Range reshuffling: Climate change, invasive species, and the case of Nothofagus forests in Aotearoa New Zealand

The impact of climate change on forest biodiversity and ecosystem services will be partly determined by the relative fortunes of invasive and native forest trees under future conditions. Aotearoa New Zealand has high conservation value native forests and one of the world's worst invasive tree problems. We assess the relative effects of habitat redistribution on native Nothofagus and invasive conifer (Pinaceae) species in New Zealand as a case study on the compounding impacts of climate change and tree invasions.


| INTRODUC TI ON
Climate change is altering the distribution of forest trees (Fei et al., 2017;McDowell et al., 2020). This is a concern because forest trees play a vital functional role in maintaining biodiversity and ecosystem services (Gamfeldt et al., 2013). As tree distributions shift, habitats and ecosystem processes may be disrupted, threatening forest biodiversity (Scheffer et al., 2001). Predictions of change in many forest systems are dire (González-Orozco et al., 2016;Scheffer et al., 2001). However, other research indicates that some tree species will expand their range as the climate warms, and these 'winners' may be both native and/or alien invasive species (Dyderski et al., 2018;Etherington et al., 2022;Kapuka et al., 2022). Subsequently, a reshuffling of species compositions in forests is a likely outcome of climate change. These ebbs and flows in forest composition will be complicated by the capacity of species to migrate, which is typically low for trees (Loarie et al., 2009), and compounded by anthropogenic landscape fragmentation. As such, conservation planning will benefit from understanding changes in potential habitat distribution for both native and exotic forests (Pecchi et al., 2019).
The archipelago of New Zealand (c. 249,960 km 2 ; Figure 1) provides an excellent model for examining compositional shifts in native and exotic forest cover with climate change. Geographic and temporal isolation has resulted in the evolution of a distinct New Zealand flora with high endemism, that was dominated by forest prior to human settlement around 800 years ago (McGlone, 1989;Wardle, 1991). Since human arrival, there has been an over 70% reduction in forest cover (Ewers et al., 2006), which has been replaced by low-statured native vegetation including grasslands and extensive areas of exotic vegetation (McGlone, 1989(McGlone, , 2001. There are major efforts to restore native forests in degraded areas (Suryaningrum et al., 2021). New Zealand also has one of the highest levels of exotic plant naturalisation in the world (Hulme, 2020), and a large exotic conifer plantation estate (1.7 million hectares), with the industry worth around $5 billion NZD (~$3.5 billion USD) per year (Brown, 2018).
Unwanted spread from these plantations and other conifer plantings has left New Zealand with one of the planet's most intractable invasive tree problems (Hulme, 2020). Invasive conifers currently cover around 1.8 million hectares of land in New Zealand (Brown, 2018), and this figure could increase to 7.5 million hectares by 2050 unless major control measures are implemented (Dickie et al., 2022). These infestations predominantly affect the low-statured native and/or exotic vegetation that replaced native forest (Dickie et al., 2022). This dynamic situation often sees invasive trees that reduce biodiversity values (Dickie et al., 2022;Peltzer, 2018) competing directly with biodiverse native vegetation (Standish et al., 2008). The extent to which the future range of native forest species and invasive conifers will overlap in New Zealand is unknown and provides an excellent case study to examine regional scale consequences of climate driven range reshuffling in forest trees.
