Journal.Nzpps.Org

Beneficial Insects 221 applied here are aligned with results from the under environmental conditions unique to empirical approach of directly manipulating enclosures. Journal of Apicultural Research Comparing traditional methods of test species stocking rates. In most crops, honey bee and 47: 162-165. selection with the PRONTI tool for host-range testing bumble bee stocking rates are notoriously diﬃcult Evans E 2017. From humble bee to greenhouse to calculate because of the challenge of finding pollination workhorse: can we mitigate risks of Eadya daenerys (Braconidae) single-pollinator sites, and the risk associated for bumble bees? Bee World, 94: 34-41. Toni M. Withers1,*, Jacqui H. Todd2, Belinda A. Gresham1 and Barbara I.P. Barratt3 with pollination failure if experimental stocking Hanan JJ, Holly WD, Goldsberry KL 1978. rates are too low. Theoretical approaches may Greenhouse management. Springer Berlin. 1Scion, Private Bag 3020, Rotorua 3046, New Zealand present a useful option for estimating appropriate Howlett BG, Read SFJ, Jesson LK, Benoist A, 2 e New Zealand Institute for Plant & Food Research Ltd, Private Bag 92169, Auckland Mail stocking rates without directly manipulating Evans LE, Pattemore DE 2017a. Diurnal insect Centre, Auckland 1142, New Zealand pollinator abundance. visitation patterns to ‘Hayward’ kiwifruit 3AgResearch, Invermay Agricultural Centre, Private Bag 50034, Mosgiel 9053, New Zealand flowers in New Zealand. New Zealand Plant *Corresponding author: [email protected] ACKNOWLEDGEMENTS Protection 70: 52-57. The authors would like to thank Colin Limmer Howlett BG, Evans LJ, Pattemore DE, Nelson Abstract A computer-based tool (PRONTI; Priority Ranking Of Non-Target Invertebrates) for orchard access and field support, Beth Kyd WR 2017b. Stigmatic pollen delivery by has been developed to assist the selection of invertebrate species for risk-assessment testing and Mary Black (Zespri Ltd.) for assistance flies and bees: Methods comparing multiple with entomophagous biological control agents (BCAs). PRONTI was used to produce a with experimental setup, and Brad Howlett and species within a pollinator community. Basic prioritised list of taxa for host-range testing with the braconid parasitoid Eadya daenerys, Warrick Nelson (PFR) for thoughtful advice and Applied Ecology 19: 19-25. a potential BCA for the eucalypt pest, Paropsis charybdis. The resulting list was compared on this manuscript. The manuscript further Jones JB 2007. Tomato plant culture: In the field, with a list developed using traditional species selection methods. Seven of the nine species benefitted from the thoughtful comments of greenhouse, and home garden. CRC Press, on the traditional list were in the PRONTI top 20. The remaining two species (Agasicles two reviewers. The project was funded through Boca Raton, Florida, 216 p. hygrophila and Cassida rubiginosa) may not have been selected if the PRONTI tool had been the Ministry of Business, Innovation and Pomeroy N, Fisher RM 2012. Pollination of used. These two species were on the traditional list because they are in the same family as the Employment programme number C11X1309 kiwifruit (Actinidia deliciosa) by bumble bees target and are considered valuable BCAs. Alternative BCAs were prioritised by PRONTI. (‘Bee minus to Bee plus and Beyond: Higher (Bombus terrestris): eﬀects of bee density and The other 13 taxa prioritised by PRONTI were not on the traditional list: the taxa are little- yields from smarter growth-focused pollination patterns of flower visitation., New Zealand known natives in the target subfamily or sister subfamily but larvae are presumed to be systems’) and co-funding support by Zespri Entomologist 25(1): 41-49 root-feeders, whereas target larvae are leaf-feeders. PRONTI can support the traditional Group Limited, Summerfruit New Zealand Pomeroy N, Stoklosinski SR 1990. Measuring the approach by providing transparent evidence to support the selection (or rejection) of non- (SNZ), Avocados NZ and the Foundation for foraging strength of bumble bee colonies. In: target species for host-range testing. Arable Research. Shriva P ed. Proceedings of the XIth Congress of the International Union for the Study of Keywords host-range testing, biosafety, host-testing list, phylogeny, risk assessment. REFERENCES Social Insects: 252-253. Bangalore, India. Alexander, MP 1980. A versatile stain for pollen, Reddy PP 2016. Sustainable crop protection fungi, yeast, and bacteria. Stain Technology under protected cultivation. Springer, 55: 13-18. Singapore, 434 p. INTRODUCTION biological control researchers with a sound Castilla N 2002. Current situation and Biological control agents (BCA) proposed for approach to follow. In general, phylogeny plays future prospects of protected crops in the release in New Zealand must first be assessed a strong part in the species selection process Mediterranean region. Proceedings of the for any risks they may pose to the receiving (Hoddle 2004). In addition to phylogeny, species International Symposium on Mediterranean ecosystem, particularly to non-target species that are valued, such as beneficial BCAs and Horticulture: Issues and Prospects, pp. 135- that could be used as alternative hosts or prey by invertebrates (endemic or exotic) with cultural 147. entomophagous BCAs (Barratt & Moeed 2005). or aesthetic significance, are also recommended Cook R, Calvin L 2005. Greenhouse tomatoes Therefore, selecting the non-target species to test to be included in test species lists (Kuhlmann change the dynamics of the North American is a very important task given the large number et al. 2006). Often this approach leads to a very fresh tomato industry. Economic Research of native and valued introduced invertebrate large initial list that, for practical reasons, must Report No. ERR-2. United States Department species in New Zealand. Traditional non-target be “filtered” by eliminating species that have of Agriculture. 86 p. https://www.ers.usda. species selection methods have been outlined disparate spatial, temporal and/or morphological gov/publications/pub-details/?pubid=45477 by the Food and Agriculture Organization and characteristics from the target species. However, Dag A 2008. Bee pollination of crop plants reviewed by Kuhlmann et al. (2006) to provide there is always room for interpretation when

