This study was initiated in the frame of the COST Action Non-Native Tree Species for European Forests – Experiences, Risks and Opportunities (FP 1403; 2014–18). From the more than 150 NNTs growing in European forests and forestry trials (Brus et al. 2019), we initially selected the 18 most important ones (Table 1), with their importance assessed according to their forest area cover (if available), and the presence in numerous European countries (Europe defined geographically, but excluding Russia and including Turkey), and/or their rapid spread. We assumed that for the species with these three characteristics, there are likely to be more studies and publications on the impacts. A species is defined as being non-native to Europe if its native range is wholly outside of Europe. Thus, tree species native to Europe but planted outside of their regional distributional range, such as Pinus nigra, Larix decidua or Picea abies were not considered as being non-native even when planted outside of their native range.

The workflow was divided into three phases. In the first phase, we searched the Web of Science (WOS) using the name of the NNT species (e.g. Prunus serotina; see Table 1; Fig. 2) as search string. The search covered all papers published until August 2021. Publications retrieved from the search were filtered in order to only retain those featuring pairwise comparisons of NNTs vs. NV regarding effects on physical and chemical soil properties and diversity attributes (abundance, species richness, diversity) of different taxonomic groups studied in European countries. We extended the comparisons with NTs to non-forest ecosystems (OE), as long as they represented a reference for naturalness in the study area. In this phase, four species (Acacia longifolia, A. saligna, Eucalyptus camaldulensis and Populus × canadensis) were excluded from further examination due to the low number of studies (n < 20). To increase the number of studies, we extended the search in the second phase to (i) scientific literature with no restriction on language, and (ii) PhD and MSc theses, or other studies published in non-WOS journals or books. Then, the statistical design of pairwise comparisons (NNTs vs. NV) of the selected studies was checked, and the results were examined for analysis of statistical significance, be it based on tests, figures with error bars, data tables allowing for calculating, e.g. t-tests, or reporting significance. Because of the great variety of indicators used in different studies for assessing differences in soil properties and diversity attributes, the parameters were aggregated, according to Tables 2, 3. Three other species (Fraxinus pennsylvanica, Abies grandis and Pinus contorta var. latifolia) were excluded from the analysis because of the low numbers of soil properties or taxa groups concerning these species.

In the third phase, we focused on NNTs having >150 comparisons (cases), where a comparison of NNTs vs. NV regarding one soil property or one species group is one case. As a result, Acer negundo (n = 21), P. sitchensis (n = 23), Pinus radiata (n = 2) and Pinus strobus (n = 8) were excluded, leaving seven species: A. dealbata, R. pseudoacacia, Quercus rubra, E. globulus, P. serotina, A. altissima and P. menziesii. For the final seven NNTs selected from the 18 focal ones, a total of 103 scientific publications (Suppl. material 1: table S1, fig. 2) with pairwise comparisons regarding soil properties and diversity attributes were included in the analysis. The number of papers found for the NNTs positively correlates with the area cover, with most studies concerning R. pseudoacacia (n = 32), P. menziesii (n = 27) and E. globulus (n = 22) (Suppl. material 1: fig. S1). For three species the number of publications was too low (P. sitchensis, n = 4; P. radiata, n = 1; and P. strobus, n = 1), even though these species are among the most widely planted NNTs in Europe.

Aggregated soil properties and diversity attributes were counted according to increasing (+1), neutral (0) or decreasing (-1) effects (p < 0.05) for the final seven NNTs. The presence of an NNT was considered to have an increasing or decreasing effect if the average values of an attribute reported for NNT stands/individuals were significantly higher or lower when compared with NV stands/individuals. The terms increasing and decreasing relate to the direction of change rather than any judgement about whether the effect on the ecosystem is beneficial/detrimental. While for diversity attributes, increasing effects translate to an increase of abundance- or diversity related attributes, increasing effects with respect to soil properties can be, for some examples, interpreted as having an adverse effect on an ecosystem. For example, an increase in C:N ratio indicates a reduction of N availability, e.g. reduced soil activity.