Although New Zealand has several distinct native forest types, this research will focus on Nothofagus (Blume) forests, which is the most common forest type remaining (Newsome, 1987;Stewart et al., 1996; Figure 1) and is also widespread across the southern hemisphere. Nothofagus is a forest tree genus with 43 species that are economically and ecologically important, often forming monospecific but also mixed stands over large areas in South America, eastern Australia, New Guinea, New Caledonia and New Zealand (Veblen et al., 1996). Nothofagus species have climate-sensitive phenological traits, such as mast seeding (Kelly et al., 2013;Monks et al., 2016), and several species have temperature regulated range boundaries (Wardle, 1984;Worth et al., 2015). Biogeographic analysis shows that Nothofagus distribution patterns have been strongly influenced by past climate change (Rawlence et al., 2021;Veblen et al., 1996;Worth et al., 2009), Gondwanan vicariance and more recent dispersal (Cook & Crisp, 2005;Knapp et al., 2005;Leathwick, 1998;Rawlence et al., 2021). Despite extensive research on the ecology and biogeography of Nothofagus species (McGlone, 1985;McGlone et al., 1996;Ogden et al., 1996;Wardle, 1984Wardle, , 1988Wardle & Lee, 1990), their response to anthropogenic climate change has only received minor attention (Leathwick et al., 1996;Worth et al., 2015).
In New Zealand, there are five Nothofagus species: N. cliffortioides, N. fusca, N. menziesii, N. solandri and N. truncata. They are all endemic and can occur from sea level to the tree line, each with different light, temperature, water and soil nutrient requirements (Table S1; Ogden et al., 1996;Wardle, 1984). A preference for cool, wet, upland sites means that they largely escaped the fires that drove anthropogenic forest decline, except in parts of central-eastern South Island and south-eastern North Island (McGlone et al., 1996).
Remnants in fire-affected areas mean the current range is relatively consistent with prehuman distributions (Wardle, 1984). Nothofagus forests in New Zealand support a myriad of animal, plant and fungal species, are important for recreation and are considered a taonga (treasure) by Māori, the indigenous people of Aotearoa (the Māori name for New Zealand). Therefore, how these species respond to climate change and invasive conifers is important from cultural and forest conservation standpoints.
There are 28 conifer species that are naturalised in New Zealand, and 13 from the Pinaceae family that are especially problematic and often listed on regional pest management plans  will be the focus of this study. Many exotic conifers in New Zealand have well-dispersed seed, and this is a key predictor in their capacity to invade (Wyse & Hulme, 2021). Most are also shade-intolerant and spread easily in native grasslands, shrublands and alpine and subalpine areas, while some can also withstand shade and invade closed canopy forests including Nothofagus forest (Burmeister, 2017;Peltzer, 2018). Invaded areas have greatly reduced ecological and biological value (Dickie et al., 2014;Peltzer, 2018), and invasive conifers are considered one of the greatest threats to indigenous biodiversity in New Zealand .
These species are predicted to see shifts in distribution as climates change, with both regional expansions and contractions .
The aim of this research was to investigate how the interacting effects of climate change and invasive conifers may impact the distribution and composition of Nothofagus forests in New Zealand.
Specifically, we use correlative species distribution models (SDMs) to: (1) predict the current and future (2070) distribution of habitat for each of five Nothofagus taxa and 13 invasive conifer species in New Zealand; (2) quantify how the distribution of habitat (losses and gains) might change for each species by 2070; and (3) examine how current and future Nothofagus habitat area compares to area environmentally suitable for introduced conifer species, and therefore the degree to which Nothofagus might be in competition with invasive conifers as distributions shift in the future. Together these analyses will directly inform conservation in New Zealand and provide new insights into future dynamics of a group of the world's most invasive trees.

