Native vegetation structure, landscape features and climate shape non‐native plant richness and cover in New Zealand native shrublands

Studies investigating the determinants of plant invasions rarely examine multiple factors and often only focus on the role played by native plant species richness. By contrast, we explored how vegetation structure, landscape features and climate shape non‐native plant invasions across New Zealand in mānuka and kānuka shrublands.


| INTRODUC TI ON
Non-native plant species can alter ecosystem functions through direct or indirect impacts on local biodiversity by competing with native species or altering ecosystem processes (Vilà et al., 2011(Vilà et al., , 2015. Consequently, the various factors underpinning the establishment success of non-native plant species across different habitats have been a major focus of research on plant invasions (Fridley et al., 2007;Jeschke et al., 2012;Nunez-Mir et al., 2017). Although many factors play a role in the establishment of non-native plants species into seminatural ecosystems, three have been repeatedly shown to be important: the native vegetation structure of the ecosystem (e.g. species richness and cover), landscape features (especially fragmentation and dispersal corridors associated with propagule pressure) and climate suitability (Kettenring et al., 2015;Rajaonarivelo et al., 2022;St. Clair & Bishop, 2019;Waddell et al., 2020). Yet studies that examine more than one of these factors on biological invasions are rare (Hulme, 2022).
Habitats with higher plant species diversity or richness are often expected to limit the successful establishment of non-native plant species (Beaury et al., 2020;Levine & D'Antonio, 1999). Although evidence of a negative relationship between native and non-native plant species richness is more often encountered in individual plots, positive relationships are often found when comparisons are made over large geographical areas (Peng et al., 2019;Tomasetto et al., 2019). These positive relationships may be explained by the habitat quality or heterogeneity, which would allow both native and non-native species to coexist (Fridley et al., 2007). These relationships can be disentangled if covariables such as environmental conditions or human activities are considered in the analyses (Beaury et al., 2020;Mungi et al., 2021;Nunez-Mir et al., 2017). Nevertheless, species richness is only one way to characterize the vegetation community and considering other variables such as species abundance or cover would provide complementary information on non-native species invasion success (Cubino et al., 2022;Fridley et al., 2021;Jeschke et al., 2012;Tomasetto et al., 2019). For instance, a high percentage of canopy or ground-layer vegetation cover has been found associated with lower levels of plant invasions in forests (Jagodziński et al., 2019;Lázaro-Lobo & Ervin, 2021;Mungi et al., 2021). In forest ecosystems, native species richness, tree density, canopy cover and the shrub/sapling layer are known to be important attributes limiting plant invasions (Gómez et al., 2019;Jagodziński et al., 2019).
The vertical structure of forest ecosystems may be an underappreciated factor limiting invasion success on the forest floor by limiting the amount of light reaching the surface of the ground (Fajardo & Gundale, 2018).
Features of the adjacent landscape include the degree to which the ecosystem concerned is fragmented, the nature of the adjacent matrix (in terms of sources of non-native propagules) and the presence of corridors that might facilitate non-native species dispersal.
A high level of fragmentation can strongly increase the invasibility of native forest or shrublands by allowing greater light penetration at fragment edges and subsequent exposure to sources of non-native plants from the adjacent matrix (Miller et al., 2014;Rossignaud et al., 2022;Sullivan et al., 2005;Timmins & Williams, 1991). The establishment success of non-native species often relies on a high level of propagule pressure as a source for seeds, new individuals and genetic diversity (Cassey et al., 2018;Lockwood et al., 2005;Simberloff, 2009). Roads and rivers have been identified as potential dispersal corridors allowing non-native plants to rapidly spread across landscapes (Lázaro-Lobo & Ervin, 2019, 2021Säumel & Kowarik, 2010;Timmins & Williams, 1991). These landscape features increase the connectivity between seminatural habitats and other land uses (such as non-native crops, forestry, human settlements) that may be sources of non-native plant propagules (Belote et al., 2008;Lázaro-Lobo & Ervin, 2021;Miller et al., 2014;Wiser & Allen, 2006). Climate has been acknowledged to strongly influence the probability of non-native species establishment (Bartomeus et al., 2012;Shea & Chesson, 2002;Tomasetto et al., 2013). High non-native species richness has often been observed in warmer and drier areas, which tend to overlap with anthropogenic land uses (Bartomeus et al., 2012;Fois et al., 2020;Ibanez et al., 2019;Marini et al., 2012;Pouteau et al., 2015).
Plant invasions have been widely investigated in grassland and forest ecosystems but to a lesser extent in shrublands (Peng et al., 2019). Early successional ecosystems, such as shrublands, can offer an alternative view on plant invasion in a relatively speciesrich and temporally dynamic habitat (Ecroyd & Brockerhoff, 2005;Fridley et al., 2021). Here we focus on mānuka (Leptospermum scoparium, Myrtaceae, J.R.Forst. & G.Forst.), kānuka (Kunzea ericoides, Myrtaceae, A.Rich.) shrublands, which are found across New Zealand (Allen et al., 1992;Ronghua et al., 1984;Stephens et al., 2005) and have been recently considered as threatened (De Lange et al., 2018). These early successional shrublands often establish in infertile and poorly drained environments (Allen et al., 1992;Ronghua et al., 1984;Stephens et al., 2005). Their high level of fragmentation and frequent proximity to anthropogenic disturbed habitat (e.g. pasture, crops) make these native shrublands highly susceptible to invasions (Davis et al., 2011;Ecroyd & Brockerhoff, 2005) and ideally suited to explore the role of native plant communities and landscape factors on invasion success. Understanding the importance of vegetation structure, landscape features and climate in the non-native plant invasion of mānuka and kānuka shrublands is crucial to protect this valuable habitat and its associated ecosystem services. Furthermore, New Zealand is viewed as one of the global regions experiencing the highest proportion of its flora dominated K E Y W O R D S biological invasions, biotic resistance, competition, corridors, introduction effort, resource availability, weeds by non-native species (Hulme, 2020) and yet whether the factors driving plant invasions into native habitats have parallels to other regions is not known.
To investigate how vegetation structure, landscape features and climate influence the non-native plant invasion in mānuka and kānuka shrublands, we combined multiple New Zealand datasets on the vegetation structure of 247 permanent plots present in mānuka and kānuka shrublands and their associated landscape and climate characteristics. We explored the invasion of non-native plant species in these native shrublands using non-native species richness and mean ground cover as proxies for establishment and spread (Cubino et al., 2022;Ibanez et al., 2019;Lázaro-Lobo & Ervin, 2021).
We expect high native tree richness to limit non-native plant invasion by reducing light availability (Ibanez et al., 2019;Rajaonarivelo et al., 2022;Rossignaud et al., 2022). We can also anticipate that landscape features and climate have an important influence on nonnative species establishment (Beaury et al., 2020;Mungi et al., 2021;Rajaonarivelo et al., 2022). Non-native richness and mean ground cover should be higher in warmer and drier areas and where the plots are found in smaller fragments with higher propagule pressure in the adjacent anthropogenic land cover (see Table 1 for data source and more detailed expectations). Understanding how these landscape features affect native plant species is critical to fully assess the invasibility of native habitats (Ibanez et al., 2019;Lázaro-Lobo & Ervin, 2021). Therefore, we tested whether vegetation structure,   (Allen et al., 1992;Pizzirani et al., 2019). Both species can be found in a wide range of habitats from coastal cliffs, sub-alpine shrublands to swamps (Allen et al., 1992;Stephens et al., 2005).
Although mānuka and kānuka are considered as early successional nurse plants, Allen et al. (1992) noticed a strong suppression of native and non-native species growth in the understorey of younger kānuka stands due to their high density. In general, the species richness of both natives and non-natives in the ground flora is low in these shrublands (Sullivan et al., 2007). Nevertheless, non-native plant species can establish in these shrublands, particularly at the edge of fragments but also in shaded areas in the understorey (Davis et al., 2011). A study of a single kānuka fragment over 30 years revealed a progressive increase in the representation of non-native plant species over time with a parallel decline in native species richness (Ecroyd & Brockerhoff, 2005).