Due to the great variety of soil properties and diversity attributes used in the studies, comparable traits were aggregated. Cases of increasing, decreasing and neutral effects were counted and used to display differences among NNTs. The numbers then served for transformations to percentages. As these balances reflect all cases found for soil properties and diversity attributes, irrespective of whether these cases refer to similar soil properties or closely related taxonomic groups in a specific reference, possible nested cases may lead to biased results. Therefore, averages of cases per aggregated soil property and diversity attribute were also calculated reference-wise and balances were re-calculated accordingly. For example, Buchholz et al. (2015) compared different insect taxonomic groups regarding abundance, species richness and beta-diversity in R. pseudoacacia and Betula pendula stands in the city of Berlin, Germany. From 17 cases, four were significantly decreasing (-1), and in 13 cases no significant differences were found (0). For this reference, the average effect on insects was calculated as -0.24 (-4/17). Three other references reported all decreasing effects of R. pseudoacacia on insects (1× in Reif et al. 2016, 1× in Hejda et al. 2017, and 2× in Kadlec et al. 2018). Averaging for all cases affecting insects, the total effect of R. pseudoacacia was calculated as -0.38 (-8/21) for this diversity attribute. In contrast, if the average effect on insects was calculated separately for the four references (-0.24, -1, -1, -1), and then the average total effect was calculated, then the total effect was -0.81 (-3.24/4).

To summarise our results of the effects of the final seven NNTs on soil properties and diversity attributes, we used a Principal Components Analysis (PCA). Effect scores for each NNT are based on total averages. Only the effects with data available for all NNTs were considered in this analysis. All analyses and graphs were developed using the statistical software R version 4.1.3 (R Development Core Team 2022) and the packages dplyr, ggplot2, rgdal and raster.

The data underpinning the analysis reported in this paper are deposited at https://doi.org/10.16904/envidat.350.

From 780 soil property comparisons collected for the seven NNTs, the aggregated properties N (n = 223), pH (n = 152), and C:N (n = 93) were the most frequently considered properties in the studies (Fig. 4, Suppl. material 1: table S2). Except for cases regarding P. menziesii, the other soil properties received little attention, in particular cation exchange capacity (CEC), Mg, K, P and Ca.

The number of cases per species and per soil property was uneven (Fig. 4). Among the NNTs, the most studied was P. menziesii with the highest number of overall soil property case studies (n=364), followed by A. dealbata (n=135) and R. pseudoacacia (n=108). The lowest number of cases was found for Q. rubra (n=32). Of all soil property cases considered, 16.9% were decreasing, 61.5% neutral, and 21.5% increasing. Out of 56 possible combinations of eight aggregated soil properties for each of the seven NNTs, the literature review retrieved information on 49 combinations. Of these, impacts were decreasing in 19 cases (39%), neutral in eight (16%) and increasing in 22 (45%).

The following clear trends could be observed: A. dealbata increased nitrogen and phosphorus and decreased pH in soils. C:N ratio decreased, e.g. soil activity became higher, in stands of P. serotina and R. pseudoacacia. In many cases ‘no changes’ was the most common outcome per species and soil property; in particular, this was observed for A. altissima for nitrogen and pH, P. menziesii for nitrogen, pH, C:N, calcium, potassium, magnesium and CEC, and R. pseudoacacia for pH and, to some extent, also for nitrogen.

Of all cases considered, the occurrence of NNTs was recorded as having a decreasing effect in 22.4% of cases, a neutral effect in 65.4% of cases, and an increasing effect in 12.1% of cases.

The number of cases per species and per diversity attribute was more even than for soil properties (Fig. 5). The highest numbers of cases were recorded for R. pseudoacacia (n=283), P. serotina (n=269) and P. menziesii (n=251); whereas the lowest numbers of cases were recorded for E. globulus (n=156) and A. altissima (n=114). The category vascular plants was the most frequently studied taxonomic group (n=720), while several other groups were rarely studied (birds, bryophytes, mammals, lichens). Most decreasing effects were reported for E. globulus and – to a considerable extent – also for A. altissima. While in a majority of cases, P. serotina presented increasing effect balances, R. pseudoacacia had one increasing effect out of seven combinations, P. menziesii two increasing effects out of five combinations, and A. dealbata two increasing effects out of six combinations. In contrast to the reviewed effects on soil properties, the consistency of the effects on taxa groups was greater. Acacia dealbata, A. altissima, E. globulus and Q. rubra had clearly decreasing effect balances on vascular plant species diversity when compared to native counterparts.

Out of 56 possible combinations, the literature review retrieved information on 38. Out of these balances of NNTs occurrences, 25 (65.8%) had a decreasing effect, 9 (23.7%) an increasing effect, and 4 (10.5%) a neutral effect.