©2018 New Zealand Plant Protection Society (Inc.) www.nzpps.org Refer to http://www.nzpps.org/terms_of_use.html New Zealand Plant Protection 71: 221-231 (2018) https://doi.org/10.30843/nzpp.2018.71.183 Beneficial Insects 222 assessing the phylogenetic and ecological considered. This resulted in an initial list of 153 Table 1 The “traditional list” of non-target species for host testing with Eadya daenerys. Species are affinities between the target host (invading pest) potential non-target beetles. To reduce the list to listed in order of phylogenetic relatedness to the target. Status is E = Exotic self-introduced pest, N = and each potential non-target. For instance, a manageable number, the list was filtered using a Native, I = Introduced beneficial biocontrol agent (BCA). Exposed larvae = larvae are known to feed published phylogenies can change over time number of ecological attributes (Kuhlmann et al. externally on the leaf surface. Ranks are included for clarity, and the two native genera Chalcolampra with taxonomic revisions, such as occurred with 2006). As E. daenerys has only one generation per and Allocharis are listed separately, whereas in the original list they were considered so similar that they Senecio species (Compositae) endemic to New year present only during springtime (November- were ranked equally. Recreated from Table 2, page 184 of Withers et al. (2015). Zealand (Paynter et al. 2004), and non-targets December), all species without spring-active, can be overlooked as valued, such as occurred leaf-feeding larvae were excluded and the Rank Species and Status Subfamily Host Primary reason for inclusion with fiddlewood (Citharexylum spinosum L.; remaining endemic beetles with an adult body Verbenaceae) in Australia (Manners et al. 2010). length greater than approximately 5-mm long 1 Trachymela sloanei E Chrysomelinae Eucalyptus spp. Same subfamily Same habitat This traditional approach for drawing (based on the minimum size of the smallest 2 Dicranosterna semipunctata E Chrysomelinae Acacia melanoxylon Same subfamily up a host testing list was followed in 2015 known host, Paropsisterna agricola (Chapuis) Adjacent habitat when New Zealand Eucalyptus (Myrtaceae) were prioritised. This process resulted in a list of 3 Chalcolampra sp. N Chrysomelinae Olearia colensoi Same subfamily growers expressed the need for a BCA to ten non-target species to try to locate for testing target the first generation of the eucalyptus with E. daenerys (Table 1). 4 Allocharis sp. N Chrysomelinae uncertain Same subfamily tortoise beetle pest, Paropsis charybdis Stål The aim of the current research was to 5 Gonioctena olivacea I Chrysomelinae Cytisus scoparius Same subfamily, BCA (Coleoptera: Chrysomelidae) (Withers et al. compare the non-target species list for E. 2015). The candidate BCA was the solitary daenerys produced using the traditional method 6 Chrysolina abchasica I Chrysomelinae Hypericum Same subfamily, BCA androsaemum larval parasitoid Eadya paropsidis Huddleston with the list produced by a new computer-based and Short (Hymenoptera: Braconidae). This tool known as PRONTI (Priority Ranking Of 7 Lochmaea suturalis I Galerucinae Calluna vulgaris Sister subfamily, exposed larvae, BCA species has since been split into three species, Non-Target Invertebrates; (Todd et al. 2015)). 8 Agasicles hygrophila I Galerucinae Alternanthera Sister subfamily, following intensive molecular and field research The PRONTI tool consists of two parts: a philoxeroides exposed larvae, BCA in Tasmania (Ridenbaugh et al. 2018). The agent Microsoft® Access 2013 database (called the Eco 9 Neolema ogloblini I Criocerinae Tradescantia Same family, exposed introduced into containment and evaluated in Invertebase); and a mathematical model that uses uminensis larvae, BCA New Zealand is Eadya daenerys Ridenbaugh the information in Eco Invertebase to produce 10 Cassida rubiginosa I Cassidinae Cirsium arvense Same family, exposed 2018 (Peixoto et al. 2018). a prioritised list of species for testing with the larvae, BCA In 2015, a non-target species list was compiled proposed BCA. Eco Invertebase is used to collate for host range testing with E. daenerys using the information: taxonomy; food web; biomass; traditional approach for entomophagous BCAs ecology; anthropocentric value and testability (Kuhlmann et al. 2006); herein referred to as (i.e. the ability to rear and test the species in control Uraba lugens Walker (Lepidoptera: MATERIALS AND METHODS the “traditional list”. Initially, a list of the closest captivity) on the invertebrate taxa found in New Nolidae) (Barratt et al. 2016; Todd et al. 2016). e candidate BCA: Eadya daenerys relatives to the target paropsine chrysomelid Zealand. It also provides data on the proposed The aim of the current study was to test the The Eco Invertebase was populated with the was made. No endemic species of paropsines BCA and the potential interaction between PRONTI tool prior to completing an actual host- following relevant biological and ecological occur in the New Zealand fauna so the closest the BCA and each taxon. The model combines range assessment for a BCA that was proposed information about the proposed BCA, E. related endemic species to the paropsines from this information (using the selection criteria for release in New Zealand. Of particular daenerys. Eadya daenerys is a solitary larval the Chrysomelinae subfamily were considered, outlined in Kuhlmann et al. (2006)) and converts interest were whether or not: (a) PRONTI endoparasitoid (Rice 2005) and is now known particularly those in the genera Allocharis, the data into a score by which the taxa can be would prioritise different or additional at-risk to be most commonly associated with a small Aphilon, Caccomolpus, Chalcolampra and ranked for host range testing with the proposed non-target species that had not been identified number of species within the genera Paropsis and Cyrtonogetus (Coleoptera: Chrysomelidae) of BCA. PRONTI has been actively tested for its using the traditional approach; and (b) if the tool Paropsisterna (Ps.) (formerly Chrysophtharta), which there are 40 species (Reid 2006). Species ability to improve the selection of invertebrate provided information that may have resulted in together known as paropsines (Peixoto et al. in the Galerucinae were also considered because species for risk assessment of entomophagous the selection of different species from those on 2018). Eadya daenerys is patchily distributed (G. they have been shown to be the phylogenetic BCA with two BCAs previously released in New the traditional list (Table 1). Allen and S. Quarrell pers. comm.) in Tasmania sister group to the Chrysomelinae, and New Zealand. These wereMicroctonus aethiopoides but has been repeatedly collected throughout the Zealand has approximately 100 endemic Loan (Hymenoptera: Braconidae) released to state since 2011, from sea level to 600 m, so is galerucine species (Reid 2014). New Zealand also control Sitona discoideus Gyllenhal (Coleoptera: considered abundant. Adults have been collected has a large number of valued introduced weed Curculionidae), and Cotesia urabae Austin & during the months of November to January with BCAs, and all 13 chrysomelids on this list were Allen (Hymenoptera: Braconidae) released to a peak in December. Adults can fly, and larval Beneficial Insects 223 assessing the phylogenetic and ecological considered. This resulted in an initial list of 153 Table 1 The “traditional list” of non-target species for host testing with Eadya daenerys. Species are affinities between the target host (invading pest) potential non-target beetles. To reduce the list to listed in order of phylogenetic relatedness to the target. Status is E = Exotic self-introduced pest, N = and each potential non-target. For instance, a manageable number, the list was filtered using a Native, I = Introduced beneficial biocontrol agent (BCA). Exposed larvae = larvae are known to feed published phylogenies can change over time number of ecological attributes (Kuhlmann et al. externally on the leaf surface. Ranks are included for clarity, and the two native genera Chalcolampra with taxonomic revisions, such as occurred with 2006). As E. daenerys has only one generation per and Allocharis are listed separately, whereas in the original list they were considered so similar that they Senecio species (Compositae) endemic to New year present only during springtime (November- were ranked equally. Recreated from Table 2, page 184 of Withers et al. (2015). Zealand (Paynter et al. 2004), and non-targets December), all species without spring-active, can be overlooked as valued, such as occurred leaf-feeding larvae were excluded and the Rank Species and Status Subfamily Host Primary reason for inclusion with fiddlewood (Citharexylum spinosum L.; remaining endemic beetles with an adult body Verbenaceae) in Australia (Manners et al. 2010). length greater than approximately 5-mm long 1 Trachymela sloanei E Chrysomelinae Eucalyptus spp. Same subfamily Same habitat This traditional approach for drawing (based on the minimum size of the smallest 2 Dicranosterna semipunctata E Chrysomelinae Acacia melanoxylon Same subfamily up a host testing list was followed in 2015 known host, Paropsisterna agricola (Chapuis) Adjacent habitat when New Zealand Eucalyptus (Myrtaceae) were prioritised. This process resulted in a list of 3 Chalcolampra sp. N Chrysomelinae Olearia colensoi Same subfamily growers expressed the need for a BCA to ten non-target species to try to locate for testing target the first generation of the eucalyptus with E. daenerys (Table 1). 4 Allocharis sp. N Chrysomelinae uncertain Same subfamily tortoise beetle pest, Paropsis charybdis Stål The aim of the current research was to 5 Gonioctena olivacea I Chrysomelinae Cytisus scoparius Same subfamily, BCA (Coleoptera: Chrysomelidae) (Withers et al. compare the non-target species list for E. 2015). The candidate BCA was the solitary daenerys produced using the traditional method 6 Chrysolina abchasica I Chrysomelinae Hypericum Same subfamily, BCA androsaemum larval parasitoid Eadya paropsidis Huddleston with the list produced by a new computer-based and Short (Hymenoptera: Braconidae). This tool known as PRONTI (Priority Ranking Of 7 Lochmaea suturalis I Galerucinae Calluna vulgaris Sister subfamily, exposed larvae, BCA species has since been split into three species, Non-Target Invertebrates; (Todd et al. 2015)). 8 Agasicles hygrophila I Galerucinae Alternanthera Sister subfamily, following intensive molecular and field research The PRONTI tool consists of two parts: a philoxeroides exposed larvae, BCA in Tasmania (Ridenbaugh et al. 2018). The agent Microsoft® Access 2013 database (called the Eco 9 Neolema ogloblini I Criocerinae Tradescantia Same family, exposed introduced into containment and evaluated in Invertebase); and a mathematical model that uses uminensis larvae, BCA New Zealand is Eadya daenerys Ridenbaugh the information in Eco Invertebase to produce 10 Cassida rubiginosa I Cassidinae Cirsium arvense Same family, exposed 2018 (Peixoto et al. 2018). a prioritised list of species for testing with the larvae, BCA In 2015, a non-target species list was compiled proposed BCA. Eco Invertebase is used to collate for host range testing with E. daenerys using the information: taxonomy; food web; biomass; traditional approach for entomophagous BCAs ecology; anthropocentric value and testability (Kuhlmann et al. 2006); herein referred to as (i.e. the ability to rear and test the species in control Uraba lugens Walker (Lepidoptera: MATERIALS AND METHODS the “traditional list”. Initially, a list of the closest captivity) on the invertebrate taxa found in New Nolidae) (Barratt et al. 2016; Todd et al. 2016). e candidate BCA: Eadya daenerys relatives to the target paropsine chrysomelid Zealand. It also provides data on the proposed The aim of the current study was to test the The Eco Invertebase was populated with the was made. No endemic species of paropsines BCA and the potential interaction between PRONTI tool prior to completing an actual host- following relevant biological and ecological occur in the New Zealand fauna so the closest the BCA and each taxon. The model combines range assessment for a BCA that was proposed information about the proposed BCA, E. related endemic species to the paropsines from this information (using the selection criteria for release in New Zealand. Of particular daenerys. Eadya daenerys is a solitary larval the Chrysomelinae subfamily were considered, outlined in Kuhlmann et al. (2006)) and converts interest were whether or not: (a) PRONTI endoparasitoid (Rice 2005) and is now known particularly those in the genera Allocharis, the data into a score by which the taxa can be would prioritise different or additional at-risk to be most commonly associated with a small Aphilon, Caccomolpus, Chalcolampra and ranked for host range testing with the proposed non-target species that had not been identified number of species within the genera Paropsis and Cyrtonogetus (Coleoptera: Chrysomelidae) of BCA. PRONTI has been actively tested for its using the traditional approach; and (b) if the tool Paropsisterna (Ps.) (formerly Chrysophtharta), which there are 40 species (Reid 2006). Species ability to improve the selection of invertebrate provided information that may have resulted in together known as paropsines (Peixoto et al. in the Galerucinae were also considered because species for risk assessment of entomophagous the selection of different species from those on 2018). Eadya daenerys is patchily distributed (G. they have been shown to be the phylogenetic BCA with two BCAs previously released in New the traditional list (Table 1). Allen and S. Quarrell pers. comm.) in Tasmania sister group to the Chrysomelinae, and New Zealand. These wereMicroctonus aethiopoides but has been repeatedly collected throughout the Zealand has approximately 100 endemic Loan (Hymenoptera: Braconidae) released to state since 2011, from sea level to 600 m, so is galerucine species (Reid 2014). New Zealand also control Sitona discoideus Gyllenhal (Coleoptera: considered abundant. Adults have been collected has a large number of valued introduced weed Curculionidae), and Cotesia urabae Austin & during the months of November to January with BCAs, and all 13 chrysomelids on this list were Allen (Hymenoptera: Braconidae) released to a peak in December. Adults can fly, and larval Beneficial Insects 224

