To identify introduced species of insects associated with woody plants in Europe, we used the list provided in Roques et al. (2016) complemented by a search of EPPO reporting services (1974–2021) and a search on Google scholar and Web of Science with many combinations of the keywords “alien (or non-native, or exotic) arthropod (or insect) species Europe”, “alien insect arthropod (or insect) + each European country”, “first report arthropod (or insect) Europe”, “first report arthropod (or insect) + each European country”. For pathogens, a list of invasive forest fungi and oomycetes detected in Europe from 1800 to 2008 was retrieved from Santini et al. (2013). Data on introduced species from 2008 until 2021 was retrieved from EPPO reporting services. We then looked for eradication programmes against each species. To identify eradication attempts against all pathogen groups (bacteria, nematodes, fungi + oomycetes and viruses or viroids), the EPPO A1 and A2 and alert lists were initially consulted for pathogens of woody plants, and all eradication programmes were searched in EPPO reporting services (1974–2020). To identify eradication attempts for both arthropods and pathogens, the EPPO Global Database (https://gd.eppo.int) was the main online database we consulted. This information was complemented with information from GERDA – Global Eradication and Response Database (http://b3.net.nz/gerda/index.php) and a search in the scientific and grey literature, including works published in scientific journals, conference proceedings, presentations, and books. Published information was searched through Google Scholar and Web of Science, using the “species name”, “alien arthropod”, alien insect”, “plant (or tree) pathogen”, “Europe (or individual countries)” and the words “eradication” and “containment” as keywords, in different combinations. Additionally, eradication reports or technical reports, pest alerts, and press releases from National and Regional Plant Protection Organizations (NPPOs and RPPOs) were consulted. The time range for eradication attempts ranged from 1945 to 2021. Finally, additional information was kindly provided by some countries’ NPPOs and RPPOs. For detailed information on some of the eradication cases which were not available in the English language, advanced searches were conducted on Google by introducing the species name and limiting results to the country for which information was missing. The information and reports obtained from these searches were then translated to extract the information required for our analyses.

A comprehensive database was constructed including the following information for each case (when available): i) species under eradication, ii) detection date, country, and location; iii) detection method, passive surveillance (i.e. casual observations reported by researchers, technician or citizens) or official survey conducted with that purpose; iv) establishment status (established or post-border interception); v) affected hosts; vi) host type (broadleaves, conifers, palms), vi) control methods used (chemical, host removal, biological, traps); vii) size of the infested area (as exact area information was not always available we defined it in categories ≤ 1 ha, > 1 ≤ 10, > 10 ≤ 100, > 100 ≤ 1000 or > 1000 ha); viii) environments infested (urban/peri-urban, protected green-houses, countryside); ix) climate, categorized as Temperate, Mediterranean or Continental according to Köppen classification system (Peel et al. 2007); x) programme start year, last detection, and date of eradication declared; xi) public education, and xii) the outcome, i.e. legal status (eradicated, under eradication, failure to eradicate). Categories used in each parameter are also described in Table 1 and Suppl. material 1.

For some parameters, information was not always available and so we defined additional criteria. For the establishment status, the pest or pathogen was considered established unless stated that it was found only on the imported plant material and not in other plants at that time or posteriorly to the destruction of the original plant material. For the outcome, we consider a pest or pathogen to be eradicated when there was an official confirmation, or if no further future records were reported. If the official status changed to restricted distribution or containment and it continued to spread, it was considered a failure. Otherwise, it was still considered under eradication.

For pathogens, in many cases, the exact location of detection was not known and thus, we used the NUTS3. If the pathogen was no longer detected during the next two-yearly surveys (or two consecutive surveys when surveys were separated by more than one year) in that region, it was considered eradicated.

Some of this information was used only for descriptive analysis whereas other parameters were used in the modelling analysis (Table 1). To obtain a sufficient number of replications per level in a factor, some levels were merged. Cramér’s V correlation between variables was estimated with the software package R, for the cases for which information for all the variables were available. (Suppl. material 2).

The statistical modelling aimed to predict the probability that the species became established (i.e., no longer found only on primary material) and next, once established, the probability of successful eradication as a function of different explanatory variables. Three main categories of factors were distinguished: i) control options, ii) characteristics of the environment/location of the outbreak, and iii) biological traits of the species. All analyses were performed for arthropods and pathogens separately.