| ME THODS
The genus Nothofagus is partitioned into four subgenera, and some authors have elevated these subgenera to genera (Heenan & Smissen, 2013), but here we follow the widely observed taxonomy of Hill and Read (1991) (Hill et al., 2015). We built SDMs for the five native New Zealand Nothofagus species as well as 13 exotic conifer species F I G U R E 1 Native and exotic forest in Aotearoa New Zealand (Landcare Research, 2020). Grey lines show regional or relevant national park boundaries, with names mentioned in the text shown in grey. Red lines and text show 'beech gaps', which are areas anomalously devoid of Nothofagus forest (McGlone et al., 1996).
that are often listed on New Zealand regional pest management plans

| Environmental data
We selected eight climate variables from the CHELSA database version 2.1 (Karger et al., 2017(Karger et al., , 2021 for potential use in the SDMs (Table 1). These variables describe temperature and precipitation extremes because it is likely climate extremes have the strongest effects on Nothofagus species (Beigaitė et al., 2022;Zimmermann et al., 2009). For the Nothofagus SDMs, we also included soil pH and soil carbon content from the New Zealand Land Resource Information Systems (LRIS) soil database (https://smap.landc arere search.co.nz/) downloaded at 100 m 2 grid cell resolution. Soil data were not included for the conifer SDMs because it was not possible to obtain global soil data of sufficient quality. Both current and predicted future climate raster layers were downloaded from the CHELSA database at a 30-second grid cell resolution. We selected future climate projections from two of the Shared Socioeconomic Pathways (SSPs) that represent the best and worst case emissions scenarios (O'Neill et al., 2014); pathway SSP126 (sustainability, where the world shifts gradually, but pervasively, towards a more sustainable path) and SSP585 (business-as-usual, where the world continues to prioritise rapid technological progress and development of human capital in the short term). These projections were averaged from 2041 to 2070 (hereafter 'future' climates). Because future projections vary based on Global Circulation Models (GCMs) used within the SSPs, we used three different GCMs to account for uncertainty in climate change predictions. The GCMs used were GFDL-ESM4, MPI-ESM-1-2-HR and IPSL-CM6A-1R TA B L E 1 Cleaned and filtered presence records used to build species distribution models for the five native New Zealand Nothofagus species and the 13 exotic conifer species often listed on New Zealand regional pest management plans, and the environmental variables selected to model each species using a variance inflation factor value of less than 10. See footnote for variable explanations and text for the data sources.

Records
Environmental variables a it is not possible to predict future soil states. Studies have demonstrated that inclusion of edaphic variables can enhance the predictive performance of SDMs even when assumed to remain static (Bertrand et al., 2012;Stanton et al., 2012).
All climate and soil variables were reprojected to the Behrmann equal-area projection at 1 km 2 grid cell resolution using the 'raster' R package version 3.5 .

| Occurrence data
Occurrence data for the five Nothofagus species were restricted to New Zealand and collected from three sources: our own fieldwork (116 entries (van Galen et al., 2023)), the New Zealand we filtered the data in both geographic and environmental space.
We retained one record per km 2 grid cell, then used the 'occfilt_ env' function from the 'flexsdm' package version 1.3.2 (Velazco et al., 2022) with 12 bins to remove records with similar values across all environmental predictor variables. We also tested using 5 and 20 bins (more bins exclude less records), but settled on 12 as this appeared to achieve the best balance between the number of records returned and the evenness of environmental values. Final occurrence records used in the SDMs ranged from 168 to 378 for Nothofagus species and 140 to 4696 for conifer species (Table 1).

