Belowground legacies of Pinus contorta invasion and removal result in multiple mechanisms of invasional meltdown.

Plant invasions can change soil biota and nutrients in ways that drive subsequent plant communities, particularly when co-invading with belowground mutualists such as ectomycorrhizal fungi. These effects can persist following removal of the invasive plant and, combined with effects of removal per se, influence subsequent plant communities and ecosystem functioning. We used field observations and a soil bioassay with multiple plant species to determine the belowground effects and post-removal legacy caused by invasion of the non-native tree Pinus contorta into a native plant community. Pinus facilitated ectomycorrhizal infection of the co-occurring invasive tree, Pseudotsuga menziesii, but not conspecific Pinus (which always had ectomycorrhizas) nor the native pioneer Kunzea ericoides (which never had ectomycorrhizas). Pinus also caused a major shift in soil nutrient cycling as indicated by increased bacterial dominance, NO3-N (17-fold increase) and available phosphorus (3.2-fold increase) in soils, which in turn promoted increased growth of graminoids. These results parallel field observations, where Pinus removal is associated with invasion by non-native grasses and herbs, and suggest that legacies of Pinus on soil nutrient cycling thus indirectly promote invasion of other non-native plant species. Our findings demonstrate that multi-trophic belowground legacies are an important but hitherto largely unconsidered factor in plant community reassembly following invasive plant removal. Published by Oxford University Press on behalf of the Annals of Botany Company.

Introduction

class="Chemical">Noclass="Chemical">n-class="Chemical">native iclass="Chemical">nvasive placlass="Chemical">nts caclass="Chemical">n alter placlass="Chemical">nt commuclass="Chemical">nities aclass="Chemical">nd traclass="Chemical">nsform laclass="Chemical">ndscapes iclass="Chemical">n obvious aclass="Chemical">nd profouclass="Chemical">nd ways (Pejchar aclass="Chemical">nd Mooclass="Chemical">ney 2009; Ruclass="Chemical">ndel ). Visual effects of iclass="Chemical">nvasive placlass="Chemical">nts oclass="Chemical">n laclass="Chemical">ndscapes caclass="Chemical">n ofteclass="Chemical">n be rapidly reversed by removal of the iclass="Chemical">nvader. However, iclass="Chemical">nvasive placlass="Chemical">nts also have major belowgrouclass="Chemical">nd effects by chaclass="Chemical">ngiclass="Chemical">ng soil class="Chemical">n class="Species">biota and altering biogeochemical processes (Ehrenfeld 2010). These belowground effects may be more difficult to reverse, resulting in both biotic and abiotic legacies that can drive changes in subsequent plant community assembly, structure and ecosystem processes (Suding ). Management of invasive plants may further augment some of these legacies, at least in the short term, for example through large pulses of organic matter inputs or disruption of plant–soil feedbacks (e.g. Symstad 2004; Martin ). Although the belowground legacies of invasive nitrogen (N)-fixing plants are well documented (e.g. Vitousek ; Liao ), the importance of legacies involving other major soil symbioses is poorly understood (Nuñez and Dickie 2014).

Invasive pines (Pinaceae) are one of the most widespread invasive trees globally (Richardson and Rejmánek 2004; Procheş ), and several lines of evidence suggest that they may have strong belowground legacies (Chapela ; Reich ; class="Chemical">Nuñez aclass="Chemical">nd Dickie 2014). The Piclass="Chemical">naceae have recalcitraclass="Chemical">nt, low-class="Chemical">n class="Chemical">calcium litter compared with other tree species, resulting in a reduction of soil pH and accumulation of organic matter in surface horizons (Reich ). Because the Pinaceae are ectomycorrhizal, the establishment of pines is frequently associated with large changes in both soil biota and nutrient cycling (Nuñez and Dickie 2014). The establishment of ectomycorrhizal trees in plantations is associated with increased mineralization of organic P and N and a loss of soil carbon (C) in deeper soil layers (Chapela ). Although less well studied than plantations, similar processes can also occur in self-established ectomycorrhizal trees (Dickie ), suggesting that these effects are driven by the presence of the tree rather than the activity of planting or associated plantation management. In particular, Dickie demonstrated that Pinus nigra invasion into grassland was associated with a loss of soil carbon (C) in mineral soil, increased P availability and rapid declines in soil invertebrate diversity.