Eradications and post-border interceptions

A total of 848 cases officially declared as eradication attempts were documented in our database, 314 against arthropods and 534 against pathogens. These cases concerned 49 species of arthropods (47 insect and 2 mite species) and 34 species of pathogens (21 fungi and oomycetes, 8 bacteria, 2 nematodes, and 3 virus/viroids). A large number of reports corresponded to post-border interceptions. These cases represented 49% (154) of reports on arthropods and 19% (87) on pathogens.

In the case of insects, these data show that for only 9% of the compiled list of 487 non-native insect species of woody plants detected in Europe, eradication measures were taken (42/487, Fig. 1). Species that are native to parts of Europe but non-native in other European countries (e.g. UK) were not included in this analysis.

Figure 1.

Cumulative number of non-native species of insects of woody plants for which eradication of established or intercepted populations was attempted, and the cumulated number of alien insect species reported for Europe until 2019.

The total number of insect species for which “eradication” measures were taken (both established populations and post-border interceptions), increased in the last two decades (Fig. 1). Still, the numbers are very modest when compared with the total number of non-native species introduced in Europe, reaching a maximum of 18% of the total number of introduced species in the last twenty years (25/136). For invasive fungi and oomycetes in European forests, eradication was attempted in only 12% of the cases (17/146).

Post- border interceptions were observed for only a few species. For arthropods, 43% of reported interceptions are linked to the oak processionary moth (Thaumetopoea processionea, OPM) in the UK, outside of the containment area in London and South East England where the pest has established, after being accidentally introduced from mainland Europe (EFRA 2022D). OPM has been intercepted in all types of environments, from nurseries to urban and peri-urban areas and in recently established plantations. Anoplophora chinensis ranks second in the number of interceptions (21 cases). This beetle was mostly intercepted in nurseries and urban and peri-urban environments, associated with imported Acer spp. plants for planting and bonsais (Branco et al. 2021). The remaining interceptions are distributed among 20 species (Suppl. material 1), mostly found in nurseries, garden centres or other confined environments (68% of cases).

Concomitantly, there is a discrepancy between the number of alien insect species reported and the number of eradication attempts by taxa. Most of the non-native species (52%) are hemipteran sap suckers, but eradication was attempted for only 6% of these (Fig. 2). Coleopterans represent 42% of the reported eradication attempts against insects and 96% of these attempts were against wood borers (Cerambycidae, Curculionidae, Buprestidae).

Figure 2.

Total number of non-native insect species on woody plants reported for Europe by order (bars) and proportion of species attempted to eradicate (established species).

For pathogens, 11 species have been intercepted outside of import-associated inspections, mostly in confined environments (72%). Phytophthora ramorum ranks first in the number of interceptions (63%), distributed among 12 European countries, followed by Cryphonectria parasitica (13%), the causal agent of chestnut blight, for which most interceptions were reported in the UK, where it has only recently established (Hunter et al. 2013; Romon‐Ochoa et al. 2022).

Success of eradication programmes in Europe

Eradication measures taken against organisms still restricted to the primary material with which they were introduced, here defined as post-border interceptions, were 100% successful. From here on we will consider only eradication programmes targeted at established populations in Europe. In total, 160 programmes were launched against 41 species of arthropods (Fig. 3) and 447 programmes against 31 species of pathogens (Fig. 4). The proportion of successes and failures varied greatly among species.

Figure 3.

Arthropod species for which eradication was attempted in Europe.

Figure 4.

Pathogen species for which eradication was attempted in Europe.

Temporal and spatial trends

For both arthropods and pathogens, the total number of eradication programmes against established populations increased abruptly in the last two decades, (Fig. 5). The success rate of eradication attempts against arthropods reached 72% in the period 2011–2020. In contrast, all programmes that started before 2000 failed. For pathogens, the success rate has also increased, being highest for the last two decades (approximately 65%). In the decades prior to 2000, the success rate was moderate, ranging between 40% and 52%.

Figure 5.

Eradication attempts in Europe by decade and corresponding rate of success of programmes targeting a arthropods and b pathogens.

In terms of geographic distribution, the highest number of eradication attempts per country were reported for France (81), Spain (61), Italy (57) and Germany (56) (Fig. 6). A clear concentration of eradication programmes exists in Western Europe (Fig. 6), with many of these programmes still ongoing.

Figure 6.

Eradication attempts against arthropods and pathogens in Europe, by country. The size of each pie is proportional to the number of eradication attempts in that country.