| Vascular plant data
We obtained information on vascular plant species and community composition from the most recently available data (2009)(2010)(2011)(2012)(2013)(2014) of the New Zealand Land Use and Carbon Analysis System (LUCAS), which is archived in the New Zealand National Vegetation TA B L E 1 List of factors associated with vegetation structure, landscape feature or climate with data source and expected relationship with non-native plant species.
The LUCAS plant species list was corrected for synonyms, subspecies names and misspellings and the classification of all species as native or non-native was based on the inclusion of species in the New Zealand non-native flora growth form dataset (Brandt, Maule, et al., 2020) or native flora growth form dataset (Brandt, McGlone, & Richardson, 2020). Plants only identified to the genus level were classified as either native or non-native if that genus only contained species belonging to one classification or 'mixed' when both native and non-native species were known to be included in the genus. This was verified using the native and nonnative flora growth form datasets and the New Zealand plant conservation network website (https://www.nzpcn.org.nz/). Plant species with unclear or unknown identification or with a mixed status were removed from the dataset.
We only sampled plots where mānuka and/or kānuka were well established and present in tier 5 or above (>0.3 m) giving a total number of 248 plots ( Figure S1). Total native and non-native richness were calculated for each plot. We generated separate measures for three vegetation tiers: 'native canopy' (species present above 5-m height), 'native understorey' (species present in and above 0.3 m but below 5 m in height) richness as well as nonnative and native 'ground' richness (species present below 0.3 m in height). We calculated two measures of cover: cumulative cover and mean species cover. Native and non-native plant species cover classes were modified to the midpoint of the percentage cover range at each tier and summed to obtain a cumulative cover value for our three vegetation tiers (i.e. ground, understorey, canopy cover) (Holdaway et al., 2017;Wiser et al., 2002Wiser et al., , 2011. The mean species ground cover (later referred to as 'mean ground cover') was calculated for native and non-native species by dividing the cumulative ground cover by the number of species present in the ground tier (ground species richness). We used cumulative cover as a proxy for the amount of plant biomass in the ground, understorey and canopy tiers, which we expected to reflect the intensity of competitive interactions among plants (e.g. light and space) (Holdaway et al., 2017;Wiser et al., 2002Wiser et al., , 2011. By contrast, the mean ground cover of native and non-native species was used as a proxy for how well these taxa performed in the plots, assuming species with higher cover would be more suited to that environment (Cubino et al., 2022). Research). We extracted mean annual temperature (°C), mean annual minimum temperature (°C) and mean annual rainfall (mm) from