Dickie also observed increases in bacterial dominance of soil food webs following class="Species">P. nigra iclass="Chemical">nvasioclass="Chemical">n. This last result is at odds with the more geclass="Chemical">neral predictioclass="Chemical">n that forest ecosystems should have greater fuclass="Chemical">ngal domiclass="Chemical">naclass="Chemical">nce of soil eclass="Chemical">nergy chaclass="Chemical">nclass="Chemical">nels thaclass="Chemical">n grasslaclass="Chemical">nds (Imberger aclass="Chemical">nd Chiu 2001; Harris 2009) aclass="Chemical">nd observatioclass="Chemical">ns of higher fuclass="Chemical">ngal domiclass="Chemical">naclass="Chemical">nce iclass="Chemical">n youclass="Chemical">ng placlass="Chemical">nted class="Chemical">n class="Species">Pinus radiata compared with pasture (Macdonald ). This discrepancy between studies remains unresolved, but is likely to be important for understanding ecosystem function. Bacterial dominated food webs are top-down controlled with increased bacterial dominance of energy channels regulated by predators such as nematodes (Wardle 2002), whereas the fungal-dominated energy channel is bottom-up controlled (i.e. resource limited). If pines increase fungal dominance, soil nutrient cycling is likely to slow and nutrients would be held in fungal biomass and become less available to plants. Conversely, if pines result in increased dominance of the bacterial channel, this is likely to increase bacterial-feeding nematodes and other predators, rather than bacteria per se, and thus result in faster nutrient cycling and increased nutrient availability (Wardle 2002).

Many of the belowground effects of pines are driven by ectomycorrhizal fungal symbionts co-introduced and co-invading with pines (Schwartz ; Pringle ). Invasive Pinaceae and their ectomycorrhizal mutualists frequently establish into grasslands and shrublands dominated by arbuscular mycorrhizal associations; as a consequence, the novel enzymatic capabilities of ectomycorrhizal fungi may cause fundamental shifts in soil nutrient class="Chemical">cycliclass="Chemical">ng, C pools aclass="Chemical">nd miclass="Chemical">neralizatioclass="Chemical">n of orgaclass="Chemical">nic class="Chemical">nutrieclass="Chemical">nts (Chapela ; Checlass="Chemical">n ; Orwiclass="Chemical">n ). Particularly over the iclass="Chemical">nitial period of Piclass="Chemical">nus establishmeclass="Chemical">nt, it is possible that this miclass="Chemical">neralizatioclass="Chemical">n of orgaclass="Chemical">nic class="Chemical">nutrieclass="Chemical">nts combiclass="Chemical">ned with high levels of C iclass="Chemical">nput may result iclass="Chemical">n iclass="Chemical">ncreased class="Chemical">n class="Chemical">N and P availability to other plants, but this possibility has not been investigated.

Once established, it is likely that the presence of ectomycorrhizal fungi around established Pinus may facilitate further Pinus invasion, thus creating a positive feedback loop (Thiet and Boerner 2007). Facilitation of seedling mycorrhizal formation by established plants is commonly observed at forest margins (Dickie , 2012) and in early primary (class="Chemical">Nara aclass="Chemical">nd Hogetsu 2004) aclass="Chemical">nd secoclass="Chemical">ndary successioclass="Chemical">n (Thiet aclass="Chemical">nd Boerclass="Chemical">ner 2007). A similar patterclass="Chemical">n is observed at the edge of Piclass="Chemical">nus placlass="Chemical">ntatioclass="Chemical">ns, with iclass="Chemical">ncreased ectomycorrhizal iclass="Chemical">nfectioclass="Chemical">n of seedliclass="Chemical">ngs class="Chemical">near placlass="Chemical">ntatioclass="Chemical">n margiclass="Chemical">ns (class="Chemical">n class="Chemical">Nuñez ). Facilitation of ectomycorrhizal infection can also occur among plant species, where a pioneering species facilitates ectomycorrhizal infection of later establishing species (Horton ), potentially increasing the competitive dominance of ectomycorrhizal plants with shared symbionts.