E. daenerys use its host for only limited dispersal in association with, Eucalyptus spp., including target taxon; (2) the potential degree of exposure economic, social and cultural value (V) of each (the infested host larva drops to the ground to some predatory beetles. (E) of each non-target to E. daenerys; (3) the non-target; and (5) the assessed ability to source pupate). Eadya daenerys can infest all larval 3. Valued Chrysomelidae: we included ten of hypothetical ecological impact (R & S) that may each non-target and to conduct tests (T), Table instars of its paropsine hosts, which range in the 11 established chrysomelid weed BCAs in result from the exposure of the non-target to the 2. To enable the model to do this, each datum in adult size from 8.2 – 11.5 mm long, Ps. agricola New Zealand. hazard posed by E. daenerys; (4) the estimated the Eco Invertebase was automatically assigned a to P. char y b di s , respectively (de Little 1979; 4. Other valued, non-native Coleoptera in the Nahrung & Allen 2004). All Eadya daenerys hosts same environment as P. char y b di s : four of the Table 2 Details of the changes to scores or weights given to questions used to inform each of the criteria are restricted to feeding on Eucalyptus spp., and six coleopteran BCA or non-native natural used in the PRONTI test with Eadya daenerys. The questions listed are only part of the data used to more than one species of host can co-occur in its enemies found in association with Eucalyptus obtain each criterion score (i.e. these changes only aﬀected part of each score). NTS = non-target main habitat of mixed native Australian forests spp. were included. species; BCA = biological control agent. as well as commercial Eucalyptus spp. plantations 5. New Zealand Euphorinae that could be at Criterion Question Reason for change in Tasmania. risk of hybridisation with E. daenerys: we included ten of the 21 taxa in this category: Hazard What is the E. daenerys poses the greatest risk to NTS closely related to (H) phylogenetic the target, so this attribute was given the highest weight in the Model criteria and weightings five native species, three BCAs, and two other separation between calculation of H by multiplying all scores by 2. This was also done Ideally, the PRONTI tool would be used to rank non-native Euphorinae. target and NTS? in the test of PRONTI with C. urabae (Todd et al. 2016). all invertebrate species that could be exposed 6. Invertebrate natural enemies of P. char yb di s : Exposure Does the NTS occur These attributes were given the highest weight (multiplied by 2) to the BCA if it were released in New Zealand. all four known predators, the two known (E) in same community or in the calculation of E when PRONTI was tested with M. In the case of E. daenerys, that would be all parasitoids, plus three spider species found habitat as the target? aethiopoides because of the risk posed to NTS in the same habitat invertebrates found in or near Eucalyptus spp., in association with Eucalyptus spp. were as the target. This was a lower risk for NTS from E. daenerys which are very widespread in parks, gardens and selected. because this agent is not known to attack unrelated species in the same habitat in Australia, so the multiplier was removed. plantations throughout New Zealand. Since such 7. Pests of Eucalyptus spp. that could be a list would be unreasonably large, the following released from competition following control Resilience Five questions on NTS Reducing the weight of attributes used to calculate E also reduced (R) mobility, abundance, the maximum risk score (i.e., H × E). In Equation 1, the NTS ten categories were used to select a set of 127 of P. char y b di s : we included 42 of 75 taxa in host range, dispersal risk score is modified by the R score (i.e. the ability of the NTS invertebrate taxa more likely to be at risk from this category, including three non-native and phenological to mitigate the risk). If the maximum risk score is reduced, the E. daenerys if it were released in New Zealand, chrysomelids, 11 other Coleoptera, and 28 overlap maximum resilience score must also be reduced so that it doesn’t and data on each were entered into the Eco non-coleopteran eucalypt leaf-feeders. have too much weight in the model. Consequently, the scores for Invertebase. This list included most of the 153 8. Chrysomelid pests of other forestry trees that each of these attributes were reduced in this test of PRONTI by dividing them by 2. taxa considered using the traditional approach. could be released from competition following However, where very little information was control of P. char y b di s : only Dicranosterna Status Is the NTS rare The S score indicates how “important” each NTS is to the (S) or are there many ecosystem (i.e., NTS with large biomass and many links to the available for individual species within a genus semipunctata on Acacia melanoxylon documented foodweb foodweb are more “important”). In previous tests of PRONTI, (e.g., the ten species of Allocharis), the data (Fabaccae) fits this category. links? rare NTS were assigned a biomass of 0.001 per m2; however, this were combined into a single entry for the genus. 9. Coleopteran hosts of Euphorinae in New was found to reduce the NTS status score unfairly. Thus, in this Although PRONTI can deal with some data Zealand: we included 21 taxa in this test, rare NTS were assigned a biomass of 0.01 per m2. Also, NTS gaps (see below), combining all available data category because they were already in that have been well studied may have many known foodweb links and obtain a much higher status score than other NTS. into a single entry for the genus increased the the Eco Invertebase, including 11 weevil Consequently, foodweb scores were all divided by 2 to reduce certainty with which these taxa could be ranked hosts of Microctonus spp. (Hymenoptera: their weight in the calculation of S. by PRONTI. Braconidae). Value Is the non-target Rare or endemic NTS often obtain a low S score because of their 1. Native Coleoptera most closely related to P. 10. Chrysomelid pests of non-forestry plants that (V) endemic or rare? predicted low biomass and few known foodweb links. However, charybdis: we selected 16 of the 17 taxa in this could be controlled by E. daenerys: all four these species are highly valued, so questions on rarity and category: 12 Chrysomelinae and Galerucinae non-native paropsine pests in this category endemicity were multiplied by 3 in the calculation of V in this taxa, plus four Chrysomelidae from other were included. test of PRONTI. subfamilies from this category (the genus Testability Can the non-target Criteria S, V and T were designed to be approximately equal Arnomus was excluded because of the lack of The 127 taxa were then ranked by the model (T) be collected, reared, because each was considered an equally important criterion or does an existing for selecting NTS. Examination of these scores in this test available data). that sits within the PRONTI tool by applying protocol for rearing showed that the average T score was higher than the average S 2. Native Coleoptera in the same environment five selection criteria to the data in the Eco exist? and V scores. In previous tests of PRONTI, the attributes used as P. char y b di s : we selected ten of the 38 Invertebase. The criteria were: (1) the potential to measure T were multiplied by 2, so these multipliers were native coleopteran taxa known to feed on, or hazard (H) posed by E. daenerys to each non- removed in this test of PRONTI to bring T in line with S and V. Beneficial Insects 225