| Landscape and climate datasets
30-year normal raster files with a spatial resolution of 500 m covering a period of 1991-2020, which overlap with the vegetation survey (Wratt et al., 2006).

| Statistical analysis
Correlations among vegetation structure (native ground, understorey, canopy richness and cumulative cover and native mean ground cover), landscape features (anthropogenic land cover, distances to the nearest road and rivers) and climate variables (including the direct measures temperature, minimum temperature and rainfall, as well as the proxies altitude, longitude and latitude) were compared using the 'Hmisc' package in R (Harrell, 2021). To avoid autocorrelation, only variables with a correlation lower than 0.6 were included within the same statistical model (Tables S2 and S3).
We first investigated the correlates of non-native plant invasions by running models predicting non-native ground richness from a combination of vegetation structure, landscape features and climate variables. If, as expected, a positive association between non-native and native richness at ground level was found, we then assessed whether native ground richness was influenced by vegetation structure, landscape features, and climate in a similar way. Using only plots that had been invaded (n = 175), we used the variables we found to be associated with non-native ground richness to assess their influence on non-native mean ground cover. However, in these models, native ground richness was replaced by native cumulative ground cover as a proxy for the potential strength of competitive relationship at ground level. Although native canopy richness was correlated with cumulative canopy cover, native ground richness was not correlated with the native ground cumulative cover (Table S3).
Finally, we tested whether the relationship between non-native and native mean ground cover revealed similar associations as found for species richness.

| RE SULTS
Across the 247 plots, 175 plots had at least one non-native plant species present (70.6%) and 71 had more than 10 (28.6%). Overall, 816 native and 257 non-native species were identified including 45 only to genus level (31 native, 14 non-native). All non-native species identified were present at ground level except one (Table 2).
Only 22 plots had non-native species in the canopy tier (above 5 m).
Non-native species had an average cumulative ground cover of 12% whereas the mean ground cover of each non-native species was 0.9% (Table 2b). The average cumulative ground cover of native plant species at ground level was about 43% across the 247 plots with a mean ground cover per species of 1.3% (Table 2a).