Many studies demonstrating plant effects on soil class="Species">biota or properties have suggested that placlass="Chemical">nt-soil feedbacks are importaclass="Chemical">nt iclass="Chemical">n driviclass="Chemical">ng above aclass="Chemical">nd belowgrouclass="Chemical">nd processes, yet few studies have completed the loop by democlass="Chemical">nstraticlass="Chemical">ng the abovegrouclass="Chemical">nd coclass="Chemical">nsequeclass="Chemical">nces of these belowgrouclass="Chemical">nd chaclass="Chemical">nges (Sudiclass="Chemical">ng ). Here we determiclass="Chemical">ne how the effects of aclass="Chemical">n iclass="Chemical">nvasive tree oclass="Chemical">n soil class="Chemical">n class="Species">biota, soil chemistry and ectomycorrhizal fungal inoculum combine to influence subsequent plant communities. Management of invasive trees typically involves removal of trees, but with little or no consideration given to how belowground legacies influence the regeneration of native vegetation, with the notable exception of N-fixing trees (e.g. Heneghan ). More generally, complex multi-trophic processes in invasion, which typify belowground legacies, have been poorly recognized (Simberloff and Von Holle 1999; Nuñez ). To address these issues, we studied the widespread non-native tree species Pinus contorta in New Zealand using a site where different management histories have resulted in Pinus invasion being prevented, reduced, allowed or allowed and then subsequently removed. These treatments, in combination with a bioassay of plant–soil interactions, enabled us to test three interlinked hypotheses:

Invasive class="Species">P. contorta facilitates the ectomycorrhizal iclass="Chemical">nfectioclass="Chemical">n of ectomycorrhizal placlass="Chemical">nts, iclass="Chemical">ncludiclass="Chemical">ng coclass="Chemical">nspecifics, the coclass="Chemical">n-familial iclass="Chemical">nvasive class="Chemical">n class="Species">Pseudotsuga menziesii and the native ectomycorrhizal pioneer Kunzea ericoides, resulting in increased dominance, nutrient uptake and growth of ectomycorrhizal plant species relative to arbuscular mycorrhizal and non-mycorrhizal species.

Management history that permits the formation of closed-canopy class="Species">P. contorta results iclass="Chemical">n greatly iclass="Chemical">ncreased soil C : class="Chemical">n class="Chemical">N ratios, increased fungal dominance of soil biota, greater immobilization of N and reduced N uptake of all plants. Furthermore, the effects of these soil changes will have greater negative effects on arbuscular mycorrhizal and non-mycorrhizal species compared with ectomycorrhizal species, due to a lower ability of arbuscular mycorrhizal plants to utilize organic nutrients (Read ; Read 1991).

Pinus invasion results in a loss of total, but an increase in available, P in soils; this should result in increased P uptake by all species (Chen ).

We note that the expectation of increased fungal : bacterial dominance in Hypothesis 2 is based on the commonly reported increased fungal dominance under forest compared with grasslands (Imberger and Chiu 2001; Harris 2009; Macdonald ), but is contrary to the results for class="Species">P. nigra iclass="Chemical">nvasioclass="Chemical">n iclass="Chemical">nto grasslaclass="Chemical">nds iclass="Chemical">n class="Chemical">n class="Chemical">New Zealand (Dickie ). One goal of the present study was to determine if the findings of Dickie could be repeated in a novel site with a different Pinus species.

Methods

We studied an invasive population of class="Species">P. contorta Loudoclass="Chemical">n (class="Chemical">n class="Species">lodgepole pine) at Craigieburn Forest Park, Canterbury, New Zealand (43°9′04″S, 171°43′52″E, ∼800 m elevation, 1447 mm mean annual rainfall with >100 mm in most months, mean summer maximum 32.9 °C, mean winter minimum −8.6 °C; soils 10–15 cm depth yellow-brown earths over greywacke/argillite rock). The study system was dominated by the ectomycorrhizal tree Fuscospora cliffortioides (Hook.f.) Heenan & Smissen (formerly Nothofagus solandri var. cliffortioides) prior to burning by Māoris in the 13th century and European settlers in the 19th century, and the subsequent introduction of grazing animals (primarily Ovis aries). This produced relatively large areas of native tussock grassland and shrubland, which currently comprise about a quarter of New Zealand's land area. A large number of non-native tree species were planted from the 1950s to the early 1980s for forestry trials, re-vegetation and erosion control plantings, of which P. contorta and P. menziesii have become particularly invasive (Ledgard and Baker 1988; Ledgard 2001).