Effects of diversity attributes were finally compared between the two approaches of averaging cases (Fig. 6). Averages using only one value for a taxa group per reference (grey bars) corresponded quite well with averages over all diversity attributes, e.g. taxa groups per NNT (black bars). Only in a few cases, such as vascular plants and bryophytes on P. serotina and lichens on R. pseudoacacia, did the use of subordinate groups contrast with the averages per reference.

Consistently available soil properties and diversity attributes were used to analyse the different effects of NNTs by Principal Components Analysis (PCA). While cases for all NNTs were available for the soil variables N, P, C:N ratio and pH, three taxa groups (insects, other arthropods and vascular plants) served for comparisons of all NNTs (Suppl. material 1: fig. S2). In the soil biplot, E. globulus and Q. rubra tended to increase C:N ratio, e.g. decreased soil activity, and increased pH in the case of Q. rubra. In the opposite direction, A. dealbata, R. pseudoacacia and P. serotina corresponded to increased N- and P-contents as well as decreased C:N ratio, indicating increased soil activity. Equally, the presence of NNTs (except for Q. rubra and A. altissima) tended to decrease pH. While P. menziesii seemed to slightly decrease soil activity (e.g. increase C:N) and slightly deplete N and P, A. altissima corresponded to an increased P content in soils.

In contrast to the soil biplot, the biodiversity biplot resulted in complex patterns of taxa groups and NNTs that are mainly driven by the strongest signals of diversity × species interactions and distorting weaker signals (Suppl. material 1: fig. S2). While A. altissima and A. dealbata clearly decreased vascular plant diversity, most NNTs decreased both insect and arthropod diversity.

The number of comparisons between tree species non-native to Europe (NNTs) and native vegetation (NV) are an indicator of the effort that has been invested by researchers in the study of different impacts of these NNTs on native ecosystems. This effort may give us information on the importance of each combination of species impact for the scientific community (e.g. Pyšek et al. 2020). Our analyses demonstrate that the most abundant studies found on pairwise comparisons between NNTs and NV matched the widespread NNTs P. menziesii, R. pseudoacacia, E. globulus and Q. rubra, but also three NNTs (A. altissima, A. dealbata and P. serotina) with comparatively small area cover. In contrast, although the conifers P. sitchensis, P. strobus and P. contorta var. latifolia have been quite widely planted, there were too few studies with pairwise comparisons to be considered in our analyses. Most likely, the reason for the disparity between the area occupied and the number of studies (or comparisons) is the invasive status of A. altissima, A. dealbata and P. serotina in several countries, the spread of these species and their impact on native ecosystems during the last decades. Many papers dealing with these species mention their invasiveness in the respective introduction sections, serving as a justification for the study. On the other hand, for the four most studied species (P. menziesii, R. pseudoacacia, E. globulus and Q. rubra) there is no apparent relationship between the area occupied by each species and the number of cases. For example, P. menziesii, which is an economically important species in terms of timber production particularly in France and Germany, is by far the most studied species in our database with 615 cases but it occupies only one third of the area of R. pseudoacacia (Brus et al. 2019), with 391 cases.

According to our database, the number of comparisons between NNTs and NV was higher for diversity in taxonomic groups than for soil properties. There may be various reasons for this. Researchers can assess a large number of taxa groups in the same study, sometimes using the same plot, as is the case for plant studies. On the other hand, there is a much larger number of taxa to be studied than soil properties. Within the universe of different soil properties, soil nitrogen, pH and C:N, were the most studied, probably because of their ecophysiological relevance for plants and ecosystems, but also because their assessment is relatively easy affordable. As for taxonomic groups, the variation in the abundance of vascular plants was more studied than the variation of all other groups. Methodological reasons, including high costs for sorting and identifying species-rich groups such as insects can explain this imbalance. In contrast, mammals and lichens were the least studied groups of our selection, with the lowest number of cases and the lowest number of NNTs. The difficulties associated with mammal censuses at the scale most NNTs were planted is probably the main reason for the dearth of studies. As for lichens, only a few available studies point to an underrepresentation in such comparisons of NNTs and NV, a phenomenon that may produce bias in the interpretation of NNT impacts (Hulme et al. 2013).