criteria. Where data were “unknown” for a taxon, of the five criteria were examined to ensure the PRONTI the middle score of five was always assigned, and data gaps (i.e. all the unknown attributes assigned this was then used to calculate the amount of a score of five) were not having undue influence 42 36 45 56 48 31 52 45 20 28 26 20 49 24 52 36 31 31 62 uncertainty in each species’ ranking (see Todd et on the scores for any of the criteria, or on the final score al. (2015) for more details). PRONTI scores. These attributes may be assigned Testability With each test of PRONTI, the attributes of higher or lower scores when known, so it is 50 17 28 12 17 63 37 45 68 45 68 68 53 83 25 78 78 78 30 score the agent, the characteristics of the agent’s hosts, important to ensure that taxa with many unknown Value and the context into which the agent would be attributes do not obtain the highest or lowest score 3 7 4 4 35 27 33 41 22 31 28 26 25 27 50 24 24 24 released were examined, and changes were made for any criterion. The top 25 non-target species 24 score where the scores or the weights given to certain ranked using the PRONTI tool are provided in Status attributes were more or less relevant to a selection Table 3, along with the rank assigned using the 32 42 47 52 51 41 47 40 44 41 43 40 49 43 48 48 45 45 criterion. For example, where an agent was only traditional method for test species selection. 48 known to attack species that were closely related, The top 20 taxa on the list produced by the score Resilience the data representing that relationship were PRONTI tool (herein referred to as the “PRONTI given a higher hazard weighting, whereas for list”) are diﬀerent to those on the traditional list 27 27 27 29 34 29 35 23 23 27 23 22 29 22 27 26 22 22 27