We established 24 sampling plots spanning the gradient of Pinus abundance created by an existing spatial mosaic of management activity of class="Species">P. contorta establishiclass="Chemical">ng iclass="Chemical">nto class="Chemical">native-domiclass="Chemical">nated grasslaclass="Chemical">nd/shrublaclass="Chemical">nd (Fig. 1). Caclass="Chemical">ndidate study plots (20 × 20 m) were established every 40 m aloclass="Chemical">ng traclass="Chemical">nsects ruclass="Chemical">nclass="Chemical">niclass="Chemical">ng perpeclass="Chemical">ndicular to the slope across the area, covericlass="Chemical">ng a total area of ∼1 km × 300 m. The cover of class="Chemical">n class="Species">P. contorta within each plot was estimated visually, and a stratified random sample of 24 plots was retained from a total pool of 45 candidate plots. This design achieved an efficient spread of sampling effort across the range of P. contorta cover present. Six plots fell into areas where continuous management had prevented Pinus invasion as evidenced by the presence of infrequent and small Pinus stumps (<5 cm basal diameter) and pulled seedlings (<2 cm basal diameter). Hereafter, these are referred to as ‘Seedling-Removal’; we consider them to be the closest approximation to an un-invaded ecosystem. Percent cover of P. contorta was <1 %. Seven plots fell into areas where periodic control efforts had resulted in Pinus growing to ca. 10 cm basal diameter and 3–5 m height before removal with subsequent reinvasion resulting in pines 1–2.5 m in height, hereafter termed ‘Sapling-Removal’, with a mean 20.2 % cover of P. contorta. Six plots were in areas where no control efforts had taken place, resulting in closed-canopy Pinus stands of ca. 10 m height and 147.6 % cover (summed across height tiers; ‘No-Removal’), and five plots fell into areas where a similar closed-canopy Pinus forest had been removed ca. 3 years prior to our measurement (‘Tree-Removal’), with 2.3 % P. contorta cover. All control operations had been performed by hand felling with chainsaws, with no removal of biomass and minimal disturbance to soils. We believe that the plots have been subject to invasion by P. contorta for approximately the same total time, with anecdotal evidence of establishment since at least 1975 (N. Ledgard, Scion NZ, pers. comm.; available at: https://www.youtube.com/watch?v=AEm9l3Yffgo; accessed 5 August 2014), and extensive invasion by 1988 (Ledgard and Baker 1988).

Figure 1.

Examples of the four levels of invasive P. contorta management (A) Seedling-Removal, with dominance by native shrubs, herbs and grasses along with some invasive herbs, (B) Sapling-Removal, where the periodic removal of P. contorta has prevented closed-canopy formation (this photo taken 2 years after sampling was conducted), (C) No-Removal, with tall closed-canopy Pinus and (D) Tree-Removal, where closed-canopy Pinus was cut 3 years prior to this study.

Examples of the four levels of invasive class="Species">P. contorta maclass="Chemical">nagemeclass="Chemical">nt (A) Seedliclass="Chemical">ng-Removal, with domiclass="Chemical">naclass="Chemical">nce by class="Chemical">native shrubs, herbs aclass="Chemical">nd grasses aloclass="Chemical">ng with some iclass="Chemical">nvasive herbs, (B) Sapliclass="Chemical">ng-Removal, where the periodic removal of class="Chemical">n class="Species">P. contorta has prevented closed-canopy formation (this photo taken 2 years after sampling was conducted), (C) No-Removal, with tall closed-canopy Pinus and (D) Tree-Removal, where closed-canopy Pinus was cut 3 years prior to this study.

Plant communities

All vascular plant species rooted within plots were identified to species, and percentage cover was measured using cover classes following standard vegetation survey protocols (Hurst and Allen 2007). Cover was measured within each of four height tiers (0–0.3, 0.3–2, 2–5, 5–12 m), such that total cover of a species can sum to >100 %.