Area affected

For both arthropods and pathogens, the success rate was the highest for infestations restricted to small areas (Fig. 7). For infested areas < 1 ha, the success rate was 82% and 90%, for arthropods and pathogens, respectively (Fig. 7). Success decreased with an increase in the affected area, with the sharpest decrease observed for pathogens.

Figure 7.

Eradication attempts conducted in Europe by infested area and corresponding rate of success against a arthropods and b pathogens. Information on the approximate area affected at the time of programme start was retrieved for 139/160 cases for arthropods and 408/447 cases for pathogens.

For areas above 1000 ha, success for pathogens was only achieved once out of 40 concluded programmes (2.5%). This unique success concerned E. amylovora in Norway. The programme started in 1986 in an infested area of 30,000 ha where all the hosts were removed (i.e., all Cotoneaster, Sorbus and Pyracantha). Within the quarantine area (70,000 ha), the production and sale of all common fire blight hosts was prohibited and bee hives were moved to areas that were free from hosts of E. amylovora. From 1993 to 2000, no new detections were made and the outbreak was declared eradicated in 1998. Although fire blight was again detected within the restriction zone in 2000, it is unknown whether a re-emergence or a new introduction occurred (Sosnowski et al. 2009).

Detection site and affected hosts

New establishments of arthropods were most often detected in urban or peri-urban areas, including residential and industrial areas (65% of cases). The rate of eradication success was highest (74%) in confined environments, where the plant materials were delimited (nurseries, glasshouses and garden centres), intermediate in residential areas (52%), and lowest (26%) in the countryside (orchards, woodlands /forests). Pathogen detection, on the other hand, occurred in the countryside in 50% of cases (mostly orchards, Fig. 8), followed by confined environments (30% of cases), and only 20% of cases were first reported in urban and peri-urban areas. The rate of pathogen eradication success was also high in confined environments (86%) and similar for the remaining environments (50% for countryside and 51% for urban/peri-urban).

Figure 8.

Eradication attempts conducted in Europe against a arthropods and b pathogens, by detection site and with the corresponding rate of success. Information on the main type of environment affected at the time of the programme start was retrieved for 157/160 cases for arthropods and 405/447 cases for pathogens. * in woodlands or forests.

Most eradication programmes targeted PPWPs attacking broadleaves (79%, both for arthropods and pathogens). For arthropods attacking broadleaves in urban and peri-urban areas, the eradication success rate was 63%, while in the five cases reported in confined environments, the success rate was 100%. The lowest success rates were reported for pests on countryside woodland and forest conifers and on urban and peri-urban shrubs, where all six launched eradication programmes have failed (Fig. 8). For pathogens, the eradication success rate on different host plants was overall similar in countryside and urban and peri/urban environments (Fig. 8), albeit slightly lower on broadleaves in woodlands and forests (30%) and on ornamentals in urban environments (44%). The only exception was the high success rate reported for pathogens attacking conifer pests in urban environments with an 80% success rate.

The results of the Cramers’ V index for the nominal variables highlighted a strong correlation between some of the variables (Suppl. material 2). For arthropods strong associations (V > 0.5) were estimated for variables mainly related with species traits, such as feeding guild and a number of other species traits (Suppl. material 2). Also, for pathogens strong associations were estimated between species traits. These correlations are solved by LASSO and regression trees, as both methods select variables based on their added predictive ability.

Statistical modelling

Results for establishment probability are given in Suppl. material 3. Here the results of LASSO regression and decision tree analysis are reported for the eradication success of programmes against established populations.

Arthropods

The LASSO regression results showed that the area affected at the start of the eradication programme was the most important factor affecting the outcome of eradication success. For areas ≤ 1 ha and > 1 ≤ 10 ha, success is similar, but above this threshold, there is a negative relationship between the area affected and the probability of a successful eradication (β = -0.72 for > 10 ≤ 100 ha; β = -1.29 for > 100 ≤ 1000 ha and β =-2.69 for > 1000 ha; coefficients are reported at a log odds scale). Other environmental factors affecting the outcome of eradication success were the main type of environment affected at the start of the program, with a higher success rate in confined environments than in the countryside (β = 0.66), and a slightly higher success rate in mainland than in island locations (β = 0.01). Regarding control measures, only the implementation of quarantine/movement restrictions was positively associated with eradication success (β = 0.52). For species traits, internal feeders had a higher probability of eradication success than external feeders (β = 0.66), oligophagous species had a lower probability of eradication success (β =-0.22), and the group of fruit/seed feeders and gall makers had higher eradication success (β =-0.32). Considering the species targeted, A. glabripennis was associated with a higher eradication success (β = 2.35) than species for which less than five cases were reported. The optimal penalty value (λ) for the model was 0.028.

