Eradicating the large white butterfly from New Zealand eliminates a threat to endemic Brassicaceae

In May 2010 the large white butterfly, Pieris brassicae L. (Lepidoptera: Pieridae), was discovered to have established in New Zealand. It is a Palearctic species that—due to its wide host plant range within the Brassicaceae—was regarded as a risk to New Zealand’s native brassicas. New Zealand has 83 native species of Brassicaceae including 81 that are endemic, and many are threatened by both habitat loss and herbivory by other organisms. Initially a program was implemented to slow its spread, then an eradication attempt commenced in November 2012. The P. brassicae population was distributed over an area of approximately 100 km2 primarily in urban residential gardens. The eradication attempt involved promoting public engagement and reports of sightings, including offering a bounty for a two week period, systematically searching gardens for P. brassicae and its host plants, removing host plants, ground-based spraying of insecticide to kill eggs and larvae, searching for pupae, capturing adults with nets, and augmenting natural enemy populations. The attempt was supported by research that helped to progressively refine the eradication strategy and evaluate its performance. The last New Zealand detection of P. brassicae occurred on 16 December 2014, the eradication program ceased on 4 June 2016 and P. brassicae was officially declared eradicated from New Zealand on 22 November 2016, 6.5 years after it was first detected and 4 years after the eradication attempt commenced. This is the first species of butterfly ever to have been eradicated worldwide.

The authors have declared that no competing interests exist. NO  General guidance is provided below.
Consult the submission guidelines for detailed instructions. Make sure that all information entered here is included in the Methods section of the manuscript.
All work described in this manuscript that involved human subjects was conducted with strict adherance to legislation described in the New Zealand Biosecurity Act 1993 (http://www.legislation.govt.nz/act/public/1993/0095/latest/DLM314623.html). The data were collected by staff from the New Zealand Department of Conservation and Ministry for Primary Industries who were authorised to do so under the New Zealand Biosecurity Act 1993. Pieris brassicae is legislated as an unwanted organism under this Act, which means authorised persons have a wide range of statutory powers to enable them to control it; including accessing, inspecting and applying treatments on privately owned properties.  to South Africa (Geertsema, 1996) and Chile (Gardiner, 1974), and may have 53 reached Nelson via its seaport as pupae on imported shipping containers, which is a 54 known pathway for P. brassicae (Anonymous, 2002;Molet, 2011). 55

Powered by Editorial Manager® and ProduXion Manager® from Aries Systems Corporation
All of P. brassicae's many host plants are brassicas (Brassicaceae) (Feltwell, 56 1982). Each female lays about 500 eggs, which are laid on host plants in batches of 57 50-150 eggs (Gardiner, 1963). Larvae feed gregariously and may defoliate several 58 plants during their development. Fifth instar larvae crawl away from their host plants 59 to pupate, typically on vertical surfaces in sheltered locations (Feltwell, 1982 (Spieth and Cordes, 2012), which suggested it could spread quickly in 78 New Zealand, and this impression was reinforced by P. rapae which took just 5-8 79 years to spread throughout New Zealand (Muggeridge, 1942). Surprisingly, however, 80 P. brassicae still appeared to be restricted to Nelson 2 years after it was first 81 recorded there (Philip, 2012). Nevertheless, MPI terminated its response in 82 The operational details of many previous eradication programs reside in 92 relatively inaccessible grey literature, which limits opportunities for learning 93 (Genovesi, 2005;Simberloff, 2009Simberloff, , 2002. This paper aims to help inform future 94 eradication programs by summarising the methods used and results obtained. 95 96

Methods 97
We define a 'detection' as the discovery of one or more P. brassicae at one location 98 at one time. Thus, detections refer to the number of inspections that revealed P. 99 brassicae rather than to the number of P. brassicae individuals found. was conducted by DOC and sought to both confirm the program was being well 120 managed and identify opportunities for improvement (Briden and Broome, 2013). 121 The second review was conducted by MPI and had similar goals to the first, plus it 122 also evaluated the program's likelihood of success (Gill, 2013a). Participants 123 included three TAG members, nine independent experts, and five MPI staff (Gill, 124 2013b). Prior to the review, participants were sent a report describing program issued with VHF and UHF radios, and team leaders carried mobile phones. Each 172 day teams were assigned to search particular properties specified via analysis of 173 previous surveillance results (see below).