Our results show inconsistent impacts on soil properties. Most studies show no significant effects on soil properties, indicating that in many conditions, other intrinsic local factors, namely parent bedrock, soil type or topography may be more important than the tree species. However, some soil impacts seem to be strongly related to particular tree species. This is the case of nitrogen, which is increased by the two Fabaceae species (A. dealbata and R. pseudoacacia). This is in line with the findings by Castro‐Díez et al. (2019) who found a strong phylogenetic signal in the effect of NNT on soil fertility mostly because of N-fixing species. However, soil fertility comes at the cost of soil acidification as a direct or indirect consequence of nitrogen fixation (Tang et al. 1999).

We would expect fast-growing species, such as E. globulus, to produce an increasing effect on nitrogen content due to increased productivity, which could contribute to increase the organic matter by stronger root growth and increased litter input (Evans 2009). However, this was not the case in the studies assessed here. In fact, the short-rotation silviculture (Tomé et al. 2021), and the slow litter decomposition (e.g. Pozo et al. 1998) in E. globulus stands, probably leads to lower soil nitrogen content and a concomitant increase in C:N (Castro-Díez et al. 2012; Mallen-Cooper et al. 2022).

The different taxonomic groups were, in a majority of cases negatively impacted by the studied NNTs when compared with the status of local NV. However, there are remarkable exceptions among the eight taxonomic groups examined and among the seven NNTs. With respect to microorganisms, for instance, there were two times more studies showing an increasing rather than a decreasing biodiversity. Most of these studies refer to A. dealbata. The results for this NNT may be linked to the results found for soil. The higher nutrient concentration found in most comparisons translates into a higher microbial activity, as found for example by Souza-Alonso et al. (2014). Bryophytes were also increasingly impacted, particularly by P. menziesii. Apparently, in this case the moist environment provided by closed and dense canopies of P. menziesii stands are likely to cause an increase of bryophytes, while the less shade-tolerant vascular plants declined (Finch and Szumelda 2007). Furthermore, several NNTs had no effect on biodiversity attributes of different taxonomic groups, e.g. regarding vascular plants under R. pseudoacacia (Sitzia et al. 2012; Vítková et al. 2017) and P. serotina (Chabrerie et al. 2010).

As for vascular plants, the most studied taxonomic group, different reasons may explain the increasing or decreasing biodiversity responses to NNTs, found in our review. A. dealbata, A. altissima and E. globulus were associated with marked detrimental impacts on plant diversity and abundance. In the case of A. dealbata, the reasons for the decrease have been related to light competition (Lorenzo et al. 2012), allelopathic effects (Lorenzo and Rodríguez-Echeverría, 2012), and changes in soil nutrients and microbial composition (Rodríguez-Echeverría et al. 2013). Similar reasons were referred to the decreasing plant diversity in stands of A. altissima, namely direct competition and allelopathic effects (Motard et al. 2015). The reasons behind the decreasing biodiversity response to E. globulus, may be related to intensive management practices. Eucalyptus globulus stands are usually coppiced every 10–12 years along three rotations, and the understory is often removed (Tomé et al. 2021). These frequent disturbances may contribute to the decrease of plant diversity (Lomba et al. 2011; Vaz et al. 2019). Other authors, however, point to intrinsic characteristics of E. globulus that may be associated to the impact on plant diversity, including the amount of light that reaches the soil, preventing the establishment of shade-tolerant species or the lack of seed-dispersing birds (Calviño-Cancela 2013). Allelopathy, which inhibits germination and root growth of understory plants, has been referred to by many authors (e.g. Souto et al. 2001; Becerra et al. 2018), while a more recent study did not find evidence of allelopathic effects of E. globulus on Californian native species (Nelson et al. 2021). This latter work suggests that other mechanisms, such as changes in osmotic potential and water or light acquisition, may better explain the suppression of understory in E. globulus plantations. As for P. menziesii, the reasons for lowered plant diversity and abundance are normally related with the dense cover of plantations before thinning, strongly shading the ground (Augusto et al. 2002; Finch and Szumelda 2007). In contrast, according to Budde (2006) and Podrázský et al. (2014), species richness and abundance can be increased in comparison to native stands of P. sylvestris, P. abies, F. sylvatica and Quercus spp., which obviously strongly depends on management intervention applied to P. menziesii stands aimed to support continuous growth and equally increase light transmission. Comparisons of R. pseudoacacia stands with NV resulted mostly in non-significant effects, i.e. vascular plant diversity did not differ in the paired stands. There were, however, more cases with increasing than decreasing responses of biodiversity, e.g. Buchholz et al. (2015). Examples with reduced biodiversity highlight the dominance of nitrophilous species in R. pseudoacacia stands in contrast to herbaceous vegetation in native forests (Benesperi et al. 2012), or the lack of native plant species adapted to low pH and nitrogen levels in favour of exotic and ubiquituous species in R. pseudoacacia plantations (Piwczyński et al. 2016). The fact that an increase in nitrogen was observed in only half of the cases for R. pseudoacacia vs. NV may explain why this NNT does not seem to have a general cascade effect on vascular plant diversity.

