From the database provided by the Spanish Association of Public Parks and Gardens (AEPJP 2010; see Appendix 1), we compiled the complete species list of woody and arboreal-like plant (i.e., trees, shrubs, arboreal cacti and palm trees) present in 46 parks from 23 peninsular Spanish cities. We used the Taxon stand package of R (Cayuela & al. 2019) to select all the accepted scientific names (consulted 22 May 2020), excluding species with invalid names. The species were classified as native or non-native. Archaeophytes (i.e., species introduced before 1500 AD) were not included in the analysis because they are poorly recorded and, for many species, their non-native status is under discussion (Pyšek & al. 2004).

We checked for the invasion status of these non-native species in Spain (Sanz Elorza & al. 2004; BOE 2019; CABI 2020) following Richardson & al. (2000). Based on that, species were classified as: invasive, established, casual, and not present in the wild. We also consulted their regulatory status by the Spanish Catalogue of Non-native Invasive Species (BOE 2019) and the List of Invasive Alien Species of Union Concern (European Commission 2019). These regulations involve the ban of possession, transport and commerce of living beings and their propagules.

Information on being invasive elsewhere was consulted in the CABI datasheets (CABI 2020) and the database of the Invasive Species Specialist Group (2015). The climatic suitability was defined as the tolerance to both, the highest historical absolute minimum temperature in Spain which was 0.2°C in Almería (9 February 1935), and the highest mean minimum temperature in the coldest month in Spain which is 10.8°C in January in Tarifa (mean recorded from data between 1981 to 2010) (AEMET 2019). These tolerances were consulted in the CABI Invasive Species Compendium (CABI 2020).

For the invasive and potentially invasive species, we assigned binary scores (yes/no) to a total of 14 potential impact types, following Blackburn et al (2014). We considered six potential impacts on native species: (1) competition, (2) hybridization, (3) disease transmission, (4) parasitism, (5) poisoning, toxicity and allelopathy, and (6) interaction with other invasive non-native species; and four impacts on ecosystems: (7) nutrient cycling, (8) physical modification of the habitat, (9) modification of natural succession and (10) disruption to food webs. We also assessed potential socioeconomic impacts, including on (11) human health (such as allergenic pollen), (12) infrastructures, (13) agriculture and forestry and (14) to other sectors (e.g., livestock, domestic animals, and fish farming).

Information on impacts was consulted in the CABI Invasive Species Compendium (CABI 2020) and the database of the Invasive Species Specialist Group (2015). Information about pollen allergenicity was found in the Allergome database (Mari & al. 2009). For each species, we calculated the number of potential environmental (0-10) and socioeconomic (0-4) impacts. The threshold for the classification between species with a high and low number of impacts was the median value for both environmental and socioeconomic impacts. The Attention list was composed by invasive and potentially invasive species in Spain with environmental or socioeconomic impacts at or above the threshold. Meanwhile species with both environmental and socioeconomic impacts below the threshold formed the Watch List.

The Google Trends tool provides monthly data of up to five keywords, in a predefined geographic and temporal range. The output values of Google Trends are always relative to the maximum absolute value of all the keywords imputed, which is set as 100. To standardize the relative values of Google trends, we used the R pack “gtrendsR” v. 1.4.2 (Massicotte & Eddelbuettel 2020), carrying out a paired systematic consultation between each species and Robinia pseudoacacia L. as the control species for Spain as the geographic region, from the 1^st of January of 2004 to the 31^st of December of 2019. Robinia pseudoacacia was selected as the control species because it was the one with the highest Google Trends value.

We used an adaptation of the invasion risk assessment (WRA) protocol for Spain (Pheloung & al. 1999; Gassó & al. 2010) on all the invasive and potentially invasive species. The WRA scores range from -14 (benign species) to 29 (maximum risk). WRA assess species as ‘rejected’ for those species likely to be of high risk (score > 6); as ‘accepted’ those with a low score (< 1); and as ‘need further evaluation’, those with intermediate scores (1-6). Some of the WRA values were obtained from several previous studies (Gassó & al. 2010; Pino & al. 2015; Álvarez & al. 2016; Bayón & Vilà 2019). For the species for which WRA had not been performed before, it was calculated ‘de novo’. Most of the scientific information to calculate the WRA came from the Invasive Species Specialist Group (2015) and CABI (2020).