Prioritising locations to search 175
The program aimed first to eliminate P. brassicae, then to continue surveillance to 176 confirm eradication. During the elimination phase, the program prioritised the 177 destruction of small peripheral P. brassicae populations to minimise spread beyond 178 the operational area, while simultaneously treating the larger central population to 179 reduce population growth and emigration pressure (Brown et  The phenology of P. brassicae was modeled (Kean and Phillips, 2013, in 196 preparation) using published data for its developmental responses to temperature 197 (e.g., Davies and Gilbert, 1985) and day length (e.g., Spieth and Sauer, 1991). The could not be searched (e.g., due to threatening dogs, locked gates or unhelpful 228 residents), contact was made again by phone or letter and access arranged. 229 Eggs and larvae were sought by systematically inspecting all host plants. Any 230 found were removed, then host plants were treated. Immature P. brassicae were 231 either killed upon detection, or kept in captivity to monitor parasitism then killed. 232 Pupae were searched for throughout the year, but were explicitly targeted during 233 winter on properties where mid-late stage larvae had been detected the previous 234 autumn. Inanimate objects such as fences, garden sheds and house exteriors were 235 searched using ladders and torches as necessary to inspect cracks and crevices. 236 Adjacent properties were also searched if it was suspected that larvae had crawled 237 off the property to pupate. 238 Adults were searched for in sunny locations with abundant nectar sources and 239 captured with hand-held nets. This was often difficult and time consuming due to P. 240 brassicae's rapid and evasive flight, but was considered worthwhile because: 241 Capturing gravid females minimised the number of eggs they could otherwise have 242 laid, potentially over many hectares; and capturing males when adult populations 243 were low potentially inhibited mate finding and reduced female fecundity. 244 Research was conducted to develop attractants for P. brassicae adults, but did 245 not produce practicably useful results (Sullivan et  A program review recommended that all P. brassicae host plants at a site should be 290 sprayed with insecticide whenever eggs or larvae were found because search 291 efficacy was likely < 100% (Briden and Broome, 2013). Consequently, the BioGro-292 certified organic insecticide Entrust ® SC Naturalyte ® was chosen as the most socially 293 acceptable option. The horticultural oil D-C-Tron ® was added to improve spray 294 coverage and increase egg mortality. Spraying was usually conducted after gaining 295 consent from property occupants, but occasionally occurred without consent when 296 the occupants could not be contacted and late-stage larvae were found. If occupants 297 were opposed to this treatment then one of the following alternatives were used: 298 which parasitises late-stage larvae and pupae (Muggeridge, 1943). Both species 314 also parasitise P. brassicae (Muggeridge, 1943)  property, an early practice was to also inspect adjacent properties within these radii 395 (Phillips et al., 2014a). However, a spatial analysis of surveillance data found no 396 evidence for clustered detections, thus it was concluded that searching properties 397 that surround an infested property was unlikely to increase detection rates above 398 searching randomly chosen properties in the same block (Phillips and van Koten, 399 2014) and the practice was discontinued. Further evidence that individual P. The UV lures were first deployed in October 2014 when detection rates had 408 already declined to low levels (Fig. 2). Pieris brassicae adults approached the lures 409 in a manner similar to P. rapae (Obara et al., 2008a, b), but never alighted on them. 410 When the eradication program began in summer 2012, there were several 440 detections outside the operational area. In summer 2012-13 (Fig. 3), one 441 (parasitised) P. brassicae larva was found ca. 25 km west of Port Nelson near Upper 442 Moutere (Fig. 1). This required intensive work to gain confidence additional P. 443 brassicae had not escaped from the operational area, including increased publicity 444 between Upper Moutere, Motueka and Nelson (Fig. 1). The larva was likely taken to 445 Upper Moutere from Nelson on an infested cabbage. Between autumn 2013 and 446 autumn 2014 (Fig. 3), several P. brassicae were detected ca. 11 km north of Port 447 Nelson at Glenduan (Fig. 1), which also required significant treatment. In summer 448 2013-14 (Fig. 3), one adult was detected ca.15 km southwest of Port Nelson at Hope 449 and another was detected ca. 10 km northeast of Port Nelson at Lud Valley (Fig. 1). 450 Intensive searching in the vicinities of these detections revealed no further P. 451

brassicae. 452
Despite such dispersal events, from autumn 2014 P. brassicae became 453 increasingly confined to central Nelson (Fig. 3), and it became apparent during 2016 454 that the last detection had occurred near central Nelson in summer 2014-15 (Fig. 3). 455 Thereafter, active surveillance persisted until winter 2016 when confidence that P. 456 brassicae had been eliminated was sufficient to terminate the program (Fig. 3). 457 458

