Introduction

Despite increased attention and enhanced regulatory efforts, non-native organisms continue to become established in new regions, primarily as a result of the increasing volume and shifting patterns of international travel and trade (Aukema et al. 2010; Hulme et al. 2008; Levine and D’Antonio 2003; Reichard and White 2001). Many countries maintain border biosecurity systems designed to detect and intercept species arriving through trade and travel routes before these organisms can become established. However, the sheer volume of global trade and travel, and the diversity of invaders, makes it inevitable that some species will evade detection at ports-of-entry, and a portion of those species will become established (Brockerhoff et al. 2006; Liebhold et al. 2012; National Research Council 2002; Work et al. 2005). When a non-native species is detected, consequent management options range from taking no action to implementing an eradication effort. The concept of eradication is beguiling; it suggests a final and permanent solution to the threat of an invader. In practice, however, the deliberate extirpation of a target species can be challenging, both biologically and economically (Myers et al. 2000; Popham and Hall 1958). Moreover, because propagule pressure tends to be positively associated with establishment success (Drake and Lodge 2006; Lockwood et al. 2005; Simberloff 2009), successful eradication of a target species could be ephemeral if propagule pressure is not mitigated.

Past work has highlighted a number of conditions thought to be critical for an eradication program to be successful (e.g., Brockerhoff et al. 2010; Dahlsten and Garcia 1989; Hoffmann et al. 2011; Knipling 1966; Myers et al. 2000; Pluess et al. 2012a; Pluess et al. 2012b; Simberloff 2003; Simberloff et al. 2005). Attributes include the ability to detect, identify and monitor an invader, understanding the potential risks and impacts posed by the invader, and having the tools to respond rapidly and effectively. Policy related attributes include the authority to intervene or take action on public and privately-owned lands, procurement of necessary funds, and the commitment and political will exhibited by affected stakeholders and the public in support of the eradication effort. As a result of these collectively complex and daunting requirements, and the notoriety of a few spectacularly failed attempts, the scientific and regulatory community has often viewed eradication with pessimism (e.g., Dahlsten 1986; Dahlsten and Garcia 1989; Myers et al. 1998; Whitten and Mahon 2005).

Despite an often negative perception of the viability of eradication efforts, the long-term benefits of successful eradication typically outweigh the costs of the program (Brockerhoff et al. 2010). A dispassionate analysis of the factors that have influenced the success of eradication programs could have tremendous practical utility and serve to inform and improve the decision-making process. In this paper, we sought to assess the importance and quantify the roles of various factors that affect eradication success of non-native arthropods by using a global database comprised of successful and failed eradication programs.

Previous reviews of attempts to eradicate non-native pest species have been largely descriptive and consist of either narratives of selected programs or generalized examples (e.g., Dahlsten and Garcia 1989; Graham and Hourrigan 1977; Myers et al. 1998; Myers et al. 2000; Popham and Hall 1958). In a recent quantitative study, Pluess et al. (2012a) used Generalized Linear Mixed Models to analyze a database of 136 eradication campaigns against a variety of terrestrial invertebrates, plants, and plant pathogens. They reported that local campaigns were more likely to succeed than regional or national programs. Other factors, including the level of biological information available about the target species, insularity, and reaction time did not significantly influence the rate of eradication success. In a subsequent classification tree analysis based upon 173 eradication campaigns, reaction time was identified as an important determinant of eradication success, along with habitat type and target taxon (Pluess et al. 2012b). The authors also highlighted the dearth of information on program costs and other socioeconomic factors in their data, which are generally considered to play major roles in the outcome of eradication campaigns.

To expand our understanding of the determinants of successful eradication programs, we acquired data from 672 eradication programs against arthropods that were undertaken around the world between 1890 and 2010. These data were compiled into a web-based Global Eradication and Response DAtabase (“GERDA,” Kean et al. 2013) that we developed and hereby present. Although GERDA also currently contains data on eradication programs targeting other taxa (e.g., nematodes, fungi, molluscs), we focused our analysis specifically on arthropods to avoid comparing taxa that differ vastly in their respective biology and invasion ecology. In addition to the current analysis presented in this paper, a long term goal of GERDA is to provide a repository to facilitate the ongoing collection and transfer of knowledge among biosecurity practitioners and the scientific community.