| Distribution modelling
We modelled the distribution of each species using the maximum entropy (MaxEnt) algorithm implemented using the R package 'maxnet' version 0.1.4 (Phillips, 2021;Phillips et al., 2017). MaxEnt is a presence-background approach based on an inhomogeneous Poisson process that is designed to deal with presence-only data (Phillips et al., 2017). We chose MaxEnt due to its flexibility and superior performance over other SDM algorithms that use presenceonly data (Elith et al., 2006;Valavi et al., 2022). To minimise the effects of sampling bias, we employed a target-group background approach where known locations of species with similar life history strategies are used to constrain background sampling (Phillips et al., 2009). We assembled two background datasets of occurrence records downloaded from GBIF. For Nothofagus models (which were restricted to New Zealand), these were records of all native New Zealand woody plants (list provided by Susan Walker, Manaaki Whenua-Landcare Research). For the conifer models (that included species introduced and native ranges), the global background points were extracted for a list of all Pinaceae species (see supporting data). We cleaned background records and performed geographical and environmental filtering as described above for the species' occurrence records.
We ran separate MaxEnt models for each of the 18 species, restricting SDMs of the five Nothofagus species to the extent of New Zealand but modelling the 13 conifer species at the global scale.
We initially explored a range of regularisation multiplier settings (1, 2 or 3) and feature class settings (including linear, hinge, quadratic and their interactions). Results were similar to all combinations of regularisation multiplier and feature class settings, so we followed Elith et al. (2010) by using the regularisation multiplier of 2 and the hinge feature class for all models. This was to produce smoother response curves that are more reliable when projecting to novel conditions such as under climate change (Elith et al., 2010;Merow et al., 2013). For each model, we selected background points from the Nothofagus or conifer background datasets that were within a 500 km buffer of species' occurrence records. We generated continuous model outputs for current and future climates fitted with a clog-log transformation as recommended by Phillips et al. (2017) using the 'predict' function, without restricting predictions to within the model training range (clamp = FALSE). To convert these continuous outputs to binary range maps indicating suitable/unsuitable habitat, we used a threshold calculated to maximise the sum of sensitivity (true predicted presences) and specificity (true predicted absences), a frequently recommended approach that reflects the prevalence of modelled species well (Jiménez-Valverde & Lobo, 2007;Liu et al., 2013Liu et al., , 2016 We also assessed the proximity of current presence records to new future habitat predicted under the SSP585 scenario. For this, we restricted the conifer data set to species with >50 records in New Zealand because of low data density and potential for missing data for some species (Discussion). Given the well-known distributions of Nothofagus and the dynamic nature of conifer distributions this analysis is likely to be conservative for conifers.
To evaluate the performance of SDMs, we used the 'ENMevaluate' function with the spatial block approach to partition occurrence and background data into four blocks (Roberts et al., 2017). The spatial block approach is recommended when the aim is cross-time model transfer (Roberts et al., 2017;Valavi et al., 2019). Through the four-block approach, we assessed discrimination of the SDMs with the threshold-independent Area Under the Curve (AUC) of the Receiver Operating Curve (ROC) and model overfit with the omission rate of the minimum training presence value (ORMTP) (Fielding & Bell, 1997;Muscarella et al., 2014). Values of AUC range from 0 to 1, with 1 indicating perfect discrimination and values below 0.5 indicating discrimination is no better than random. The ORMTP values range from 0 for models that are not overfitted to 1 that are overfitted. Due to concerns, traditional ROC applications such as AUC favour a broad spectra of possible commission errors and limitations derived from the choice of background records (Lobo et al., 2008), we also used partial ROC (pROC) as it focusses on predictions with acceptable levels of omission by rescaling one axis of the ROC to reflect the proportional area identified as suitable (instead of commission error as used in AUC) (Peterson et al., 2008). We calculated this using the customised function described by Kass et al. (2022) using the 'kuenm' package (Cobos et al., 2019

| Model performance
All five Nothofagus models and 13 conifer models performed well, with AUC values greater than 0.735 (Table S2). The ORMTP values were low in all cases (<0.042 for Nothofagus and <0.007 for conifers; Table S2), indicating there were no problems with overfitting. All models were statistically significant based on the partial ROC; this means the models had acceptable levels for an omission of α = 0.05.

The MESS (Multivariate Environmental Similarity Surface) maps of
Nothofagus models showed that very little area predicted as environmentally suitable was based on extrapolation, other than some area of future predictions for N. truncata in the far north of New Zealand ( Figure S1).

| Current Nothofagus suitable area
Predictions of current habitat mostly encapsulated the presence records well for all five Nothofagus species, although N. cliffortioides and N. menziesii were somewhat poorly predicted in western Fiordland ( Figure 2a-see Figure 1 for the locations of place names referred to in the text). Some areas outside of current presence records were also predicted to be environmentally suitable, particularly Interestingly, all models accurately predicted the Westland and Taranaki 'beech gaps' (Figure 1) that are devoid of Nothofagus forest. Total environmentally suitable area predicted for each species ranged from 81,790 km 2 (N. cliffortioides) to 130,178 km 2 (N. solandri; Figure 3a). Of the total land area of New Zealand included in the models (249,960 km 2 ), 87.4% (218,523 km 2 ) was predicted as environmentally suitable for at least one Nothofagus species (Figures 3a   and 4a).  (Figure 4), and so show changes in total Nothofagus or conifer area suitability.   predicted loss was also high in South Island (Figure 2d,e). When examining changes in combined total Nothofagus suitable area (Figure 4a-c), under the SSP585 scenario, total suitable area was predicted to reduce by 12.1% (Figure 3b), with ~3600 km 2 of currently suitable area lost (Figure 3c).