Soils

Soils for chemical and biological analysis were homogenized from five locations within each plot (at the centre and at four orthogonal points 7.07 m from the centre) using a custom-made 65-mm diameter class="Chemical">metal coriclass="Chemical">ng device to sample the top 100 mm of miclass="Chemical">neral soil. This depth is the most relevaclass="Chemical">nt duriclass="Chemical">ng early establishmeclass="Chemical">nt of placlass="Chemical">nt species. Litter was geclass="Chemical">nerally sparse, except uclass="Chemical">nder the declass="Chemical">nsest piclass="Chemical">ne iclass="Chemical">n the class="Chemical">n class="Chemical">No-Removal treatment, and was not included in samples. Even under closed-canopy pine, there was an abrupt transition between litter and mineral soil and no F or H layer had developed. Soil samples were analysed for pH, total C, total N, KCl extractable NH4–N and NO3-N, total P and Bray-extractable P.

Microbial communities were characterized by class="Chemical">phospholipid fatty acid (class="Chemical">n class="Chemical">PLFA) profiling, using gas chromatography. Phospholipid fatty acid profiling was chosen over DNA-based methods as it is uniquely able to directly compare bacterial and fungal biomass (Baldrian ). We used 18 : 2ω9, 12 to represent fungi, 16 : 1ω5 to represent arbuscular mycorrhizal fungi and summed cy-17 : 0, cy-19 : 0, i-15 : 0, a-15 : 0, i-16 : 0, i-17 : 0 and 16 : 1ω7 to represent bacteria. Soil invertebrates were extracted and enumerated using the tray method (Yeates and Bongers 1999) from ca. 80 g (dry mass) of each soil sample. All enchytraeids and nematodes within the sample were counted. Approximately 100 individual nematodes from each sample were identified to nominal genus and assigned to functional guilds based on feeding types. For both PLFA and nematode data we calculated fungal : bacterial dominance as F/(F+ B) where F is the fungal PLFA marker or fungal-feeding nematode abundance and B is the sum of bacterial PLFA markers or bacterial-feeding nematode abundance. This has the advantage of a symmetric distribution under either fungal or bacterial dominance and is equivalent to the commonly calculated nematode-channel ratio for fungal dominance of nematode-feeding groups, but differs from a straight F: B ratio often reported in PLFA data as F/B.

Plant growth responses: ex situ bioassay

To measure the response of different plant species to treatment effects, we planted six plant species that co-occur in the field and represent a range of mycorrhizal status and growth forms into intact soil cores (Table 1). From the centre of each of the 24 plots we removed six, intact, 110-mm depth soil cores using a 65-mm diameter class="Chemical">metal-coriclass="Chemical">ng device directly iclass="Chemical">nto 65-mm PVC tubes oclass="Chemical">n 11 December 2007, at the start of the Austral summer. The top 10 mm of soil was sliced from the top of each sample with a kclass="Chemical">nife to remove placlass="Chemical">nts aclass="Chemical">nd root crowclass="Chemical">ns. Sampliclass="Chemical">ng equipmeclass="Chemical">nt was sterilized with a 10 % household bleach solutioclass="Chemical">n betweeclass="Chemical">n plots. Soil cores were immediately wrapped iclass="Chemical">n a class="Chemical">n class="Chemical">aluminium foil, stored in a cool storage box and returned within 8 h to a 4 °C refrigerator. The subsequent day, cores were moved to an unheated greenhouse, aluminium foil was removed, bottoms of tubes were covered with 1.5-mm mesh screen secured by adhesive tape and cores were placed into individual 100 × 100 mm aluminium trays to prevent any water flow between pots (which might have transmitted soil biota).

Table 1.

Species utilized in bioassay. Here we use ‘non-mycotrophic’ to refer to a plant species that is typically a non-host of mycorrhizal fungi, as opposed to ‘non-mycorrhizal’, which we use for individual plants that do not have any mycorrhizal infection present but that typically do host mycorrhizal fungi.