For each species in the Attention list, we calculated a Priority Index (PI _i ) based on Bayón & Vilà (2019) according to the following equation:

Where: PI _i = Priority Index for species i; E_i = number of environmental impacts for species i; S _i = number of socioeconomic impacts for species i; WRA _i = Weed Risk Assessment score for species i; STV _i = Standard trending value for species i. The number of potential environmental impacts was set to 11 and not to 10 to be consistent with the Priority Index applied for the first time to nursery species that also included impacts on biofouling (Bayón & Vilà 2019).

To test if invasive and potentially invasive species occurrence (i.e., number of parks present) is related to the STV and the WRA, we calculated the Pearson’s correlation coefficient, and performed dispersion dot plots by the R ggplot2 package (Wickham 2016); smooth line was performed by ‘lm’ method.

For the non-native species included in the categories of invasive or potentially invasive species, we seek to describe the occurrence of each potential impact, as well as the associations among impacts and their relative statistical weight.

To answer how frequently species with potential impacts were found in parks, we related how many species were causing each type of impact with their occurrence (mean ± SE) across urban parks. These pairs of values were represented in a scatter plot, including the average values for each variable as reference dash lines.

The complete species database underpinning the analysis, with their potential impacts, reported in this paper, are deposited in the Zenodo repository at https://doi.org/10.5281/zenodo.4295845 (Bayón & al. 2020).

From the 486 plants of the database, eight were not at the species level, or had not valid names. Of the remaining, there were 84 native species, 6 archaeophytes and 388 non-native species. From the non-native species, 16 were invasive; four are regulated by the Spanish catalogue (BOE 2019), conforming the Priority List: Acacia dealbata Link, present in four parks (8.7%); Buddleja davidii Franch., present in two parks (4.3%); Opuntia ficus-indica (L.) Mill., present in two parks (4.3%); and Ailanthus altissima (Mill.) Swingle, present in 20 parks (43.5%). This last species is also regulated by the European list (European Commission 2019).

Only one species, Ficus benjamina L., was classified in the Green list. From the rest of species, 39 were established, 72 were casual, and 33 were not found in the wild. We found information of potential impacts for only 59 invasive and potentially invasive species.

The median number for environmental impacts was four and the median for socioeconomic impacts was one. Forty-seven species (12.1% of total non-native and 79.7% of the invasive and potentially invasive species) had potential impacts at or above these threshold number of potential impacts, conforming the Attention list (Table 1) and 12 species (3.1% of the total non-native and 20.3% of the invasive and potentially invasive species) did not reach the threshold of impacts, and thus conformed the Watch list (Table 2).

Finally, the Uncertainty list (Appendix 2) has 96 species (24.7% of total non-native species). Unfortunately, we could not find information on the 228 remaining species (58.8%), and thus, they were classified in the Data Deficient list (Appendix 3).

There are 21 parks (46%) with some regulated species. On average, parks have 0.57 ± 0.11 (mean ± SE) invasive regulated species from the Priority List; more concretely, two parks have three regulated species, one park has two regulated species, and 18 parks (39%) have one regulated species. In addition, parks have 8.65 ± 0.97 invasive or potentially invasive species (i.e., species listed in the Attention and Watch Lists) from which 3.15 ± 0.39 are casual, 1.98 ± 0.24 are established, and 2.09 ± 0.27 non-regulated invasive in Spain. A very relevant finding is that we did not find a park without any invasive or potentially invasive species.

There is a significant positive correlation between the occurrence of these 59 invasive and potentially invasive species with the Standard Trending Value (Pearson’s R = 0.52; p-value < 0.001). However, the correlation between species occurrence and their WRA score was not significant (Pearson’s R = -0.23; p-value = 0.085) (Fig. 2).

Almost the 80% of the 59 invasive and potentially invasive species have the potential to compete with native species. At the ecosystem level, the most relevant potential impact is the physical modification of the habitat (71% species), followed by impacts on natural succession (56%). The most common potential socioeconomic impact is on human health, with 53% species. The least represented potential impacts are disruption of the food webs (15%), impacts on infrastructures (15%), parasitism (10%) and hybridization with native species (10%) (Fig. 3).