Materials and methods

Compilation of data

We compiled a database of eradication attempts targeting a total of 130 non-native arthropod species implemented in 91 countries (Fig. 1). The species targeted for eradication were generally defined as actionable pests in that their expected economic and ecological harm was sufficiently great to warrant an eradication attempt. We acquired data from a variety of sources including scientific literature, government documents and press releases, unpublished reports and other components of the grey literature, searches of reputable internet sites (i.e., those associated with universities or government agencies), and interviews with biosecurity personnel. The arthropod eradication data used for our analysis is documented by 453 references, all of which are recorded in GERDA.

Fig. 1
figure 1

Country representation of arthropod eradication programs currently within GERDA

Full size image

One major challenge we encountered when compiling these data was defining whether or not a management activity constituted an “eradication attempt.” To be included in GERDA, an eradication program had to satisfy the following three conditions:

  1. 1.

    Management intent Eradication, defined as the complete removal of the target species from a defined area, had to be identified as a goal of the program. Pest management programs were excluded if the objectives included reduction or containment, but not local extinction, of the target population. One of the greatest technical challenges of any eradication campaign is demonstrating the absence of the very last target individual; mere pest reduction, rather than complete extirpation, thus omits one of the most important aspects of eradication. However, we did encounter a few cases where eradication was fortuitously achieved following large-scale pest reduction programs. We included data from these cases because they provided valuable information regarding the effort and expenditures required to achieve eradication. Also, some newly-discovered incursions were managed initially for containment while information was gathered to evaluate potential costs, benefits, and feasibility of eradication. Such cases were not included unless there was evidence that the management goal changed from containment to eradication.

  2. 2.

    Spatial distinctness Eradication programs were considered unique when populations of a target pest occupied discrete areas separated by more than twice the typical dispersal distance of the target species, although in most cases the distances were far greater. In such cases, we argue that while the eradication effort was directed at the same target pest, programs targeting individual populations would have proceeded irrespective of other populations and therefore, each was uniquely informative. Some eradication programs against Lymantria dispar (L.) in North America (Hajek and Tobin 2009), for example, met this criterion. Conversely, an eradication program against Teia anartoides (Walker) in New Zealand (1999–2006) that targeted three population epicenters was considered as one program because each epicenter was close enough to be within the dispersal capability of male moths (Suckling et al. 2007). It was not always possible to deduce from the available data whether one large eradication program constituted several independent local programs; in these cases, we considered the program as a single effort. It was also sometimes difficult to determine if eradication programs against the same taxon in adjoining geopolitical areas (e.g., across US or Australian states) represented distinct and unique programs. In these cases, programs were included separately when separate management agencies were responsible for the program, or considered as one program when the management action was conducted under the auspices of a common agency.

  3. 3.

    Temporal distinctness The International Standards for Phytosanitary Measures Number 9 (Food and Agricultural Organization of the United Nations 2006) specifies that “the minimum period of time of pest freedom to verify eradication will vary according to the biology of the pest, but should take into consideration factors such as sensitivity of detection technology, ease of detection, life cycle of the pest, climatic effects, and efficacy of treatment.” Under current international standard practices, declaration of eradication of most pests can be made provided that suitable surveillance activity has resulted in no subsequent detections for at least 2–3 times the normal generational time of the target taxon. We followed this convention whenever there was doubt about whether a subsequent detection in a previously treated area constituted a new invasion or an unsuccessful eradication.