agents known to attack species in a particular (Table 1), although the lists have seven taxa in score feeding niche, these data were given the highest common. Deciding which taxa to test from the Exposure weighting. Changes made to scores and weights PRONTI list requires interrogation of the data 84 78 78 84 90 78 90 84 84 78 84 85 65 85 84 78 77 77 for E. daenerys are given in Table 2. in the table: the PRONTI tool was developed 78 score To produce the prioritised list of species, as a decision-support system such that the Hazard several questions and therefore several scores information in the list should be used to support were combined to provide a total score for each decisions on the taxa to undergo risk assessment criterion. The scores for each criterion were then with the BCA. For instance, it may be decided combined using the following equation: that taxa with low uncertainty in their ranking (e.g. less than 50%), and where the hazard is high Chrysomelinae Galerucinae Galerucinae Chrysomelinae Subfamily or or Subfamily Family Chrysomelinae Galerucinae Chrysomelinae Chrysomelinae Chrysomelinae Galerucinae Chrysomelinae Chrysomelinae Eumolpinae Chrysomelinae Chrysomelinae Galerucinae Galerucinae Galerucinae PRONTI score = [(H x E)/R ] × (S+V+T) (e.g. a hazard score greater than 75) should be Galerucinae Equation 1 selected, and only one species from each genus should be included in the test species list. Using Where H = hazard (criterion 1); E = exposure these criteria, the following ten Chrysomelidae

(criterion 2); R = resilience (ability of the species to could be selected from the PRONTI list for . N . N . N

mitigate the risk; criterion 3); S = the status of the testing with E. daenerys: Gonioctena olivacea, . N . N