In the regression tree analysis, the optimised tree resulted in only one split, with higher eradication success for areas below 10 ha than for larger infested areas (82.6% vs 28.6%). When the area as explanatory factor was removed, a secondary tree was obtained (Fig. 9) for which eradication success was higher when quarantine/movement restrictions were implemented (69%) than when they were not (31%). The variables host type, location and phytophagous specialisation were also excluded from this tree construction due to association with the affected area. Further tree divisions highlighted the main type of environment affected, the main reproduction method, and climate as explanatory variables affecting eradication success.

Figure 9.

Optimal classification tree (after removing the size of the affected area) for factors affecting eradication success and failures of 102 eradication programmes against non-native arthropods of woody plants in Europe. In the model, every species was given equal weight and thus the records of the same species were weighted by the inverse of the number of records per species. Light grey in bars represents successful eradication, dark grey represents failure to eradicate.

Pathogens

The LASSO regression estimated that the affected area was the most important factor associated with eradication failure and the higher the area the stronger the association (>1 ≤ 10 ha: β = -0.798; > 10 ≤ 100 ha: β = -1.825; > 100 ≤ 1000 ha: β = -3.246; > 1000 ha: β = -4.334). The type of environment affected also influenced the outcome, with success more likely in confined than in urban/peri urban (β = -0.196) and countryside environments (β = -0.237). Eradication success was more likely when the eradication programme started within the first year after detection (β = 0.369), in temperate than Mediterranean climates (β = 0.434), when surveys were conducted at least annually (β = 0.566), and when host removal alone was used compared to combined methods (β = 0.063). Possible saprotrophic species were harder to eradicate, although the effect was small (β = -0.050). At the species level, Fusarium circinatum (β = 0.771), Plum pox virus (β = 0.438) and Cryphonectria parasitica (β = 0.223) were easier to eradicate than species with lower than five eradication attempts, and Hymenoscyphus fraxineus (β = -2.013), Phytoplasma mali (β=-0.998), Xanthomonas arboricola pv. Corylina (β = -2.453), Xylella fastidiosa (β = -0.511), Citrus tristeza virus CTV (β=-0.059) and Lecanosticta acicula (β = 0.196) were harder to eradicate. The optimal penalty value (λ) for the model was 0.011.

In the regression tree analysis, the optimized tree resulted in only one split where the area was the only variable included, like the results for arthropods. However, here the separation occurred for areas below 1 ha, which had higher eradication success than larger areas (88.1% vs 34.6%). When the area was removed from the model (Fig. 10), the presence of the species in adjacent NUTSIII units resulted in the first split, with higher eradication success when the species was not yet present in adjacent NUTS. The variables host type and location were also removed due to their close association with area. Other environmental factors affecting the outcome of an eradication programme included the climate and the main type of environment affected at the start of the program. Several species traits were shown to influence the probability of a successful eradication, namely the main dispersal mechanism, the host range, the incubation period, the possibility to remain asymptomatic for long periods or indefinitely and the existence of resting spores or stages. Differences in the influence of these factors between groups were observed. Regarding control and monitoring options, the implementation of restrictions on movement/quarantine and preventive felling affected the outcome of an eradication programme.

Figure 10.

Optimal classification tree (after the area affected removed) for factors affecting eradication success and failures of 344 eradication programmes against pathogens of woody plants in Europe. In the model, every species was given equal weight and thus the records of the same species were weighted by the inverse of the number of records per species. Light grey in bars represents successful eradication, dark grey represents failure to eradicate.

An increasing number of non-native forest pests and pathogens was observed in the last century in Europe (Santini et al. 2013; Roques et al. 2016, 2020; Ghelardini et al. 2017). Here we report a concomitantly increasing number of eradication programmes conducted in Europe against pests and pathogens affecting woody plants. However, eradication was attempted for only 9% of the non-native insect species and for 12% of the pathogens. In contrast, for hundreds of species eradication was ever attempted. Nevertheless, these figures are better than those reported for alien insects in North America, where eradication was attempted for only 1.8% of established species (Liebhold et al. 2016).