GERDA data fields

A summary of data fields included in GERDA is presented in Table 1. Most fields were not challenging to populate since they were based on various taxonomic details and basic life history of the target taxon, geographic location of the eradication program and its climate, and details of the agency responsible for the program. Data that were more difficult to obtain, at times resulting in missing data fields, included the method of initial detection and specific treatments or tactics used for eradication. The most challenging data to acquire were the program costs. Even when expenditures were reported, they often only included direct costs of the treatment and not the total program cost (i.e., personnel, pre-treatment environmental assessments, public outreach and meetings). Of the 672 arthropod eradication programs used in this analysis, adequate cost data were available for 141 programs.

Table 1 A summary and description of the primary data fields currently included in the Global Eradication and Response DAtabase (GERDA)
Full size table

Standardization of costs

All cost data were converted to the 2005 United States dollar (USD) to allow for a direct comparison of eradication programs worldwide. To do this, historical annual average exchange rates were first used to convert local currencies to the USD (Officer 2011); USD amounts were then standardized to the year 2005. Various inflation rates can be applied for this standardization, some of which will inflate early years more than others (Williamson 2011). We used the GDP deflator, which “represents the mean price of all the goods and services produced in the economy” (Williamson 2011). Since eradication programs involve a combination of fixed and variable costs for labor and materials, the GDP deflator likely provides a more accurate index of cost than other methods, which generally rely on either the price of household consumables or the cost of unskilled labor.

Data analysis

Many of the quantitative and categorical variables from Table 1 were used in our analyses, and we merged some data fields or categories within a given data field. This reduced the frequency of missing data and minimized redundancy when categories within the field were conceptually equivalent. For example, when considering the size of the eradication program, we used the maximum extent of the regulated area, the maximum area treated, or the larger of the two values when both fields were populated. Due to the large number of target species that typically have a univoltine lifecycle, we considered species to be either “univoltine” or “not univoltine.” The variables and their respective categories used in this analysis are summarized in Table 2. Some variables were transformed using log10 to satisfy assumptions of normality (Table 2).

Table 2 Variables from GERDA and their categories used in the logistic regression analysis
Full size table

We first used logistic regression, using both forward and backward selection methods independently, to develop a subset of variables to explore further. The binary response variable was eradication success or failure (Table 2). Eradication programs in which the outcome was “in progress” or “unknown” were omitted in the logistic regression analysis (167 programs). Once we compiled a subset of variables through the stepwise technique, we used the likelihood ratio, G2, to assess variable significance. Odds ratios, when appropriate, were formed after partitioning G2 into non-significant components (Agresti 1996).

Separately, we used multiple correspondence analysis to assess the use of control tools by the taxonomic order of the target species to determine if certain orders were more frequently targeted with a specific control tool or set of tools. In this analysis, we focused on only five of the 13 orders currently included in GERDA (Coleoptera, Diptera, Hemiptera, Hymenoptera, and Lepidoptera), which collectively represented 617 eradication programs. Of the omitted orders, four contained only 1 or 2 eradication programs, while for the remaining four orders we lacked sufficient control tool information. We also used logistic regression to determine if eradication success differed among these five orders. Finally, we used least squares regression to analyze the relationship between the infested area (km2) and the cost of the program (millions USD), after normalizing both variables using the log10 transformation. All analyses were conducted in SAS (1999).

Results

Summary information for all data currently in GERDA, including the data presented in this paper, are available online (Kean et al. 2013); thus, we briefly summarize some of the characteristics of the database for arthropods. Of the 672 arthropod eradication programs considered in this analysis, 395, 110, and 167 were considered to be successful, a failure, or either in progress or unknown, respectively. The 167 programs either in progress or unknown were comprised of 68 species, of which 35 species were represented by one program, and 61 species were represented by <5 programs. The remaining 505 programs that were considered to be successful or a failure were comprised of 111 species.

Considering all 672 arthropod eradication programs, the most numerically-prevalent orders of target pests were Diptera (259 cases), Coleoptera (133 cases), Lepidoptera (133 cases), Hymenoptera (61 cases), and Hemiptera (31 cases). The most numerically prevalent target species were L. dispar dispar (L.) (73 cases), Ceratitis capitata (Wiedemann) (56 cases), Bactrocera dorsalis Hendel (40 cases), Aedes aegypti (L.) (33 cases), and Agrilus planipennis Fairmaire (25 cases). There were 112 target species for which the number of eradication programs was <10 and 51 species were targeted by only a single eradication program. A time series of the number of eradication programs by the year of commencement is presented in Fig. 2.