Despite the net loss in environmentally suitable area predicted
for all Nothofagus species (other than N. truncata under SSP126; Figure 3b), new habitat area was projected to become suitable for all species, particularly in South Island (Figures 2d,e and 3d). New area gained was greatest for N. solandri (~19,800 km 2 under SSP585), but all species were predicted to gain at least 2878 km 2 under SSP126 and 5469 km 2 under SSP585 (Figure 3d). In total, approximately 9730 km 2 of area previously unsuitable for all Nothofagus species was predicted to be gained under SSP585 (Figure 3d).

| Conflict with invasive conifers
The Examining the overlap between new area predicted to become environmentally suitable for each Nothofagus species in future (blue in Figure 2d,e) and conifer suitability showed that 60-80% of this new area is potentially also suitable for at least one conifer species (under SSP585; Figure 5; and see SSP126 Figure S5). Key conflict zones could include much of Central Otago (Figure 1) where the environment was predicted to be suitable for at least eight conifer species and up to four Nothofagus species (Figures 4 and 5). When considering the minimum migration distance required to reach this new and contested habitat in 2070 ( Figure 6, Figure S6), we see that the conifers are often closer than the Nothofagus, particularly in the large areas of new habitat in the southern South Island (Figure 6c).
However, 20% to 40% of potential new Nothofagus habitat was predicted to be unsuitable for any conifers, particularly along South Island's west coast ( Figure 5).

| DISCUSS ION
Our results indicate that habitat for both native and exotic forest in New Zealand will be substantially redistributed under future climates. Potential habitat for the Nothofagus species and invasive conifers is likely to decrease overall in response to climate change, but substantial loss in some areas will be partially offset by gains in other areas. A major finding was that much of the newly suitable habitat for native Nothofagus species in 2070 is also predicted to be suitable for at least one invasive conifer species. Key conflict zones where Nothofagus will be competing with multiple conifer species for habitat were extensive, especially in South Island. The combined effects of habitat change and competition could see widespread shifts in forest composition that favour exotic conifers. However, for all five Nothofagus species, we also identified areas of suitable habitat that are outside the current and future envelope of the wilding conifers.

| Model performance
The predicted habitat for the Nothofagus species under current conditions corresponded well with their known distributions (Wardle, 1984), except for areas in western Fiordland where both N.
menziesii and N. cliffortioides forest were relatively poorly predicted.

Western Fiordland had sparse sampling compared with eastern
Fiordland where the models performed better. The area also has extremely high annual rainfall (up to 12,000 mm; Ogden et al., 1996) and is an outlier in that Nothofagus forest is extensive to low elevations where podocarp-broadleaf forest often dominates ; Table S1). It is therefore possible that a combination of sparse sampling and environmental outliers is causing the poor prediction in this area.
It is likely that some of the invasive conifer species modelled are not in equilibrium with environmental space in New Zealand, especially those that have been recently introduced or have low numbers of records such as Pinus monticola and P. uncinata, violating an important assumption of SDMs (Araújo & Peterson, 2012). This is a common problem when modelling invasive species, and as recommended (Mainali et al., 2015), the conifer models were built on global data to provide the best possible estimate of the realised niche of these species in New Zealand. It is more likely that the Nothofagus species are in equilibrium with the environment because they have been present in New Zealand since at least the Pliocene, perhaps much longer (McGlone et al., 1996). It is possible that they may still be refilling parts of their range following the last glaciation (Leathwick, 1998;Wardle, 1988), and anthropogenic land clearing has reduced Nothofagus forest especially in central-eastern South Island and south-eastern North Island (Wardle, 1984). However, we believe that the training data represent their realised niche relatively well, because there are often remnant stands in these cleared areas (Figure 2;Wardle, 1984), and broad distributions mean the environmental conditions where the species are 'missing' are largely captured in other parts of their ranges (e.g. see MESS analysis, Figure S1).