	Origin	Mycorrhizal status	Growth form
P. contorta	Exotic	Ectomycorrhizal	Tree
P. menziesii	Exotic	Ectomycorrhizal	Tree
K. ericoides	Native	Dual ecto/arbus	Tree
Coprosma robusta	Native	Arbuscular	Shrub
Carex comans	Native	Non-mycotrophic	Sedge
Poa cita	Native	Arbuscular	Grass

Soil cores were class="Chemical">watered to field capacity aclass="Chemical">nd each was placlass="Chemical">nted with two seedliclass="Chemical">ngs of oclass="Chemical">ne species (Table 1). Seeds were germiclass="Chemical">nated iclass="Chemical">n sterile perlite before beiclass="Chemical">ng traclass="Chemical">nsplaclass="Chemical">nted 1–4 weeks followiclass="Chemical">ng germiclass="Chemical">natioclass="Chemical">n (sowiclass="Chemical">ng dates were staggered to syclass="Chemical">nchroclass="Chemical">nize placlass="Chemical">nticlass="Chemical">ng oclass="Chemical">n 12 December 2007). Oclass="Chemical">n 5 February 2008, oclass="Chemical">ne raclass="Chemical">ndomly selected seedliclass="Chemical">ng was removed iclass="Chemical">n the case of both seedliclass="Chemical">ngs surviviclass="Chemical">ng. Seedliclass="Chemical">ngs were harvested oclass="Chemical">n 15 May 2008. Roots were washed from the soil, aclass="Chemical">nd both roots aclass="Chemical">nd shoots were dried aclass="Chemical">nd weighed. All abovegrouclass="Chemical">nd shoot tissue was grouclass="Chemical">nd aclass="Chemical">nd aclass="Chemical">nalysed for class="Chemical">n class="Chemical">N and P content. For the three ectomycorrhizal species (Pinus, Pseudotsuga and Kunzea), ectomycorrhizal infection was quantified under a stereomicroscope with the presence of mantle confirmed under compound microscopy. All terminal roots (‘root tips’) on seedlings were evaluated as being either non-mycorrhizal or mycorrhizal. Arbuscular mycorrhizal infection was not quantified.

Statistics

Field data were analysed by analysis of variance and Tukey's class="Disease">HSD tests to determiclass="Chemical">ne sigclass="Chemical">nificaclass="Chemical">nce, with arcsiclass="Chemical">ne square-root traclass="Chemical">nsformatioclass="Chemical">n of ratio data. Placlass="Chemical">nt commuclass="Chemical">nity compositioclass="Chemical">nal shifts amoclass="Chemical">ng maclass="Chemical">nagemeclass="Chemical">nt strategy treatmeclass="Chemical">nts were determiclass="Chemical">ned by summariziclass="Chemical">ng species compositioclass="Chemical">nal data as ordiclass="Chemical">natioclass="Chemical">n axes usiclass="Chemical">ng class="Chemical">noclass="Chemical">n-metric multi-dimeclass="Chemical">nsioclass="Chemical">nal scaliclass="Chemical">ng (class="Chemical">n class="Chemical">NMDS). We tested the significance of management treatment using nonparametric permutational multivariate analysis of variance (PERMANOVA) as implemented in the function adonis, vegan package, R version 3.01 (R Core Team 2013). Ectomycorrhizal infection of seedlings was tested for the presence/absence of ectomycorrhizas using a χ2 test. Pinus and Pseudotsuga responses to ectomycorrhizal infection were tested using ANCOVA with terms for treatment, mycorrhizal infection (% of root tips) and their interaction with non-significant terms sequentially removed. Plant growth responses were analysed by comparing the three treatments with pine establishment (Sapling-Removal, No-Removal, Tree-Removal) with the Seedling-Removal treatment. Biomass (log transformed) and N- and P-concentration responses were calculated as the difference between the treatment-mean value and the mean value from the Seedling-Removal treatment divided by the maximum of the treatment or Seedling-Removal values. This puts all seedling species on a similar response scale for the effect of pine establishment on subsequent seedling growth and nutrient concentrations. It also provides a parallel metric to a common calculation of plant–soil feedbacks with symmetry around zero (e.g. Diez ). We also tested results with log-response ratios (log (treatment/control)) and note where this makes a qualitative difference to the outcome (one instance).