Fig. 2
figure 2

Cumulative number of initiated eradication programs, and when programs were declared to be either successful or a failure (excluding programs in progress or if the outcome was not known), 1890–2010. The insert graph represents the total number of initiated programs on a non-transformed scale

Full size image

The stepwise regression procedure initially identified the following variables as potentially significant predictors of eradication success: infestation size, method of detection, relative detectability of the organism, voltinism, program duration, host range, and primary feeding guild. In a subsequent full model with this subset of variables, duration and voltinism were not significant (P > 0.3). All other variables were significant (P < 0.01), and the Hosmer and Lemeshow lack-of-fit test was not significant (P = 0.94), suggesting that the full model would not be improved by the inclusion of additional variables. Because two species, L. dispar dispar and C. capitata, accounted for 10.9 and 8.3 % of the arthropod eradication programs, respectively, we also conducted our analysis with these two species excluded. In addition to both being the two most numerically dominant species, they also provided contrasting case examples, such as in the commodities affected (forest vs. agricultural), voltinism (univoltine vs. multivoltine), and feeding guild (folivore vs. frugivore). Following the stepwise regression procedure, the same variables were observed to be significant in this reduced dataset with the exception of one variable: the method of detection was not significant in this reduced dataset (P = 0.39).

An interesting aspect of these analyses was that the climate suitability variable was not significant due to the fact that eradication programs were rarely conducted in the absence of climate suitability. For example, in 95.1 % of the programs (N = 245 programs) in which we could confidently assign a Köppen climate group (Kottek et al. 2006) to both the native habitat of the target species and the habitat where the eradication program was conducted, the climate group overlapped. Furthermore, in only two cases was the level of climate mismatch by more than one climate group (i.e., Köppen C, warm temperate, to Köppen D, snow). The lack of climate suitability could result in a failure to establish following arrival and preclude the need to initiate eradication. Exceptions to this pattern could include pest species that can exploit climate-controlled environments, such as greenhouses, homes, and other structures.

The area of the eradication program, or infestation size, was both a significant predictor of the probability of eradication success (P < 0.01, N = 255, Fig. 3a) and program costs (P < 0.01, N = 141, Fig. 3b). Regarding eradication success, the odds of a successful eradication program were 1.3 times less likely (95 % CI 1.1–1.5) for every log10 increase in area. In terms of cost, we observed a positive relationship between area and cost [log10(costs, millions USD) = −0.254 + 0.416(log10(area, km2)); R2 = 0.52]. We also illustrate separately eradication attempts against non-native forest insects for which our cost data were particularly robust (Fig. 3b). We did not observe a significant effect of outcome (i.e., success or failure, P = 0.87), or a significant interaction effect between area and outcome (P = 0.87) on costs. Moreover, we also did not detect a significant effect of target group (forest insects whose programs were dominated by those against L. dispar, versus other arthropods whose programs were dominated by those against C. capitata and B. dorsalis, P = 0.53), or a significant interaction effect between target group and area on costs (P = 0.86). Thus, the primary driver of program costs appears to be the area of the infestation and not the target group or the eventual program outcome.