We may deduce that species with higher economic or ecological impacts were those selected for eradication programmes. A low benefit: cost ratio has been suggested as one of the reasons for eradication not to be attempted (Kean et al. 2022). Still, there are species with high economic impact, such as Gonipterus platensis in the Iberian Peninsula (Cordero-Rivera et al. 1999; Valente et al. 2018) for which eradication was never attempted. Several possible reasons for not attempting to eradicate an invasive population have been proposed (Liebhold et al. 2016; Kean et al. 2022): 1) the fact that the pest or pathogen was already too widespread or abundant, or that spread was too rapid; 2) an underestimation of the potential impacts of the pest or pathogen; 3) the lack of adequate surveillance mechanisms to detect an invasion early enough; 4) the lack of effective control tools; 5) the existence of open pathways for re-introduction; 6) or policymakers did not consider eradication to be a realistic option. Since the decision to carry out an eradication programme is typically taken at the national level, and frequently imposed at the European level, this leads to responsibilities in the scientific community in transferring information to policymakers so that more species could be targeted. We hope this revision work may contribute in this regard.

At the European level, eradications were more concentrated in Western regions with minor numbers in the northern and eastern European countries. In part, this distribution coincides with the hotspots of first detections in Europe (Branco et al. 2019), that is the countries where most incursions of high-impact invaders occurred. But there is still a contrast between the high number of first detections in central-eastern countries and the low number of eradication attempts.

An optimistic conclusion of our study is that the overall rate of eradication success has been increasing over time for both pathogens and arthropods, and especially for the latter. In the last decade, eradication success attained levels of 76% for arthropods and 68% for pathogens. Yet, these figures include officially declared eradication measures taken against PPWP on imported materials or against adult insects, i.e. before establishment. The success for arthropods is similar to that reported by Tobin et al. (2014) (78%), whereas for pathogens it is higher than the 55% reported by Smith et al. (2017), with both of these studies being based on data from the Global Eradication Database GERDA. However, care should be taken with the interpretation of these results, as we found they might be overestimated. In fact, a significant part of successful programmes commonly referred to as “eradication” in EPPO and NPPO reports, and consequently in GERDA, concern non-established species; i.e. arthropods or pathogens that were still restricted to the materials with which they were introduced. We propose that these cases could be more accurately defined as interceptions not associated with import inspections. Cases in which pests and pathogens were still restricted to the materials with which they were introduced were mostly found in nurseries and in greenhouses, but a few of them were also on trees already planted in the field. An example of this latter situation was reported in 2019 on oak trees recently planted in the UK, imported from the Netherlands and Germany, which carried oak processionary moth caterpillars. These trees were destroyed and this was considered a successful eradication (UK GOV 2019). In the present study, we treated these cases as “post-border interception”. In total, we counted 241 cases in this category and the eradication success rate in these cases was 100%. This outstanding result confirms that the surveillance of plant materials imported and moved inside countries should be done with incessant efforts. A higher investment in surveillance and detection has been identically defended by many authors (Simberloff 2003; Tobin et al. 2014; Ganley and Bulman 2016; Liebhold et al. 2016). Many previous studies considered these cases as eradications as in the GERDA database. However, if one considers eradication in sensu stricto only for the programmes against established species, the rate of eradication success is significantly lower. In fact, when we consider only the established species, the overall eradication success rate decreases to 50% for arthropods and 61% for pathogens. However, despite being lower than in other reviews, this value is still relatively high. Furthermore, the success rate of programmes has been increasing in recent decades. Therefore, more optimism is justified about the likelihood of eradication programmes being successful and worthwhile.

On the other hand, eradication success relies on external drivers, and some species might be particularly difficult to eradicate which may be related to environmental factors or species traits. With this aim, we tried to understand the main factors determining eradication success.