species in the receiving environment (criterion Trachymela spp., Chaetocnema spp., Chrysolina . N 3); V = the value of the species (criterion 4); abchasica, Aphilon spp., Pleuraltica cyanea, and T = the ‘testability’ of the species (criterion Allocharis spp., Longitarsus jacobaeae, Alema 5). More details of scores, weights and criteria spp., Chalcolampra speculifera. Whichever Gonioctena olivacea I Lochmaea suturalis I Phyllotreta undulata N undulata Phyllotreta Species and Origin Species and Dicranosterna semipunctata E Chaetocnema spp sloanei E Trachymela Trachymela catenata E catenata Trachymela Chrysolina abchasica I abchasica Chrysolina Allocharis spp Allocharis Pleuraltica cyanea N Pleuraltica Aphilon spp Aphilon Cyrtonogetus crassus N crassus Cyrtonogetus Eucolaspis brunnea N Chalcolampra speculifera N speculifera Chalcolampra Chrysolina hyperici I hyperici Chrysolina Alema spp Alema Allastena spp Allastena spp Bryobates Longitarsus jacobaeae I and how they are applied and calculated in the species are selected, the data in the table can be to rounded scores their PRONTI by ranked daenerys. Species are Eadya of testing host for PRONTI speciesby listed 25 non-target Top PRONTI tool can be found in Todd et al. (2015). used to justify those decisions in any application 1 19 18 3 4 2 Rank from PRONTI 5 10 9 8 7 6 16 15 14 12 13 11 to release the BCA submitted to the New Zealand 17 Table 3 is (BCA).Subfamily agent beneficialbiocontrol I = Introduced pest, N = Native, self-introduced Origin E = Exotic is number. whole the closest each species. for unknown was that score used the PRONTI calculate data to of the percentage Chrysomelidae. indicates % uncertainty for given Beneficial Insects 227 score, ranging from 0 to 10, with 10 assigned to RESULTS AND DISCUSSION ------7 2 5 1 6 4 those attributes that were most informative to the A number of diagnostic tools were used to check 3 Table 1 Table selection criteria. For example, species known to the model was working correctly and that data had Rank in feed on eucalypt leaves were assigned a score been entered consistently. Firstly, ‘dummy’ high, of 10 for criterion 2 (exposure), while species low and medium risk species were used to check 13 11 32 19 17 17 36 26 45 31 45 57 14 47 11 45 66 18 66 feeding on other parts of eucalypt trees, or on that the scores were being applied appropriately ainty other plant species, were assigned lower scores. in the model and were being ranked as expected % uncert- A number of diﬀerent pieces of data in the Eco (Todd et al. 2015; Barratt et al. 2016). Secondly, the Invertebase were used to inform each of the five scores obtained by the non-target species for each 6660 4386 4416 4739 6371 6481 6388 5913 5094 5131 5270 5352 5819 4803 4805 4973 5020 5089 5020 score

criteria. Where data were “unknown” for a taxon, of the five criteria were examined to ensure the PRONTI the middle score of five was always assigned, and data gaps (i.e. all the unknown attributes assigned this was then used to calculate the amount of a score of five) were not having undue influence 42 36 45 56 48 31 52 45 20 28 26 20 49 24 52 36 31 31 62 uncertainty in each species’ ranking (see Todd et on the scores for any of the criteria, or on the final score al. (2015) for more details). PRONTI scores. These attributes may be assigned Testability With each test of PRONTI, the attributes of higher or lower scores when known, so it is 50 17 28 12 17 63 37 45 68 45 68 68 53 83 25 78 78 78 30 score the agent, the characteristics of the agent’s hosts, important to ensure that taxa with many unknown Value and the context into which the agent would be attributes do not obtain the highest or lowest score 3 7 4 4 35 27 33 41 22 31 28 26 25 27 50 24 24 24 released were examined, and changes were made for any criterion. The top 25 non-target species 24 score where the scores or the weights given to certain ranked using the PRONTI tool are provided in Status attributes were more or less relevant to a selection Table 3, along with the rank assigned using the 32 42 47 52 51 41 47 40 44 41 43 40 49 43 48 48 45 45 criterion. For example, where an agent was only traditional method for test species selection. 48 known to attack species that were closely related, The top 20 taxa on the list produced by the score Resilience the data representing that relationship were PRONTI tool (herein referred to as the “PRONTI given a higher hazard weighting, whereas for list”) are diﬀerent to those on the traditional list to rounded scores their PRONTI by ranked daenerys . Species are Eadya 27 27 27 29 34 29 35 23 23 27 23 22 29 22 27 26 22 22 27

Equation 1 selected, and only one species from each genus E should be included in the test species list. Using Where H = hazard (criterion 1); E = exposure these criteria, the following ten Chrysomelidae

(criterion 2); R = resilience (ability of the species to could be selected from the PRONTI list for . N . N . N

mitigate the risk; criterion 3); S = the status of the testing with E. daenerys: Gonioctena olivacea, . N . N

Environmental Protection Authority (EPA). because they are not common or well distributed nine endemic New Zealand galerucines in ------There were many similarities between the nationwide. From the PRONTI list, L. jacobaeae the top 20 that were not on the traditional list Table 1 Table

Rank in two lists, including four of the species on the could arguably have been selected as a test because the larvae were all thought to be root- traditional list also selected from the PRONTI species: it is a BCA of ragwort in the Galerucinae feeders. This prompted a re-examination of the list using the criteria given above, plus both lists sub family, and although the larvae feed on the scarce literature and collections with much more 11 40 19 18 57 26

ainty include a species of Trachymela. It is perhaps roots, they emerge from eggs on the foliage so care, and a reconsideration of whether or not

% uncert- more interesting to focus on the diﬀerences could be at-risk from parasitism before moving any endemic species may be external leaf-feeders between the two lists. below ground. This species has lower resilience, after all. No additional evidence for external leaf The ten genera or species ranked highest from with only one generation per year in some parts -feeding larvae was uncovered and Lochmaea 4231 4258 4131 4134 4157 4198 score

score subfamilies: Neolema ogloblini and Cassida larval feeding habits of these species, with larvae investigating NTS with less available data, and Testability Testability rubiginosa. Unfortunately, N. ogloblini was never having been collected it is presumed they therefore higher uncertainty in their ranking, is accidentally excluded from the set of 127 taxa feed on roots. Similarly, little is known about the obvious. This is encouraged by the EPA. 30 48 20 56 63 10 score Value Value ranked by PRONTI, so it is not possible to examine endemic chrysomeline genus Aphilon, which that species further here. Cassida rubiginosa was was ranked highly by PRONTI but not included CONCLUSION 20 27 31 59 32 31 on the traditional list because it is a useful BCA for on the traditional list because of the diﬃculty Compared with the traditional method of test- score Status Californian thistles, Cirsium spp. (Compositae), of collecting and testing this species, and their species selection for biological control risk in areas of the Bay of Plenty and Wairarapa (M. small size. Host range tests with E. daenerys and assessment, PRONTI adds value by delivering a 36 49 44 47 43 38 Cripps, AgResearch pers. comm.). Also, this the two native Chrysomelinae on the traditional more consistently-applied, transparent evidence- score species met the ecological criteria for phenology list (i.e., Chalcolampra sp. and Allocharis sp.) may based ranking of each non-target species. Also, it Resilience overlap, with leaf-feeding larvae, a moderate indicate whether further tests with other native provides for the examination of how the rankings body size, and potential habitat overlap where Chrysomelinae are needed. and test species selection have been achieved. 24 24 27 29 22 33

score infested weeds might grow beneath or adjacent PRONTI ranked some other species highly Consequently, researchers are able to quantify