Fig. 3
figure 3

Relationship between area of the infestation and the probability of eradication success (a), and the program costs (b). In a, the size of the circle reflects the number of cases, while the solid and dashed lines are the predicted probabilities and 95 % confidence intervals from logistic regression, respectively. In b, the solid line is the least squares regression fit to all arthropod data

Full size image

The probability of eradication success was significantly affected by both the relative detectability of the target species (P < 0.01, Fig. 4a) and, when considering all species, the method of detection (P < 0.01, Fig. 4b). Because more than 95 % of target species in GERDA were classified as having either a high or low detectability, we excluded species that were classified as having a medium detectability. Generally, species were classified as having high detectability if a sensitive monitoring tool exists, such as traps baited with species-specific pheromones. Eradication programs were 8.1 times (95 % CI 3.9–16.6) more likely to be successful if the target species was classified as having a high, rather than a low, detectability. With regard to the method of detection, programs that used active methods, such as targeted traps using species-specific semiochemical attractants, were 4.6 times (95 % CI 1.4–15.4) more likely to result in eradication success than programs relying on passive detection methods (e.g., non-specific traps or public vigilance), and 26.8 times (95 % CI 9.6–74.6) more likely to result in eradication success than programs relying on host and habitat searches (Fig. 4b). Interestingly, programs that relied on passive detection methods were 5.9 times (95 % CI 2.0–17.4) more likely to be successful than those that relied on host or habitat searches. When L. dispar dispar and C. capitata were excluded, we did not detect a significant difference in the method of detection.

Fig. 4
figure 4

Number of eradication programs that were successes or failures based upon the relative detectability of the target species (a) or the method of detection (b) (cf. Tables 1, 2). Different lowercase letters in a, b denote significant differences in eradication success (P < 0.05)

Full size image

Host range of the target species significantly affected the probability of eradication success (P < 0.01, Fig. 5a). Polyphagous species were 6.2 times (95 % CI 3.6–10.6) more likely to be eradicated than the combined group of oligophagous and monophagous species, both of which were not significantly different (P = 0.06). This observation could be reflected by the number of programs that have targeted the polyphagous species L. dispar, C. capitata, and B. dorsalis, which collectively accounted for 169 programs. These species also represent high impact plant pests that are often aggressively targeted for eradication when detected, which could explain the higher success rate of eradication programs against polyphagous species.

Fig. 5
figure 5

Number of eradication programs that were successes or failures based upon the host range (a) and primary feeding guild (b) of the target species. Different lowercase letters in a, b denote significant differences in eradication success (P < 0.05)

Full size image

The primary feeding guild also significantly affected the probability of eradication success (P < 0.01, Fig. 5b). Folivores and frugivores formed a non-significant group (P = 0.09) and had the highest rate of eradication success, followed by parasites, root feeders, sap suckers, and omnivores, which formed a separate non-significant group (P = 0.97). Wood and subcortical phloem feeders had the lowest rate of eradication success and were 14.6 (95 % CI 7.1–30.1) and 6.2 (95 % CI 3.1–12.1) times less likely to be successfully eradicated than the combined group of folivores and frugivores, and the combined group of parasites, root feeders, sap suckers, and omnivores, respectively (Fig. 5b).

A multiple correspondence analysis revealed that in arthropod eradication programs the use of specific control tools differed significantly among orders (P < 0.01). When partitioning χ2 ≈ 89 % of the variation was explained by two dimensions (Fig. 6). The most pronounced associations between insect orders and the control tools used for eradication were the use of mass trapping or lure and kill, and the sterile insect technique against Dipteran pests, and mating disruption and biopesticides against Lepidopteran pests (Fig. 6). Host removal and destruction were most often associated with programs against Coleoptera and Hemiptera (Fig. 6). The probability of successful eradication also differed significantly among orders (P < 0.01). For eradication programs targeting Lepidoptera (N = 115 programs) and Diptera (N = 189), for which there were specific available control tools, 86.1 and 86.8 %, respectively, were considered successful. In contrast, 71.4, 68.2, and 59.1 % of programs targeting Hymenoptera (N = 49 programs), Hemiptera (N = 22), and Coleoptera (N = 88), respectively, were considered to be successful. We recognize that reporting bias in favor of successful programs almost certainly exists, which could alter the proportion of programs deemed successful. However, given the extensive sample size and the fact that many control tools are linked to the biology of the targeted organisms, we contend that the varying levels of success observed for different orders is likely a robust pattern.