On average, 19.21 ± 3.59 species can cause impacts, and species with potential to cause impact are present in 6.02 ± 1.59 parks. By relating the number of species causing each impact type with their occurrence, according to their average values, we found the following patterns. First, there are three impacts (i.e., human health, physical modification of habitats and competition) that are caused by more species than average. The most frequent impact is on human health, with species causing this impact present in 7.29 ± 1.26 parks. Second, there is one impact (i.e., natural succession) caused by more species than average (33), but they are present in less parks than average (5.58 ± 0.94). Third, there are five impacts (i.e., disease transmission, interaction with other invasive species, and disruption of food webs, on nutrient cycling and on agriculture) that are also caused by less species than average, but they are present in more parks than average. Finally, five impacts (i.e., hybridization, parasitism, poisoning, toxicity and allelopathy, impacts on infrastructures and other socioeconomic impacts) are caused by less species than average and are present in less parks than average (Fig. 4).

There was no negative correlation between any pair of impacts but 18 with significant positive correlations (Fig. 5). The highest correlation was between poisoning, allelopathy and toxicity and impacts on agriculture (Pearson’s R = 0.52); toxicity is also highly related to human health (R = 0.30) and other socio-economic impacts (R=0.37). There were also high correlations, between competition, the different types of toxicity, the alteration of natural succession and the modification of the structure of the habitat, with Pearson’s coefficients between R = 0.34 and R = 0.42. In general terms, there were positive correlations not only between impacts within the same category (i.e., on native species, on ecosystems, and on socioeconomics), but also among impacts across categories.

The introduction of plants for ornamental purposes is the main deliberate pathway for plant invasions (van Kleunen & al. 2018). Unfortunately, this path includes some of the most harmful invasive plant species (Hulme 2007). In Spain, no park is free of planted invasive or potentially invasive species. One striking result is that almost half of the parks studied have at least one regulated invasive species (Priority List). In the parks analyzed, there are four invasive species (Ailanthus altissima, Acacia dealbata, Buddleja davidii and Opuntia ficus-indica) that are regulated and thus their plantation is forbidden (Fifth Transitory Provision, BOE 2013). Moreover, A. altissima, present in 43.5% of the parks studied, is also listed as invasive by the European Regulation (European Commission 2019). These species shall not be kept, bred or permitted to reproduce or grown, unless it is done in contained holding, where are physically isolated and cannot escape or spread by unauthorized persons (European Commission 2014). Therefore, these species should be removed from the Spanish urban parks.

On the other hand, our analysis identified only one species with no risk to be invasive. Ficus benjamina is the only species in our Green list. Based on the consulted scientific evidence (CABI 2020), this species can be planted safely. However, the criteria to classify a species in the Green list is that they are not established in Spain, not invasive anywhere and the climate in mainland Spain is not suitable. The propagule pressure is an important factor determining invasion (Lockwood & al. 2005, 2009), so it is worth considering that planting species in large quantities and in many locations can change this scenario. There are possibly many species in the Data Deficient list that could be moved to the Green list but our methodology is conservative and based on the precautionary principle. We cannot encourage planting non-native species for which there is not enough information on their invasion potential.

One of the highest interests for management are the species belonging to the Attention and the Watch lists. These are invasive and potentially invasive species in which we performed a detailed analysis of the impacts they can cause according to the mechanisms proposed in Blackburn & al. (2014). We recommend that species in the Attention list should be considered for regulation, especially all the species with high Priority Indexes such as Ricinus communis L. (PI = 73), Robinia pseudoacacia (PI = 73), Parkinsonia aculeata L. (PI = 67) and Eucalyptus globulus Labill. (PI = 65). The remaining species of the Attention list should be at least part of a monitoring program to prevent their spread in natural areas adjacent to parks. Moreover, the 12 species in the Watch list should also be included in early warning and active surveillance programs such as the lists of non-native species potentially capable of competing with native species, altering their genetic purity or ecological balances, set out by the Ministry of Ecological Transition of Spain (BOE 2020).