E flea beetle). Both Galerucinae on the traditional (Table 1) and the PRONTI list (Table 3) are We thank the EPA and DOC, in particular list are introduced weed BCA: Agasicles primarily because the traditional list uses filters Clark Ehlers, and Chris Green, respectively, for hygrophila is an important BCA of alligator weed (e.g. phenology or testability) that remove taxa helpful comments on the usefulness of these two in warmer climates such as Northland, Auckland completely from the list, whereas PRONTI keeps approaches in risk assessment. Carl Wardhaugh, . N and the Bay of Plenty (Stewart et al. 1999), and all species on the list, but ranks those with low Scion, further delved into the habits of New Lochmaea suturalis was introduced to control testability or mismatching phenology further Zealand beetles for us and we are most grateful. heather in central North Island. Both species down the list. There is no requirement to select Ivan Arnerich and Sharon Burrell, Scion, have phenology overlap with the target, with the top ‘x’ species from the PRONTI list. The provided invaluable IT assistance. This research leaf-feeding larvae, and there is potential habitat value of PRONTI is that the model considers a was funded by Scion and The New Zealand Bruchidius villosus I Bruchidius Trachytetra spp Chrysolina quadrigemina I quadrigemina Chrysolina Oemona hirta E hirta Oemona Atrichatus ochraceus N Atrichatus Species and Origin Species and melanocephala Aporocera overlap should infested alligator weed or heather wider ecological range of non-target risks than Institute for Plant & Food Research Ltd as part be found growing amongst eucalypt plantations. the traditional method currently does (Todd et of the Better Border Biosecurity (B3) (www.b3nz. 22 21 20 25 24 23 Rank from PRONTI Both species are lower on the PRONTI list al. 2015). The PRONTI list (Table 3) included org) research collaboration. Table 3 continued Table Beneficial Insects 229

PRONTI PRONTI the PRONTI list above, are all Chrysomelinae or of the country (Gourlay et al. 2005). The three suturalis and Agasicles hygrophila remained Galerucinae (Table 3), whereas the traditional native galerucine taxa could also be alternative the only representatives of the Galerucinae on list (Table 1), included two species from other test species, although little is known about the the final test-species list. However, the value of 55 36 49 36 31 36 score subfamilies: Neolema ogloblini and Cassida larval feeding habits of these species, with larvae investigating NTS with less available data, and Testability Testability rubiginosa. Unfortunately, N. ogloblini was never having been collected it is presumed they therefore higher uncertainty in their ranking, is accidentally excluded from the set of 127 taxa feed on roots. Similarly, little is known about the obvious. This is encouraged by the EPA. 30 48 20 56 63 10 score Value Value ranked by PRONTI, so it is not possible to examine endemic chrysomeline genus Aphilon, which that species further here. Cassida rubiginosa was was ranked highly by PRONTI but not included CONCLUSION 20 27 31 59 32 31 on the traditional list because it is a useful BCA for on the traditional list because of the diﬃculty Compared with the traditional method of test- score Status Californian thistles, Cirsium spp. (Compositae), of collecting and testing this species, and their species selection for biological control risk in areas of the Bay of Plenty and Wairarapa (M. small size. Host range tests with E. daenerys and assessment, PRONTI adds value by delivering a 36 49 44 47 43 38 Cripps, AgResearch pers. comm.). Also, this the two native Chrysomelinae on the traditional more consistently-applied, transparent evidence- score species met the ecological criteria for phenology list (i.e., Chalcolampra sp. and Allocharis sp.) may based ranking of each non-target species. Also, it Resilience overlap, with leaf-feeding larvae, a moderate indicate whether further tests with other native provides for the examination of how the rankings body size, and potential habitat overlap where Chrysomelinae are needed. and test species selection have been achieved. 24 24 27 29 22 33

score infested weeds might grow beneath or adjacent PRONTI ranked some other species highly Consequently, researchers are able to quantify

Exposure Exposure to eucalypt plantations. Cassida rubiginosa was that had not been included on the traditional list their thoughts around why a species should, or ranked lower by PRONTI, primarily because of for a number of reasons. Eucolaspis brunnea was should not, be considered for testing against a 61 78 68 44 65 63 its more distant relationship to the target. If host ranked highly on the PRONTI list because it is particular BCA, and examine those carrying score Hazard Hazard range testing shows that C. rubiginosa is attacked a chrysomelid, it has many links to the foodweb higher uncertainty. PRONTI will become easier by E. daenerys then improvements will need to be (they are generalist plant feeders), it is native and and faster to use as more species are added to made to the PRONTI tool. easy to collect and test. Although the adults are the Eco Invertebase, and is available for use with Of the four Galerucinae selected from known to feed on eucalypt leaves, the larvae are other entomophagous BCAs under consideration the PRONTI list, three are native taxa (i.e. root-feeders and so they were excluded from the for release in New Zealand. Chaetocnema spp., P. c y ane a , and Alema spp.) traditional list. Bruchinae Galerucinae Chrysomelinae Cerambycidae Eumolpinae Cryptocephalinae Subfamily or or Subfamily Family and one is a weed BCA (L. jacobaeae, the ragwort The diﬀerences between the traditional list ACKNOWLEDGEMENTS flea beetle). Both Galerucinae on the traditional (Table 1) and the PRONTI list (Table 3) are We thank the EPA and DOC, in particular list are introduced weed BCA: Agasicles primarily because the traditional list uses filters Clark Ehlers, and Chris Green, respectively, for hygrophila is an important BCA of alligator weed (e.g. phenology or testability) that remove taxa helpful comments on the usefulness of these two in warmer climates such as Northland, Auckland completely from the list, whereas PRONTI keeps approaches in risk assessment. Carl Wardhaugh, . N and the Bay of Plenty (Stewart et al. 1999), and all species on the list, but ranks those with low Scion, further delved into the habits of New Lochmaea suturalis was introduced to control testability or mismatching phenology further Zealand beetles for us and we are most grateful. heather in central North Island. Both species down the list. There is no requirement to select Ivan Arnerich and Sharon Burrell, Scion, have phenology overlap with the target, with the top ‘x’ species from the PRONTI list. The provided invaluable IT assistance. This research leaf-feeding larvae, and there is potential habitat value of PRONTI is that the model considers a was funded by Scion and The New Zealand Aporocera melanocephala E melanocephala Aporocera Bruchidius villosus I Bruchidius Trachytetra spp Chrysolina quadrigemina I quadrigemina Chrysolina Oemona hirta E hirta Oemona Atrichatus ochraceus N Atrichatus Species and Origin Species and overlap should infested alligator weed or heather wider ecological range of non-target risks than Institute for Plant & Food Research Ltd as part be found growing amongst eucalypt plantations. the traditional method currently does (Todd et of the Better Border Biosecurity (B3) (www.b3nz. 22 21 20 25 24 23 Rank from PRONTI Both species are lower on the PRONTI list al. 2015). The PRONTI list (Table 3) included org) research collaboration. Table 3 continued Table Beneficial Insects 230