Fig. 6
figure 6

a Multiple correspondence analysis of control tools used by order of target pests, showing variation primarily across two dimensions. The percentages indicate the percent variation explained by each respective dimension. b Corresponding plot of the number of programs, by order, that have used the indicated control tools. Orders are shown in uppercase (DIPT, Diptera; LEPID, Lepidoptera; HYMEN, Hymenoptera; HEMIP, Hemiptera; and COLEOP, Coleoptera) and control tools are shown in lowercase (sit, sterile insect technique; trap, mass trapping or lure and kill; biopest, biopesticide; md, mating disruption; pest, pesticide; removal, host or habitat removal; biocon, release of natural enemies; and quar, quarantine and movement control)

Full size image

Discussion

The increase in eradication attempts beginning in the late 1980s (Fig. 2) could partially reflect an increased number of arrivals of non-native species (Levine and D’Antonio 2003; Liebhold et al. 2006; McCullough et al. 2006; Work et al. 2005). It could also reflect increased awareness of the inimical effects of some non-native species and technological advances leading to new or improved control tactics and strategies. Only a minority of introduced species cause substantial economic, social and ecological harm (Aukema et al. 2010; Mack et al. 2000), and eradication is rarely warranted unless the impacts of the non-native species are expected to be severe. We acknowledge sampling bias could exist in the temporal pattern of data compiled in GERDA given that more recent data are easier to locate and compile, and that successful programs could be reported more often due to the reluctance to publicize failures. We contend, however, that the extensive database, which encompasses 672 arthropod eradication programs, is robust enough for this initial analysis of the primary drivers of eradication success and failure.

For a small proportion of potentially high impact, non-native pests, eradication could be cost-effective and the preferred management option for governments, providing that certain conditions are met. Our analysis, which encompassed many examples of eradication efforts targeting arthropod pests with diverse life history strategies, has revealed some consistent patterns. For example, the probability of eradication success declines as the area that is infested increases (Fig. 3a), a pattern that is both intuitive and reflected by prior observations (Liebhold and Tobin 2006; Pluess et al. 2012b; Rejmánek and Pitcairn 2002). Conversely, the cost of the eradication effort increases over the area that is infested (Fig. 3b), which likewise is intuitive. Quantification of the general relationships between the area of an infestation and the predicted probability of eradication success and program costs should be useful in future planning efforts when eradication is considered as a management response.

High relative detectability of the target species is a primary component of successful eradication programs (Fig. 4a). With regard to L. dispar eradication efforts, for example, there was a dramatic increase in eradication programs in North America following the identification and synthesis of the L. dispar sex pheromone (Bierl et al. 1970), which is now routinely used in L. dispar monitoring programs (Tobin et al. 2012). It is extraordinarily challenging to manage any species when the ability to detect the target species is limited, especially in an eradication program. Furthermore, the failure to detect small incipient populations caused by the lack of a sensitive survey tool could result in the infestation being larger when finally detected, and therefore, less likely to be successfully eradicated in an economically feasible manner (Fig. 3, Epanchin-Niell et al. 2012).

It was not unsurprising that the relative detectability of target pests was an important factor in eradication success. Programs that can rely on active methods of detection, such as species-specific lures and trapping devices, are most likely to be successful (Fig. 4b), reinforcing the importance of detection systems to improve invasive species management (Government Accountability Office 2006; Jarrad et al. 2011; Simberloff et al. 2005). Probability of success was also greater when passive means of detection, such as through private citizens reporting pest presence, were used rather than pre-emptively searching sites considered to be at high risk of species arrival (e.g., nurseries, industrial sites, sawmills), which was least likely to be associated with eradication success (Fig. 4b). This finding could reflect an overall increase in public awareness of non-native pests and the effectiveness of public relationship campaigns coordinated by universities, biosecurity and resource management agencies. The lack of significance of the method of detection when L. dispar dispar and C. capitata are excluded from the analyses could reflect the historical target trapping for these two pests. The availability of relatively inexpensive detection traps, and control tools, for both these high risk pests enables countries to maintain trapping networks and respond to detection with an eradication program. This likely explains the numerical dominance of L. dispar dispar and C. capitata. When these two species were excluded from the analyses, detection method was no longer significant, which could reflect insufficient variation in the method of detection represented by the remaining programs.

