Frankliniella occidentalis (Per- Pickett, C. H., F. E. Gilstrap, R. K. Morrison, and L. F. Bouse. gande) (Thysanoptera: Thripidae). Can. Entomol. 124: 1987. Release of predatory mites (Acari: Phytoseiidae) 1123Ð1128. by aircraft for the biological control of spider mites Gonzalez, D., B. R. Patterson, T. F. Leigh, and L. T. Wilson. (Acari: Tetranychidae) infesting corn. J. Econ. Entomol. 1982. Mites: a primary food source for two predators in 80: 906Ð910. San Joaquin cotton. Calif. Agric. 36: 18Ð20. Polis, G. A., C. A. Myers, and R. D. Holt. 1989. The ecology Grafton-Cardwell, E., J. Granett, and S. M. Normington. and evolution of intraguild predation: potential compet- 1991. Inßuence of dispersal from almonds on the popu- itorsthat eat each other. Annu. Rev. Ecol. Syst.20: 297Ð lation dynamics and acaricide resistance frequencies of 330. spider mites infesting neighboring cotton. Exp. Appl. Ac- Ramakers, P.M.J. 1993. Coexistence of two thrips predators, arol. the anthocorid Orius insidiosus and the phytoseiid Am- Helle, W., and M. W. Sabelis. 1985. World crop pests: spider blyseius cucumeris on sweet pepper. Bulletin OILB/SROP mitestheir biology, natural enemiesand control, vol. 2b. 16: 133Ð136. Elsevier Publishers B. V., Amsterdam, The Netherlands. Roda, A., J. Nyrop, M. Dicke, and G. English-Loeb. 2000. Howarth, F. G. 1991. Environmental impacts of classical Trichomesand spider-mitewebbing protect predatory biological control. Annu. Rev. Entomol. 36: 485Ð509. mite eggsfrom intraguild predation. Oecologia (Berl.) Hoy, M. A. 1982. Aerial dispersal and Þeld efÞcacy of a 125: 428Ð435. genetically improved strain of the spider mite predator Rosenheim, J. A. 1998. Higher-order predatorsand the reg- Metaseiulus occidentalis. Entomol. Exp. Appl. 32: 205Ð212. ulation of insect herbivore populations. Annu. Rev. En- Hoy, M. A., W. Barnett, and W. D. Reil. 1982. Large scale tomol. 43: 421Ð447. releases of pesticide resistant spider mite predators. Calif. Rosenheim, J. A. 2001. Source-sink dynamics for a generalist Agric. 35: 8Ð10. insect predator in a habitat with strong higher-order pre- Leigh, T. F., V. L Maggi, and L. T Wilson. 1984. Develop- dation. Ecol. Monogr. 71: 93Ð116. ment and use of a machine for recovery of arthropods Rosenheim, J. A., H. K. Kaya, L. E. Ehler, J. J. Marios, and B. A. from plant leaves. Environ. Entomol. 77: 271Ð276. Jaffee. 1995. Intraguild predation among biological-con- Leigh, T. F., and P. F. Wynholds. 1980. Do insecticides trig- trol agents: theory and evidence. Biol. Control. 5: 303Ð335. ger spider mite reproduction? Calif. Ariz. Cotton 15: 19Ð Rosenheim, J. A., L. R. Wilhoit, and C. A. Armer. 1993. 21. Inßuence of intraguild predation among generalist insect 852 ENVIRONMENTAL ENTOMOLOGY Vol. 32, no. 4

predators on the suppression of an herbivore population. lative releases of predatory mites Metaseiulus occidentalis Oecologia (Berl.). 96: 439Ð449. and Amblyseius californicus (Acari: Phytoseiidae) in the SAS Institute. 2000. JMP statistical discovery software for San Joaquin Valley of California. Ph. D. dissertation, Uni- Microsoft Windows, version 3.2.2. SAS Institute, Cary, versity of California, Davis, CA. NC. Trichilo, P. J., and T. F. Leigh. 1986. Predation on spider Schausberger, P., and B. A. Croft. 2000. Cannibalism and mite eggs by the western ßower thrips, Frankliniella oc- intraguild predation among phytoseiid mites: are aggres- cidentalis (Thysanoptera: Thripidae), an opportunist in a siveness and prey preference related to diet specializa- cotton agroecosystem. Environ. Entomol. 