REFERENCES biocontrol: a nationwide survey for impacts Todd JH, Barratt B, Withers TM, Berndt LA, Barratt BIP, Moeed A 2005. Environmental safety on non-target plants. New Zealand Plant Gresham BA, Avila GA, Malone LA 2016. of biological control: Policy and practice in Protection 57: 102–107. A comparison of methods for selecting New Zealand. Biological Control 35: 247–252. Peixoto L, Allen GR, Ridenbaugh RD, Quarrell non-target species for risk assessment of Barratt BIP, Todd JH, Malone LA 2016. Selecting SR, Withers TM, Sharanowski BJ 2018. When the biological control agent Cotesia urabae non-target species for arthropod biological taxonomy and biological control researchers BioControl 62: 39–52. control agent host range testing: evaluation of unite: species delimitation of Eadya Withers TM, Allen GR, Reid CAM 2015. a novel method. Biological Control 93: 84–92. parasitoids (Braconidae) and consequences Selecting potential non-target species for de Little DW 1979. A preliminary review of for classical biological control of invasive host range testing of Eadya paropsidis. New the genus Paropsis Olivier (Coleoptera: paropsine pests of Eucalyptus. PLoS ONE in Zealand Plant Protection 68: 179–186. Chrysomelidae) in Tasmania. Journal of the press. Australian Entomological Society 18: 91–107. Reid CAM 2006. A taxonomic revision of the Gourlay AH, Fowler SV, Rattray G 2005. Australian Chrysomelinae, with a key to Abundance of ragwort flea beetle (Longitarsus the genera (Coleoptera: Chrysomelidae). jacobaeae) at five sites on the west coast, Zootaxa 1292: 1–119. South Island, New Zealand. In LC0405/131 Reid CAM 2014. Chrysomeloidea Latreille, 1802. Manaaki Whenua, Landcare Research. In: Leschen RAB, Beutel RG ed. Handbook Hoddle MS 2004. Chapter 4. Analysis of fauna of Zoology, Vol IV (Arthropoda: Insecta), in the receiving area for the purpose of Part 38 Coleoptera, Beetles. Morphology and identifying native species that exotic natural Systematics. De Gruyter, Berlin. Pp. 11–16. enemies may potentially attack. In: Van Rice AD 2005. The parasitoid guild of larvae of Driesche RG, Reardon R ed. Assessing host Chrysophtharta agricola Chapuis (Coleoptera: ranges for parasitoids and predators used for Chrysomelidae) in Tasmania, with notes on classical biological control: a guide to best biology and a description of a new genus and practice. USDA Forest Service, Morgantown, species of tachinid fly. Australian Journal of West Virginia. Pp. 24–39. Entomology 44: 400–408. Kuhlmann U, Schaﬀner U, Mason PG 2006. Ridenbaugh RD, Barbeau E, Sharanowski BJ 2018. Selection of non-target species for host Descriptiom of four new species of Eadya specificity testing. In: Bigler F, Babendreier (Hymenoptera, Braconidae) parasitoids of the D, Kuhlmann U ed. Environmental impact Eucalyptus tortoise beetle (Paropsis charybdis) of invertebrates for biological control of and other Eucalyptus defoliating leaf beetles. arthropods. CABI International, Wallingford, Journal of Hymenoptera Research (In press). UK. Pp. 15–37. https://doi.org/10.3897/jhr.@@.24282. Manners AG, Palmer WA, Dhileepan K, Stewart CA, Chapman RB, Emberson RM, Hastwell GT, Walter GH 2010. Characterising Syrett P, Frampton CMA 1999. The eﬀect insect plant host relationships facilitates of temperature on the development and understanding multiple host use. Arthropod- survival of Agasicles hygrophila Selman Plant Interactions 4: 7–17. & Vogt (Coleoptera: Chrysomelidae), a Nahrung HF, Allen GR 2004. Sexual selection biological control agent for alligator weed under scramble competition: mate location (Alternanthera philoxeroides). New Zealand and mate choice in the eucalypt leaf beetle Journal of Zoology 26: 11–20. Chrysophtharta agricola (Chapuis) in the Todd JH, Barratt BIP, Tooman L, Beggs J, Malone field. Journal of Insect Behavior 17: 353–366. LA 2015. Selecting non-target species for Paynter QE, Fowler SV, Gourlay AH, Haines ML, risk assessment of entomophagous biological Harman HM, Hona SR, Peterson PG, Smith control agents: evaluation of the PRONTI LA, Wilson-Davey JRA, Winks CJ, Withers decision-support tool. Biological Control 80: TM 2004. Safety in New Zealand weed 77–88.