The detectability of target pests undoubtedly plays an important role at the end of an eradication program because it largely determines whether the apparent absence of a target pest represents eradication success. The reappearance of a species thought to be extinct, referred to as the “Lazarus effect” (Flessa and Jablonski 1983; Morrison et al. 2007), and has been implicated, for example, in the recurrence of C. capitata detections in California (Carey 1996; but see Liebhold et al. 2006). The inability to detect the last few remaining individuals of a population targeted for eradication, even in the face of aggressive detection efforts has important consequences in the analysis and interpretation of data from eradication programs. We attempted to minimize misclassifications of eradication success, such as those due to the Lazarus effect, by using a criterion of apparent absence of at least 2–3 times the normal generational time of the target taxon. Any remaining misclassifications would largely lie with the most numerically dominant taxa within GERDA; however, our results were largely unaffected when the two most dominant species, L. dispar dispar and C. capitata, were excluded, suggesting that potential Lazarus effects have not greatly affected our conclusions.

The importance of detectability could also be reflected in the primary feeding guild of the target species (Fig. 5). Damage caused by externally feeding arthropod folivores or frugivores is likely to be noticed, whereas more clandestine species that feed within their host plant are likely to escape detection for some time, at least during the early stages of the invasion process. Subcortical wood and phloem feeders have been a particularly challenging group to eradicate, reflecting their cryptic life history, often low detectability, and the lack of control options available for use over large areas. Research to develop better detection technology and control options for these organisms should be a priority, particularly given the sharp increase in new detections of non-native subcortical borers over the past 20–30 years (Langor et al. 2009; Aukema et al. 2010). The use of “citizen scientists” has also generated much interest with regard to a diversity of ecological research topics (see the recent special issue introduced by Henderson 2012). For example, in New Zealand, approximately half of all new plant pest detections are reported by the general public (Froud et al. 2008), and it is believed that every known Anoplophora glabripennis (Motschulsky) infestation in the USA was discovered by private citizens. The engagement of citizen scientists to aid in surveys for invasive pests is a promising management tool that deserves more attention (e.g., Beetle Busters 2012; Crall et al. 2010; Ingwell and Preisser 2011).

The ability to use taxon-specific tools in an eradication program also appears to be an important determinant of eradication success. Analysis of the current GERDA database reveals that Diptera and Lepidoptera had the highest rate of eradication success and both were strongly associated with more specific control tools, such as mass trapping, lure and kill, and the sterile insect technique (Diptera), or mating disruption and biopesticides formulated with the use of taxon-specific entomopathogens (Lepidoptera). In contrast, more general methods, such as host and habitat removal, tended to be associated with orders that have recorded less eradication success. Host-specific tactics could also benefit from wider societal acceptance if they are associated with fewer undesirable effects on the environment and to non-target organisms. The use of multiple tactics, especially those that are species-specific, that can act synergistically to decrease the target population could also lead to greater eradication success (Blackwood et al. 2012; Suckling et al. 2012).

We highlighted factors that are critical determinants of successful arthropod eradication programs, which should assist in the development of improved management responses to non-native species. Also, because we envision GERDA (Kean et al. 2013) as an online repository for eradication program data, the addition of new data could facilitate future analyses that provide greater insight into factors affecting the outcome of eradication programs. Indeed, since this manuscript was initially submitted, an additional 57 eradication programs and 25 non-native arthropod species that were not previously included have now been entered into GERDA. The inclusion of cost data for eradication programs, failed programs, and control tools used in the effort would greatly facilitate future analyses. Identifying constraints and determinants of success can provide a basis for prioritizing and enhancing future eradication attempts. Moreover, we anticipate that GERDA will help to inform biosecurity practitioners and the larger scientific community by providing rapid access to the experiences of others in the decision making process.