| Non-native richness relationships
The best GAM model predicting non-native ground richness explained 59% of the deviance and included native ground richness, native canopy richness, anthropogenic land cover, distance to nearest road and river, minimum temperature and rainfall as explanatory variables (Table S4). One plot had to be removed from the analysis to balance the residuals (n = 247). Overall, vegetation structure (native canopy and ground richness) together appeared to explain greater variation than all the landscape features (Tables 3a and S6a).
Interestingly, non-native ground richness was positively associated with native richness at ground level ( Figure 1a) but negatively associated with native canopy richness (Figure 2a and S2). However, non-native ground richness only showed a clear increase when native ground richness was above 50 species, which involved only 48 plots (19%). As expected, non-native ground richness increased with adjacent anthropogenic land cover and proximity to a road or river (Table 3a, Figure S2). When considering climate, higher non-native richness was observed in drier areas regardless of the minimum temperature (Table 3a, Figures 2c and S2).

| Native richness relationships
In the same way as the model predicting non-native ground richness, the model predicting native ground richness was based on 247 plots and included native canopy richness, anthropogenic land cover, distance to nearest road and river, minimum temperature and rainfall  Figure S3). Native canopy richness explained most of the deviance. In contrast to the model predicting non-native ground richness, native ground richness was positively associated with native canopy richness (Figure 2a,d).
Similarly, native ground richness declined with increasing adjacent anthropogenic land cover (Figure 2b,e). Native ground richness was not influenced by rainfall but increased with minimum temperature before declining when the mean annual minimum temperature was more than 10°C ( Figure S3).

| Non-native cover relationships
The model predicting non-native mean ground cover explained 49% of the deviance. The same explanatory variables as the models predicting non-native species richness were considered but native cumulative ground cover replaced native ground richness (Table S4). Only two explanatory variables were significant with anthropogenic land cover explaining most of the deviance followed by native canopy richness (Tables 4a and S6a, Figure S4). Similar to the non-native richness model, non-native mean ground cover was positively associated with adjacent anthropogenic land cover ( Figure 3b) but negatively associated with native canopy richness (Figures 2a and 3a). Interestingly, non-native mean ground cover showed a strong decline until native canopy richness reached about 10 species (Figure 3b). Contrary to the model predicting non-native ground richness, native cumulative ground cover and climate did not influence non-native mean ground cover (Table 4a, Figures 1b and S4).

| Native cover relationships
The model predicting native mean ground cover used the same variables as the model predicting non-native mean ground cover but did not explain much deviance (20%). As for the model predicting non-native mean ground cover, native canopy richness and anthropogenic land cover were the only variables with significant relationships (Tables 4b and S7b, Figure S5). However, both variables were negatively associated with native mean ground cover (Figure 3c,d).

TA B L E 3
Factors influencing non-native (a) or native (b) ground richness (≤ 0.3-m height) ranked by their importance using the chisquare values with associated p-values; n = no. of plots and D 2 = deviance explained. Sign of the linear relationship described as +: positive, −: negative or ns: nonsignificant. See variable description in Table S1 and full model output in

| DISCUSS ION
Compared with shrublands elsewhere in the world (Pyšek et al., 2010;Talluto & Suding, 2008), mānuka and kānuka shrublands in New Zealand appeared on average to have a relatively low level of non-native plant invasions but the level of invasion can increase in areas with a high proportion of adjacent anthropogenic land cover.
Although 70% of the plots in this study were invaded, we measured low non-native mean species richness (6.9 spp.) with only a limited number of sites presenting non-native species above 5-m height (22 sites or 9%). Nevertheless, this level of invasion is high compared with native forest in New Zealand that has been shown to have a F I G U R E 2 Relationships between non-native or native ground richness (≤0.3-m height) and native canopy richness (above 5-m height) (a,d), proportion of anthropogenic land cover at 1-km radius (b,e) and mean annual rainfall in mm (c,f) with 95% confidence intervals. Fitted means calculated by assuming other predictors were at their mean value. Marks above the x-axis indicate the distribution of observations. ns, nonsignificant, ***p < .001.

TA B L E 4
Factors influencing (a) non-native mean ground cover and (b) native mean ground cover (≤ 0.3-m height) in invaded plots ranked by their importance using the F values with associated p-values; n = no. of plots and D 2 = deviance explained. 'Native ground cover' refers to 'native cumulative ground cover'. Sign of the linear relationship described as +: positive, −: negative or ns: nonsignificant. See variables description in Table S1 and full model output in Table S7, Figures S4 and S5. Significant relationships are highlighted in bold.  much lower level of invasion with a mean species richness of 1.45 (Rossignaud et al., 2022), while by contrast, New Zealand grasslands represented one of the most heavily invaded ecosystems with a non-native mean species richness of 16.4 (Tomasetto et al., 2013).
However, species richness only tells part of the story and the examination of cover highlights that despite the presence of several non-native plant species in the survey plots, the actual mean ground cover in shrublands was low (0.9%). While this might indicate that such species have a limited role in these ecosystems it should be noted that non-native plants have been found to have significant impacts on native richness even when at low abundance (Bernard-Verdier & Hulme, 2019).