F I G U R E 4
Overlaid predicted environmentally suitable habitat across New Zealand (based on the binary thresholds; see Figure 2) under (a,d) current conditions and (b,c,e,f) future 2070 conditions for emissions scenarios SSP126 (best case) and SSP585 (worst case) of the five native Nothofagus species and the 13 invasive conifer species. Shading shows the number of species with predicted suitable habitat in each grid cell. Individual Nothofagus maps are shown in Figure 2, and individual conifer maps are provided in Figures S2-S4. Area suitable for at least one species is 218,523 km 2 for (a), 208,430 km 2 for (b), 192,151 km 2 for (c), 233,195 km 2 for (d), 223,168 km 2 for (e), and 217,581 km 2 for (f). To compare the predicted environmental niche space of Nothofagus and conifers, (g) shows values of 10,000 randomly selected grid cells from (a) (right y-axis; green triangles) and (d) (left y-axis; purple circles) of the eight climatic variables used to build the models. Temperature variables are in °C and are the maximum temperature of the warmest month (bio5), minimum temperature of the coldest month (bio6), and mean temperatures of the warmest and coldest quarters (bio10 and bio11, respectively). Precipitation variables are in kg m −2 and show the precipitation of the wettest month (bio13), driest month (bio14), wettest quarter (bio16) and driest quarter (bio17).
The cause of conspicuous gaps in Nothofagus distributions ('beech gaps') have been keenly debated in New Zealand (Burrows, 1965;Haase, 1990;McGlone, 1985;Ogden et al., 1996;Wardle, 1964Wardle, , 1988Wardle et al., 1988;Wardle & Lee, 1990). Consensus indicates that multiple processes probably contribute to beech gaps, including dispersal barriers, environmental suitability and biotic interactions . Our results support this multiprocess explanation. The Westland and Taranaki gaps ( Figure 1) were unsuitable based on our models (Figure 2), indicating they have a climatic/environmental basis. In contrast, the Stewart Island, Central Canterbury and Manawatū gaps ( Figure 1) were predicted to be largely suitable suggesting dispersal barriers and/or biotic interactions are contributing factors in these areas.

| The redistribution of Nothofagus habitat
Unlike studies that predict some winners under climate change (Dyderski et al., 2018;Kapuka et al., 2022), the native species studied here were all predicted to see reductions in habitat suitability in 2070, except for N. truncata, which had a minor gain of habitat under emissions scenario SSP126, but a loss under SSP585. Some