Temporal changes in detection rates 459
Eggs, larvae and adults of P. brassicae were more detectable than pupae, thus there 460 were peaks in detection rates during spring and autumn when they were more 461 prevalent than pupae (Fig. 2). Monthly rates peaked in September 2013 when P. 462 brassicae (including all life stages) was detected on 9% of 2931 inspected 463 properties. By this time, staff had been fully trained, P. brassicae was relatively 464 abundant, and most of the population was exposed to control (i.e., few pupae). 465 Thereafter, rates generally declined, though they showed regular smaller peaks each 466 autumn and spring until the end of 2014. They declined to zero in January 2015 and 467 remained there until 4 June 2016 when surveillance ended (Fig. 2). 468 Early data entry issues included a GIS interface that: Allowed users to inadvertently 498 enter incorrect/invalid inspection dates and misspelled addresses; and provided 499 users with inadequate confirmation that new records had been successfully entered 500 and saved, which often provoked duplicate entries. These issues were compounded 501 by the Nelson cadastre initially being incomplete and out of date, which sometimes 502 created confusion for field staff about the spatial locations of addresses and resulted in inspection records being assigned to incorrect addresses. These problems 504 created a dataset that was time-consuming to correct before it could be reliably used 505 for analysis. In November 2014, a data manager with GIS expertise was assigned 506 full time to the eradication program, and remaining issues with the cadastre and GIS 507 interface were resolved by early 2015. 508 Numerous aspects of the eradication program additional to P. brassicae's 538 conspicuousness and accessibility likely contributed to its success at relatively low 539 cost. The program engendered strong public support and received valuable reports 540 of sightings that accounted for ca. 20% of all P. brassicae detections. This support 541 was fostered by comprehensive publicity, rapid responses to public reports, 542 respectful and communicative staff, and the availability of an effective organic 543 insecticide which was more acceptable to many residents than synthetic chemical 544 alternatives. The bounty particularly excited public interest, plus it eliminated some 545 P. brassicae and provided independent evidence that the population had been 546 correctly delimited. It was also helpful that in 2001 MPI had declared P. brassicae an 547 Unwanted Organism under the New Zealand Biosecurity Act 1993 because it gave 548 authorised staff the legal right to search and treat private properties for P. brassicae. 549 Moreover, some DOC staff had this authorisation before the program began, and 550 after it commenced they expedited training to authorise additional staff. 551 Sometimes when nonnative organisms are discovered in new regions, little 552 technical information is available to support effective responses (Pluess et al., 2012). 553 However, numerous studies of P. brassicae in its native range were available to 554 support aspects of the eradication attempt including species diagnosis, identifying 555 effective chemical treatments, defining the butterfly's host range and natural 556 enemies, and developing a phenology model and lure. Unfortunately, such 557 information had not been used to develop preparedness plans prior to the 558 establishment of P. brassicae in New Zealand, which might have further increased 559 the probability of eradication success (Pluess et al., 2012). 560 Several aspects of P. brassicae's New Zealand habitat and ecology were 561 fortuitously helpful to the program. Numerous P. brassicae natural enemies were 562 present in Nelson and probably facilitated population suppression. These included: 563 the insect parasitoids C. glomeratus and Pt. puparum (Muggeridge, 1943); and 564 insect predators such as Vespula vulgaris, V. germanica (Brodmann et al., 2008), 565 Polistes chinenis antennalis (Clapperton, 1999), various species of ants (Jones, 566 1987), spiders, harvestmen and predatory beetles (Dempster, 1967) and birds 567 (Baker, 1970). The butterfly's potential population growth rate in Nelson was also 568 limited by a proportion of the population entering aestivation, which reduced that part The data management issues experienced predominantly during the first 2 years 597 of the program reduced operational and analytical efficiency, but did not create 598 serious doubt about achieving the feasibility criterion, "Programme is effectively 599 managed, and its status is reliably monitored and accurately recorded" (Phillips et