| Role of vegetation structure, landscape features and climate
As has previously been found in continental ecosystems, high native tree richness reduced the richness and cover of non-native plants (Iannone III et al., 2015;Jagodziński et al., 2019;Lázaro-Lobo & Ervin, 2021;Rossignaud et al., 2022). High native canopy richness was associated with low non-native ground richness but not native ground richness. Greater diversity of native trees in the canopy would favour competitive exclusion by limiting access to resources, especially light (Ibanez et al., 2019;Rajaonarivelo et al., 2022). The lowlight availability under a closed canopy of tall trees is an important limiting factor for the establishment of early successional plants including non-native species (Jesson et al., 2000;Martin et al., 2009).
However, only focussing on species richness can restrict our ability to clearly identify the processes behind the patterns in non-native plant invasions. Examining species ground cover provides additional insights as it is a proxy for abundance and establishment success (Lázaro-Lobo & Ervin, 2021). Here, non-native mean ground cover declined with increasing native canopy richness suggesting native tree diversity also limits non-native species growth or spread at a site (Gómez et al., 2019).
Many studies have shown the importance of propagule pressure and disturbance on the success of non-native species establishment (Lockwood et al., 2005;Miller et al., 2014;Waddell et al., 2020).
Although propagule pressure and disturbance are often confounded in observational studies (e.g. Beaury et al., 2020), separately considering species richness and mean ground cover allowed us to assess different aspects of plant invasion (Cubino et al., 2022;Lázaro-Lobo & Ervin, 2021;Mungi et al., 2021). As expected, a high level of non-native plant richness was associated with high anthropogenic land cover and proximity to roads and rivers. High propagule pressure can benefit non-native species by increasing their chances of finding a suitable environment (Lockwood et al., 2005;Simberloff, 2009). Landscape features also influenced non-native species ground cover, particularly through the degree to which the ecosystem is fragmented. Highly fragmented native forest or shrublands can be more susceptible to invasion due to edge F I G U R E 3 Relationships between non-native or native mean ground cover (≤0.3-m height) and (a,c) native canopy richness (>5-m height), (b,d) proportion of anthropogenic land cover at 1-km radius with 95% confidence intervals. Fitted means calculated by assuming other predictors were at their mean value. Marks above the x-axis indicate the distribution of observations. *p < .01, ***p < .001.
effects that combines greater disturbance, light penetration and propagule pressure from adjacent land cover (Davis et al., 2011;Ecroyd & Brockerhoff, 2005;Rossignaud et al., 2022;Sullivan et al., 2005;Timmins & Williams, 1991). The New Zealand landscape is dominated by anthropogenic land cover such as pasture, crops or pine plantations, which host large numbers of non-native plant species (Aikio et al., 2012;Brockerhoff et al., 2003;Ecroyd & Brockerhoff, 2005;Tomasetto et al., 2013). Since the most frequent non-native plant species found in our dataset were herbaceous species frequently found in pastures and ruderal habitats (see Table S5), anthropogenic land cover could serve as source for these species, which would more easily invade fragmented native shrublands (Ecroyd & Brockerhoff, 2005;Sullivan et al., 2005;Timmins & Williams, 1991;Wiser & Allen, 2006). While the anthropogenic adjacent land cover undoubtedly contributed to the establishment success of propagules that invaded the shrublands, there seemed to be no significant effect of roads and rivers despite being potential dispersal corridors of non-native plant species into the shrublands.
We tested whether climate influenced the invasion success of non-native plant species using mean annual rainfall and mean annual minimum temperature. Climate conditions are considered to be one of the first barriers that non-native species have to overcome to become established in a new environment (Bartomeus et al., 2012;Gioria et al., 2023;Shea & Chesson, 2002;Tomasetto et al., 2013). In our study, native shrublands in high rainfall areas were less invaded by non-native species regardless of the temperature. Rainfall explained part of the non-native richness variation but showed no relationship with mean ground cover supporting its role as an environmental filter in the early stage of invasion (Bartomeus et al., 2012;Ibanez et al., 2019). However, climate can also be correlated with human population density and activity (Marini et al., 2012). The South Island of New Zealand presents strong rainfall variation between the wetter west and drier east coast (Salinger & Mullan, 1999). The west coast of the South Island not only has high rainfall but greater native forest cover. Such spatial covariation makes it difficult to disentangle only the effect of rainfall on non-native plant invasion across New Zealand (Rossignaud et al., 2022;Wiser & Allen, 2006).