15: 821Ð825. tion? Exp. Appl. Acarol. 24: 709Ð725. Trichilo, P. J., and L. T. Wilson. 1993. An ecosystem analysis Schausberger, P., and A. Walzer. 2001. Combined versus of spider mite outbreaks: physiological stimulation or single species release of predaceous mites: predator- natural enemy suppression. Exp. Appl. Acarol. 17: 291Ð predator interactions and pest suppression. Biol. Control 314. 20: 269Ð278. van den Bosch, R., and K. S. Hagen. 1966. Predaceousand Sclar, D. C., D. Gerace, and W. S. Cranshaw. 1998. Obser- parasitic arthropods in California cotton Þelds. California vationsof population increasesandinjury by spidermites Agricultural Experiment Station Bulletin 820. (Acari: Tetranychidae) on ornamental plantstreated Van Driesche, R. G., and T. S. Bellows. 1996. Biological with imidacloprid. J. Econ. Entom. 91: 250Ð255. control. Chapman & Hall, Inc., New York. Schoener, T. W. 1993. On the relative importance of direct Von Ende, C. N. 1993. Repeated-measures analysis: growth versusindirecteffectsin ecological communities,pp. and other time-dependent measures, pp. 113Ð137. In S. M. 365Ð411. In H. Kawanabe, J. E. Cohen, and K. Iwasaki Scheiner and J. Gurevitch [eds.], Design and analysis of [eds.], Mutualism and community organization: behav- ecological experiments. Chapman & Hall Inc., New York. ioral, theoretical and food web approaches. Oxford Uni- Walde, S. J., J. M. Hardman, and C. N. Magagula. 1997. versity Press, Oxford, United Kingdom Direct and indirect speciesinteractionsinßuencingwith- Simberloff, D., and P. Stiling. 1996. How risky is biological in-season dynamics of apple rust mite, Aculus schlechten- control? Ecology 77:1965Ð1974. dali (Acari: Eriophyidae). Exp. Appl. Acarol. 21: 587Ð614. Sokal, R. R., and F. J. Rohlf. 1995. Biometry, 3rd ed. W. H. Wilson, L. T., P. J. Trichilo, and D. Gonzalez. 1991a. Natural Freeman and Company, New York. enemiesof spidermites(Acari: Tretranychidae) on cot- Snyder, W. E., and A. R. Ives. 2001. Generalist predators ton: density regulation or casual association? Environ. disrupt biological control by a specialist parasitoid. Ecol- Entomol. 20: 849Ð856. ogy 82: 705Ð716. Wilson, L. T., Trichilo, P. J., and Gonzalez, D. 1991b. Spider Snyder, W. E., and A. R. Ives. 2002. Interactionsbetween mite (Acari: Tetranychidae) infestation rate and initia- specialist and generalist natural enemies: parasitoids, tion: effect on cotton yield. J. Econ. Entomol. 84: 593Ð600. predators, and pea aphid biological control. Ecology 84: Wittmann, E. J., and S. R. Leather. 1997. Compatibility of 91Ð107. Orius laevigatus Fieber (Hemiptera: Anthocoridae) with Snyder, W. E., and D. H. Wise. 2001. Contrasting trophic Neoseiulus (Amblyseius) cucumeris Oudemans(Acari: cascades generated by a community of generalist pred- Phytoseiidae) and Iphiseius (Amblyseius) degenerans ators. Ecology 82: 1571Ð1583. Berlese (Acari: Phytoseiidae) in the biological control of Sunderland, K. D., J. A. Axelsen, K. Dromph, B. Freier, J. L. Frankliniella occidentalis Pergande (Thysanoptera: Hemptinne, N. H. Holst, P.J.M. Mols, M. K. Petersen, W. Thripidae). Exp. Appl. Acarol. 21: 523Ð538. Powell, P. Ruggle, H. Triltsch, and L. Winder. 1997. Pest Yodzis, P. 1988. The indeterminancy of ecological interac- control by a community of natural enemies. Acta Jut- tions. Ecology 69: 508Ð515. landica. 72: 271Ð326. Tijerina-Chavez, A. D. 1991. Biological control of spider Received for publication 18 January 2002; accepted 22 April mites(Acari: Tetranychidae) on cotton through inocu- 2003.