One of the most consensual conclusions of previous studies is that eradication success is greater the smaller the affected area. Tobin et al. (2014) reported that eradication was 1.3 times less likely for every log10 increase in the infested area for arthropods, and Pluess et al. (2012a) identified a critical threshold of 4905 ha for the infested areas of alien invertebrates, plants, and plant pathogens, above which the probability of successful eradication reduced to half (66.7% vs 32.5%). Not surprisingly, our results are consistent with this conclusion, although our thresholds are different. Additionally, our analyses indicated lower thresholds than in previous studies: below thresholds of 10 ha for arthropods and below 1 ha for pathogens, the success rate is very high, corresponding to 82% for arthropods and 90% for pathogens. But above these thresholds eradication success significantly decreases. Still, an eradication success of 60% was attained between 10 ha and 100 ha for arthropods and between 1 ha to 10 ha for pathogens. This means that for infested areas up to these sizes, eradication may still be considered feasible. Nevertheless, if circumstances are favourable, even eradication of infestations over larger areas may be successful, provided sufficient resources are available. Interestingly, the decreasing slope was steeper for pathogens than for arthropods, which suggests that the factor area is likely to be even more important for the former group. The length of time elapsed since the first detection also led to a decrease in the success rate, as previously mentioned by Pluess et al. (2012b). However, it is difficult to separate the factor time from the area (as the affected areas usually increase over time). Therefore, the main recommendation is that governments and organizations involved should invest in surveillance and detection to achieve detection as soon as possible. Furthermore, preparedness in terms of having tools and plans available is critical to enable timely responses to newly detected incursions to start eradication programmes quickly and increase their likelihood of success. This also implies being well informed about new potential risk organisms and how to eradicate them when they arrive, prior to their invasion.

We observed that species with a higher number of eradication attempts are also those with the highest eradication success. This may reflect increased knowledge on how to deal with these particular invasive species, accumulated in the previous eradication attempts, which would also facilitate a quick reaction before its spreading and becoming then impossible to eradicate.

Blackburn et al. (2011) proposed a conceptual framework for biological invasions which has recently been updated by Paap et al. (2022) to accommodate forest pathogens. Both frameworks are composed by a series of stages, namely transport, introduction, establishment and spread, each stage having a particular barrier that a population needs to overcome to reach the next stage. Different management options may apply at different stages, and early detection of the invader and a fast response time, increase the feasibility of successful eradication. For instance, in nurseries and greenhouses PPWP are usually detected in the introduction phase, when populations still must overcome the limited distribution barrier. This applies to the reported cases still restricted to the primary material in which it was introduced, which had 100% eradication success. By contrast, ash dieback, caused by H. fraxineus, is an example of a species that was identified as invasive too late, when large areas of European forests were already invaded and there was no possibility to react anymore (Pautasso et al. 2013).

The similarity of symptoms to native or previously introduced species can mask the presence of invasive species for long periods, as occurred for Phytophthora cinnamomi and Heterobasidion irregulare, with similar symptoms as Phytophthora × cambivora and Heterobasidion annosum, respectively (Brasier et al. 1993; Vettraino et al. 2005; Garbelotto et al. 2022). For these cases, the development and/or implementation of new genomic biosurveillance tools is critical, so that the taxonomy of the invasive organism can be clarified unequivocally and effective eradication program can be started rapidly (Hamelin and Roe 2020; Luchi et al. 2020).

All these cases reinforce the concept that we should be able to identify potential invaders before they leave the country of origin to be prepared in advance. A good example is the case of P. ramorum, the causal agent of sudden oak death in California. The high potential risk identified early for this species, and the fear of having a similar epidemic in Europe, boosted the early detection and the rapid implementation of containment measures. However, not all potential invaders with high economic and ecological impacts have demonstrated this potential in its native region or other invaded regions. Frequently, an organism only becomes emergent in the invaded range, since resistance of native host plants and the communities of natural enemies keep them at low or imperceptible levels in its native range (Elton 1958; Jeffries and Lawton 1984; Wolfe 2002).

As expected, species found in confined or limited environments, usually subject to frequent intervention, such as greenhouses, are easier to eradicate. The same results were found by Pluess et al. (2012b). Both pathogens and arthropods became established more easily, and thus more difficult to eradicate, in urban or peri-urban areas and in the countryside (woodlands, forests or orchards) than in confined environments. An interesting outcome from our work is that urban and peri-urban areas have similar establishment probabilities and equal difficulties of eradication as countryside. The relevance of urban forests and urban trees for the establishment and spread of invasive forest species has been gaining relevance (Poland and McCullough 2006; Paap et al. 2017; Branco et al. 2019; Dale et al. 2022; Nunes et al. 2023) which is reinforced by our results.

Climate may play a role in the success of eradication programmes. Warmer climates may favour higher growth rates for arthropod populations. On the other hand, Mediterranean climates with harsher summer conditions, or a continental climate with severe winters may explain a lower probability of establishment and a higher probability of eradication success for some groups of insect pests and pathogens in these conditions. Still, a general trend of climate in the eradication success did not emerge from our analysis. The differences in the climate of origin and the one of the invaded range could have played a role in the eradication of specific species. Additional studies could address this hypothesis. Further, our dataset does not completely allow to disentangle climate effects from other factors, namely cultural and socio-economic ones.