| Differential response of native and nonnative species
When focussing on the ground tier, we observed a positive relationship between native and non-native richness, which has often been interpreted as native and non-native species responding similarly to environmental variables (Levine & D'Antonio, 1999;Nunez-Mir et al., 2017;Shea & Chesson, 2002). However, we found native and non-native ground richness responded differently to canopy richness, adjacent anthropogenic land use as well as temperature and rainfall. Higher native canopy richness was associated with higher native ground richness but lower mean ground cover. Non-native species in New Zealand possess different life-history traits from native species, which are more adapted to grow under native shrubland conditions (e.g. higher level of shade tolerance) (Brandt et al., 2021;Timmins & Williams, 1991;Tomasetto et al., 2013). Interestingly, non-native mean ground cover showed no relationship with native cumulative ground cover suggesting that either the habitat quality or heterogeneity allows both native and non-native species to coexist or non-native species occupy different ecological niches than native species (Fridley et al., 2007;Levine & D'Antonio, 1999;Shea & Chesson, 2002). The average cumulative cover of both native (43%) and non-native (12%) species indicates that interaction among plants at the ground tier might be weak or infrequent. Therefore, the native plant community at the ground tier would likely play a limited role in plant invasions in these shrublands.
Native and non-native plant species did not show similar responses to landscape features or climate (Lázaro-Lobo & Ervin, 2021;Pouteau et al., 2015;Tomasetto et al., 2013). Landscape features influenced native plant richness and mean ground cover with a greater proportion of anthropogenic land cover associated with lower native richness and mean ground cover but higher nonnative richness and mean ground cover (Lázaro-Lobo & Ervin, 2021).
A higher proportion of adjacent anthropogenic land cover associated with forest clearing and fragmentation may reduce the opportunities for native propagules to reach native shrublands (Sloan et al., 2016). Climate influenced native and non-native species richness differently with native richness variation partly explained by minimum temperature whereas non-native richness varied with rainfall (Pouteau et al., 2015;Tomasetto et al., 2013).
In conclusion, our results provide detailed information on nonnative plant invasions in native shrublands by considering both nonnative and native plant species responses to vegetation structure, landscape features and climate. Although New Zealand is an island nation with a high level of plant invasions, the broad patterns appeared consistent with invasions in forests and shrublands in continental regions. We showed that native canopy richness followed by adjacent anthropogenic land cover were the key factors associated with non-native plant invasions. This points out the importance of considering different vegetation tiers when assessing the relationship between native and non-native species richness. These findings show the importance of maintaining high native tree richness to support native shrubland resistance to plant invasion by limiting light availability. A high proportion of adjacent anthropogenic land cover and low native canopy richness was associated with higher non-native richness and mean ground cover. Climate and proximity to dispersal corridors only influenced non-native richness.
Interestingly, our results also suggested that anthropogenic land cover may not only favour non-native species invasion directly but also indirectly by negatively impacting the composition of the local native community. We clearly identified that mānuka and kānuka shrublands that are highly fragmented and isolated from native forest are more at risk of plant invasion. In addition, our results highlight the necessity to disentangle the impact of anthropogenic land cover on the local native community from its impact on non-native species establishment to improve management practices and conservation programmes.

ACK N O WLE D G E M ENTS
We acknowledge the use of data drawn from the National The costs of this publication have been partially met by the Lincoln University Open Access Fund.

FU N D I N G I N FO R M ATI O N
The research was funded by the New Zealand Tertiary Education Commission through its support of Bioprotection Aotearoa, a National Centre of Research Excellence.

CO N FLI C T O F I NTE R E S T S TATE M E NT
The authors declare no conflict of Interest.

PE E R R E V I E W
The peer review history for this article is available at https:// www.webof scien ce.com/api/gatew ay/wos/peer-revie w/10.1111/ ddi.13713.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available in Philip E. Hulme https://orcid.org/0000-0001-5712-0474