F I G U R E 5
Overlap between new area gained across New Zealand for Nothofagus species in 2070 under the SSP585 (worst case) emissions scenario (blue area in Figure 2e) and the predicted suitability for conifers ( Figure 4f). Overlap under the SSP126 scenario is provided in Figure S5. Shading shows the number of conifers with predicted suitable habitat in each grid cell. Insets show how much area is predicted to be suitable for different numbers of conifer species, with text showing the percentage of area unsuitable for all conifers and the percentage suitable for at least one conifer species. As the climate changes the ability of species to persist in their current range will depend partly on phenotypic plasticity (acclimation) and adaptive genetic variation (adaptation) (Alberto et al., 2013;Valladares et al., 2014). Plasticity in ecophysiological traits has been reported in South American Nothofagus species along elevation and rainfall gradients (Mathiasen & Premoli, 2016;Premoli & Brewer, 2007), and New Zealand Nothofagus species also exhibit considerable phenotypic variation along these gradients at both local and regional scales Wardle, 1984). Adaptive genetic variation among populations has also been reported in these species (Wilcox & Ledgard, 1983), so it is possible that selection on standing genetic diversity could result in relatively rapid adaptation to new conditions (Barrett & Schluter, 2008;Bitter et al., 2019;Kremer et al., 2012). Long distance geneflow in wind pollinated trees such as Nothofagus can also facilitate regional adaptation by enabling movement of adaptive alleles among widely distribute populations (Kremer et al., 2012). However, these processes are limited by generation times in long-lived trees (Aitken et al., 2008). Therefore, while acclimation could allow time for rapid adaptive processes to operate, it will require that populations persist through substantial changes in their bioclimatic niche.
Forest microclimates can provide important biotic resistance to climate change and may delay shifts in forest distributions (De Frenne et al., 2021;Zellweger et al., 2020). Subcanopy microclimates buffer external conditions so that seedling recruitment and other ecological processes may be relatively unaffected by changes in climate (De Frenne et al., 2021). Nothofagus forests are often found in gullies and south-facing slopes that may offer some refuge from regional conditions . Indeed, many temperate Nothofagus species are thought to have survived arid and cold periods in microrefugia within otherwise unsuitable environments during glaciations (Newnham et al., 2013;Worth et al., 2009). Disturbance is likely to be a key antagonist to microclimate-based resilience (Thom et al., 2017). Nothofagus forests in New Zealand have relatively low levels of natural disturbance such as fire (Perry et al., 2014;Wyse et al., 2018), so forest microclimates may potentially delay or prevent population losses under suboptimal conditions in some areas.
However, evidence of tree decline associated with climate change is mounting globally (Carnicer et al., 2011;Davis et al., 2019;Goberville et al., 2016;Kapuka et al., 2022) and has been reported in South America Nothofagus forests (Tarabini et al., 2021). Stand dieback in New Zealand Nothofagus forests has been linked to drought (Hosking, 1993; Wardle & Allen, 1983), highlighting their vulnerability to F I G U R E 6 Proximity analysis. The minimum distance between newly suitable habitat identified under the SSP585 scenario and current occurrence records (the cleaned and filtered training data) for (a) the five Nothofagus species and (b) the five conifer species for which >50 occurrence records were present in New Zealand (Pinus contorta, Pinus nigra, Pinus pinaster, Pinus radiata, and Pseudotsuga menziesii). Grid cells predicted as newly suitable for more than one species show the average minimum distance of those species. Individual maps for each species are provided in Figure S6. (c) the difference in distance between Nothofagus (a) and conifers (b) in grid cells predicted to be newly suitable for at least one Nothofagus and one of the conifer species. Green indicates Nothofagus occurrence records are closer to the new habitat and purple indicates conifers are closer, the shading shows by how much. Nothofagus (Wardle, 1984).
The potential for Nothofagus species to exploit the new habitat predicted under future climates is likely to be limited by their poor dispersal and establishment rates. Despite occasional evidence of long distance dispersal (Burrows & Lord, 1993;Haase, 1989), Nothofagus forests have slow migration rates over geological time frames (Wardle, 1964(Wardle, , 1988, and often seemingly stable forest boundaries following contemporary disturbances (van Galen et al., 2022). The seed is relatively poorly dispersed (usually <40 m; Wardle, 1984), but establishment appears to be a key limitation after large-scale disturbances (van Galen et al., 2021;van Galen et al., 2022). Nothofagus are obligately ectomycorrhizal and new suitable habitat is often devoid of ectomycorrhizal fungi (Baylis, 1980). This symbiont limitation has been widely discussed (Baylis, 1980;Dickie et al., 2012;van Galen et al., 2021van Galen et al., , 2022; however, recent work has demonstrated the additional importance of microsite conditions and competition with grasses (van Galen et al., 2021(van Galen et al., , 2022. Much of the newly predicted habitat east of the Southern Alps (that run along the eastern Westland boundary; Figure 1) is currently dominated by low-statured vegetation including extensive areas of native and exotic grasslands (Dymond et al., 2017) where ectomycorrhizal fungi are usually absent. Therefore, despite the model predictions of new habitat under future climates, natural migration of Nothofagus forests into these areas is likely to be limited and highlights a possible need for assisted migration for these species.