Regarding species traits, we could associate some traits with a higher difficulty of eradication. For arthropods, the most remarkable outcome is the extremely low success in eradicating Hemipteran species. This is probably explained by several traits shared by many hemipteran species, such as the high dispersal ability, frequently mediated by wind and their difficulty of detection at low densities due to their often small size and cryptic stages, high fecundity and short life cycles. Concomitantly, in the LASSO models, species traits associated mostly with hemipterans in our group of species, such as parthenogenesis, were found to be relevant. An example is the psyllid T. erytreae, for which six eradication programmes were launched, and none succeeded, despite the huge effort invested in it. For pathogens, as expected, eradication proved to be harder for species with high saprotrophic abilities, for species dispersed by wind, for species that may remain indefinitely asymptomatic and for species with resting spores or stages. Unexpectedly, however, species with intermediate incubation periods (>1 ≤ 12 months) were overall easier to eradicate than those with shorter (≤ 1 month) or longer periods (>12 months). Short incubation periods may lead to faster population growth and dispersal thus challenging eradication efforts. The concomitant harder difficulty to eradicate species with incubation periods longer than one year may be associated with poor detection before planting infected material: if disease symptoms may appear after plantation, with a lag that may reach several years for some pathogens, the infected area may become large, hampering eradication success (Migliorini et al. 2015). This highlights the need for detection attempts on asymptomatic plants that indeed improved the eradication prospect (see later).

When calculating correlation between variables, most strong associations (V = 0.5) were obtained between pairs of different species traits, both for arthropods and pathogens. These correlations are justified given the high number of cases concentrated on only a few species. The statistical modelling used were able to deal with collinearity to an extent: for the LASSO regression when multiple variables are correlated they will be penalized leading to one unique predictor becoming important; for the tree-regression it chooses the variables that lead to the best split in the data. Nevertheless, it is important to note, that potentially some of the correlated variables could have been used as surrogate in the LASSO or tree regression.

An outstanding result of our study is that management options did not emerge as a relevant predictor variable of eradication success. This might be due to the fact that host plant removal, almost always combined with other treatments, was the commonly used management strategy for both arthropods and pathogens. Chemical control alone leads to very low success rates (25%). Other management options are very species-specific, such as the use of tree climbers for A. glabripennis monitoring, nest removal for oak and pine processionary moth control, and vector control for several vector transmitted pathogens and thus, do not allow extrapolation to general guidelines. Also, generally similar eradication measures were applied everywhere for a given species, because frequently these measures are mandatory according to European regulations.

Another main significant outcome of our review is the importance of quarantine measures for the success of arthropod eradication. For pathogens, however, the implementation of such measures was relevant only when the target organisms were wind-borne. The intensification of surveys, at least in an annual rhythm, was shown to be relevant both for the detection of pathogen infection before establishment and for the success of eradication. Giving up the efforts of surveillance and control after a while, especially when the populations are under low levels and difficult to detect, is a common error leading to unsuccessful eradication campaigns (Simberloff 2002; Liebhold and Tobin 2008; Tobin et al. 2014). Therefore, persisting in monitoring for several years after the last detection can be crucial for the success of programmes. Tobin et al. (2014) reported that the existence of a sensitive monitoring tool (such as pheromone traps) was one of the most important predictors in the success of an eradication campaign against arthropods. However, in our study, the use of a semiochemical lure neither affected the outcome of an eradication campaign nor the probability of establishment, and the use of traps for monitoring only slightly increased the eradication success. In the study by Tobin et al. (2014), however, the authors highlighted that when Lymantria dispar and Ceratitis capitata were excluded from the analysis, the use of targeted traps or lures was no longer significant in the outcome of an eradication attempt. The high efficiency of the available semiochemical lures for these two species may contrast with the lures currently available for the arthropod species targeted for eradication in Europe. For A. glabripennis, for e.g., a recent study has shown that the available pheromone traps, although recommended for monitoring and mass trapping of this insect, are inefficient at intercepting the pest (Marchioroand Faccoli 2021). For pathogens, we observed that an overall higher success rate of eradication was observed when asymptomatic plants were also sampled (60% to 92%), compared to when only symptomatic trees were sampled 51% to 74%.