The higher numbers of increasing vs. decreasing biodiversity responses to P. serotina are surprising and reflect the context of the studies considered in the analyses. For P. serotina, many increasing cases originate from two studies by Dyderski and Jagodziński (2021a, 2021b), in which several diversity attributes are listed. Most of them refer to P. serotina growing in Pinus sylvestris stands or plantations and were compared mostly with dense and species-poor Fagus sylvatica stands. In contrast, the paper-balanced score in Fig. 6 (in black; only one average value per taxonomic group per paper), results in a distinct decreasing response in vascular plant diversity. This corresponds with many studies that report a reduction in light levels caused by the presence of P. serotina (e.g. Starfinger et al. 2003; Chabrerie et al. 2010; Dyderski and Jagodziński 2019).

In summary, it is challenging to disentangle the different factors responsible for a certain impact and to ascertain which factors are more important when it comes to cultivated NNTs (Augusto et al. 2002; Tomé et al. 2021). Similarly to E. globulus, plantations of other NNTs are often intensively managed in relatively short rotations to maximise timber production, which results in specific disturbance and light availability regimes across the cultivation cycle (Augusto et al. 2002). These disturbance regimes may include pruning, thinning, understory removal and a clear cut at the end of a rotation. Therefore, cultivation history needs to be considered to better understand the long-term impacts of NNTs on biodiversity (e.g. Carneiro et al. 2008). However, the control for management influence, allowing its separation from the intrinsic species characteristics, is not included in most studies that compared NNTs and NV. The discussion on this topic becomes even more problematic with cultivated NNTs that spread spontaneously and mix with both native and non-native species, such as in the case for E. globulus in the Iberian Peninsula (Tomé et al. 2021) and R. pseudoacacia in Central Europe (Vítková et al. 2017). More sound conclusions can be drawn from those species that are currently not cultivated and which are considered noxious weeds by national legislations, such as A. altissima or A. dealbata (Pötzelsberger et al. 2020b). Given their invasive behaviour, there is a strong probability that most stands have been naturally-regenerated and that no management operations have influenced the impacts they cause.

According to our results, the NNTs that cause the strongest impact on biodiversity are those that are phylogenetically distant from European plant species. This is in line with other studies showing the importance of congeneric plant species in the establishment and survival of other living organisms that are part of the ecosystem (Harvey et al. 2012; Spafford et al. 2013). A. dealbata, E. globulus and A. altissima are associated with the lowest biodiversity in comparison to NV. While, as previously stated, the role of management should be taken into account in the case of E. globulus, this role is not a significant factor in A. dealbata and A. altissima. Therefore, one possible explanation is the lack of eco-evolutionary history with local native species, making it more difficult to establish ecological interactions, such as pollination, herbivory, seed dispersal and entire food webs, in particular when the planted area is small and thus interactions with NV are rare. Nonetheless, there are cases of strong ecological integration of NNTs, such as the case reported for the pollination of E. globulus in Galicia, Spain (Calviño-Cancela and Neumann 2015). However, this might occur at the expense of reproductive success and maintenance of native plant populations (Arceo‐Gómez and Ashman 2016). The development of these NNT-based novel ecosystems should therefore be taken into consideration when assessing and analysing the impact of NNTs on native species (Hobbs et al. 2006). Besides intrinsic factors of NNTs related to the phylogenetic relatedness (e.g. secondary plant compounds) also structural properties (bark, canopy architecture) as well as co-introductions of associated species can be important in some cases as shown for P. menziesii (Gossner et al. 2005; Gossner and Ammer 2006; Gossner 2016).