| Conflict with invasive conifers
The capacity for Nothofagus species to naturally migrate could be further limited if newly suitable habitat is occupied by invasive conifer species. In stark contrast to native Nothofagus species, the invasive conifers are well-equipped for rapid migration and establishment (Dickie et al., 2022). Most have high dispersal ability (Ledgard, 2001;Wyse & Hulme, 2021), reproduce early, and grow rapidly in open communities such as shrublands and grasslands . Propagule pressure is also high due to their widespread use in forestry and amenity plantings (Peltzer, 2018). Potential conflict between conifer and Nothofagus species for new habitat is therefore a key conservation concern because there is a high likelihood that conifers will dominate (Dickie et al., 2022;Peltzer, 2018;Standish et al., 2008). Conifer invasions have profound effects on ecosystems, altering community composition, reducing species diversity and changing ecosystem processes such as nutrient cycling (Dickie et al., 2022;Nuñez et al., 2017;Salomón et al., 2018). In fact, legacy effects on soil communities often mean that even after removal of conifer infestations, recovery of native species is limited and reinvasion by conifers or other invasive plants is more likely (Dickie et al., 2022;Peltzer, 2018).
Nothofagus forest could be invaded directly by some conifer species. For example, Pseudotsuga menziesii has invaded Nothofagus forest in South America and New Zeland, reducing biodiversity value and ecosystem integrity (Burmeister, 2017;Orellana & Raffaele, 2010;Salomón et al., 2018;Simberloff et al., 2002). However, it is disturbed forest and low-statured vegetation that are most at risk from conifer invasion (Dickie et al., 2022;Nuñez et al., 2017;Simberloff et al., 2002), and invasive conifer control will likely be required to ensure the establishment and successes of native forests.
There are regions where conifers seem unlikely to persist, especially on the west coast of South Island (this study ;Etherington et al., 2022). Reassuringly, all five Nothofagus species are expected to gain new habitat where no conifer species are predicted to occur under current or future conditions. Areas without habitat for invasive conifers are restricted to high rainfall parts of the country (this study; Etherington et al., 2022), and our results indicate that this is a potentially important axis of niche differentiation between Nothofagus and the exotic conifer species. It appears that Nothofagus species possess a much broader rainfall niche in that they can occupy much wetter areas. These wet parts of the Nothofagus realised niche are likely important refugia from interactions with invasive conifers.

| Implications for future forests
Like many recent studies, our work indicates potential for major redistributions of native and exotic forest habitat in response to climate change (Dyderski et al., 2018;González-Orozco et al., 2016;Kapuka et al., 2022;Pecchi et al., 2020). Such research is typically based on correlative SDMs that describe some elements of niche space, and the actual impact on species distributions will depend on additional natural and anthropogenic factors (Araújo & Peterson, 2012). As discussed above, ecological and evolutionary processes such as plasticity and genetic variation, self-maintaining forest microclimates, species' dispersal ability and biotic interactions with other species might limit or delay actual changes in forest distributions (Alberto et al., 2013;Bouchard et al., 2019;Zellweger et al., 2020). Despite potential for delays, the scale of predicted change makes it likely that forest distributions will be impacted to some degree by changing climate. One additional complication is that our analysis lacks information on land use and management. The predicted distributions we present include urban, conservation, agriculture and forestry land that may or may not be available for afforestation in 2070. Future research should examine how interactions with land use and social policies will impact the potential for Nothofagus restoration and invasive tree management (Dickie et al., 2022). For example, computable general equilibrium (CGE) modelling offers useful methods for jointly considering social, bioclimatic and species distribution information to provide more precise guidance to conservation planners (Wei & Aaheim, 2023).
Our work highlights that climate change impacts on native for-

ACK N O WLE D G E M ENTS
We thank Kath Dickinson for helpful discussions during the development of this project. We acknowledge Susan Walker for providing the list of New Zealand woody species. We acknowledge the use of data drawn from the Natural Forest plot data collected between

CO N FLI C T O F I NTER E S T S TATEM ENT
The authors declare no conflicts of interest.

PEER 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.13767.