Because our method is reproducible, new available information has served to provide here some updates compared to a previous risk analysis on nursery plants (Bayón & Vilà 2019). Specifically the species Cupressus arizonica Greene and Jacaranda mimosifolia D.Don., that were in the Uncertainty list are now classified in the Watch list because the low number of impacts according to CABI (2020). However, there are still many species (96) in the Uncertainty list and up to 228 species (more than the half non-native species of the whole dataset) in the Data Deficient list, with no data available at all. This is a recurring problem that deserves to be addressed as soon as possible (Bayón & Vilà 2019). We think that periodic updates of species listing are needed, as new information is available (Gallardo & al. 2016), a task that requires a collective effort of a larger team of experts (González-Moreno & al. 2014; Roy & al. 2019).

We found a positive correlation between the standard trending values (STV) of the species and their occurrence in the parks. This result reinforces the notion that a greater prevalence of species at urban parks generates greater interest in people. This can cause a fashion-creating effect that can act as a snowball process influencing in turn the interest on some particular species to be planted for instance in private gardens. It has been proven that an increase in Google, reflected in the Google Trends statistics, is directly related to an increase in private commerce (Vosen & Schmidt 2011). Thus, the presence of invasive and potentially invasive species may be generating an interest, which translates into a greater use of these species, increasing their trade and commerce, and so the dispersion of propagules through the private breeding and cultivation, favoring invasions (Lenda & al. 2014; Hulme & al. 2018). On the other hand, promoting the use of ornamental native species, can generate an awareness effect that favors people’s interest on the native flora, and urban parks can also be considered of conservation interest (Alvey 2006; Vaz & al. 2018).

The use of non-native species in urban parks is worrisome for the impacts they can cause if they escape and invade natural ecosystems. Eighty percent of the invasive and potentially invasive species can compete with native species, 71% can modify the habitat and 52% can cause human health impacts. The species with the potential to cause these impacts have higher occurrence than average. In particular, there are more than 30 species, which can cause human health problems. Considering that these species cause these impacts due to their traits, (i.e., toxicity, allergenicity, have thorns, etc.) even if they do not establish in the wild, we recommend these species to be removed from parks.

The ability to modify native habitats, to compete with native species, to modify the natural succession, and to be toxic are impacts highly correlated to each other. Our analysis does not demonstrate causality. However, some direct links might take place. For example, allelopathy, included as a toxicity impact, might increase plant competition by influencing soil nutrient availability (Inderjit & del Moral 1997; Medina-Villar & al. 2017); competition with native plants also changes species composition and diversity, and thus, succession and habitat structure (Schoener 1974). Moreover, as this category of impact includes not only allelopathy, but also toxicity and poisoning, they can also cause impacts on human health. In fact, poisoning, allelopathy and toxicity are the impacts with the highest number of correlations, being also highly related to the impacts on agriculture because allelopathy can have an inhibitory effect on crop production (Qasem & Foy 2001), and cause livestock poisoning.

Hybridization and disruption of the food webs are not correlated to any other impact. That the disruption of the food webs is not correlated to other impacts such as competition with native species, the physical modification of the habitats or changes in nutrient cycling is an unexpected result. We think that a deeper analysis of species impact profiles is needed to test whether these unexpected result holds in other species assemblages, or whether our non-significant correlations are due to limited information from the pool of species analyzed. Finally, it is worth mentioning that we did not found cases of counterbalancing impacts, thus synergies are more common than trade-off between pairs of impacts (Vilà & Hulme 2017).

Possible positive effects of introduced non-native species have not been considered in this paper. From a cultural ecosystem service perspective, the presence of non-native trees are highly valued by aesthetics, eco-tourism and as cultural heritage (Vaz & al. 2018) and, although invasive plants are negatively perceived by most people, their management is not perceived as a high priority relative to other environmental risks (Potgieter & al. 2019). Furthermore, non-native trees can also enhance regulating ecosystem services such as climate regulation, soil fertility and erosion control. However in the context of, not only non-native, but also invasive or potentially invasive species, the valuation of impacts in ecosystem services are strongly context dependent varying across tree types, climatic and socioeconomic conditions of the area of introduction (Castro-Díez & al. 2019). For this reason, analyses as the ones described in this paper are of great scientific and management interest.