Results

Vegetation survey

Vegetation was dominated by native shrubs with a smaller component of herbs (slightly >50 % exotic) and native grasses where invasive class="Species">P. contorta was preveclass="Chemical">nted from establishiclass="Chemical">ng (Seedliclass="Chemical">ng-Removal; Fig. 2A). The raclass="Chemical">nk order of the domiclass="Chemical">naclass="Chemical">nce of differeclass="Chemical">nt placlass="Chemical">nt groups was similar uclass="Chemical">nder Sapliclass="Chemical">ng-Removal aclass="Chemical">nd class="Chemical">n class="Chemical">No-Removal, but with a substantial reduction in shrub cover and increase in tree cover in the No-Removal treatment (Fig. 2B and C). Vegetation structure in the Tree-Removal treatment, in which Pinus had established and then been removed, shifted strongly with increases in non-native grasses (P3,20 < 0.0001, Tree-Removal significantly different from all other treatments, no other contrasts significant) and non-native herbs (P3,20 = 0.019; Fig. 2D, Tree-Removal significantly different from Sapling-Removal and No-Removal, no other contrasts significant). There was also significantly lower cover of native shrubs under both No-Removal and Tree-Removal relative to Seedling-Removal and Sapling-Removal (P3,20 < 0.0001).

Figure 2.

Vegetation structure as percent cover of native (grey bars) and non-native (open bars) plants of different growth forms. Cover values are summed across different height tiers, such that total cover can exceed 100 %. Error bars indicate standard errors within native (offset to left) and exotic (offset to right) categories.

Within-plot plant species richness was 42.3 ± 0.67 (mean ± standard error) in Seedling-Removal and 50.6 ± 1.8 species in Sapling-Removal, but significantly lower in class="Chemical">No-Removal (33.8 ± 4.0) aclass="Chemical">nd Tree-Removal (34.8 ± 2.2) treatmeclass="Chemical">nts (P3,20 = 0.00022). Uclass="Chemical">nder the Seedliclass="Chemical">ng-Removal treatmeclass="Chemical">nt, vegetatioclass="Chemical">n was domiclass="Chemical">nated by the class="Chemical">native Ericaceous shrubs Dracophyllum uclass="Chemical">niflora aclass="Chemical">nd class="Chemical">n class="Species">D. longifolium (36 % cover), the non-native herb Hieracium pilosella (9 % cover), the native shrub Ozothamnus letophyllus (6 %) and the native grasses Chionochloa spp (4 %) and Poa colensoi (3 %). Non-metric multi-dimensional scaling analysis showed a strong movement along NMDS axis 1 with increasing Pinus invasion (Fig. 3). Under the Tree-Removal treatment, axis 1 scores returned to similar values to the Seedling-Removal treatment, but plots were strongly separated on axis 2, reflecting dominance of non-native grasses Agrostis capillaris (22 % cover), Anthoxanthum odoratum (12 %) and the non-native herb H. pilosella (19 %). Permutational multivariate analysis of variance showed a strong effect of treatment (P3,20 < 0.001, R2 = 0.76). This included a significant treatment difference between Seedling-Removal and Tree-Removal (P1,9 = 0.003, R2 = 0.54).

Figure 3.

Non-metric multi-dimensional scaling ordination of plant communities in plots under the four management regimes (points) and the position of all native and non-native (exotic) plant species found in more than three plots (six letter codes as the first three letters of genus and species, full names are in ). Lines show the outer hull around all plots within a treatment. Arrows indicate the progressive change with increased P. contorta invasion along Axis 1, and the return of community composition on Axis 1 but strongly shifted plant community along Axis 2 following P. contorta removal.

class="Chemical">Noclass="Chemical">n-metric multi-dimeclass="Chemical">nsioclass="Chemical">nal scaliclass="Chemical">ng ordiclass="Chemical">natioclass="Chemical">n of placlass="Chemical">nt commuclass="Chemical">nities iclass="Chemical">n plots uclass="Chemical">nder the four maclass="Chemical">nagemeclass="Chemical">nt regimes (poiclass="Chemical">nts) aclass="Chemical">nd the positioclass="Chemical">n of all class="Chemical">native aclass="Chemical">nd class="Chemical">noclass="Chemical">n-class="Chemical">native (exotic) placlass="Chemical">nt species fouclass="Chemical">nd iclass="Chemical">n more thaclass="Chemical">n three plots (six letter codes as the first three letters of geclass="Chemical">nus aclass="Chemical">nd species, full class="Chemical">names are iclass="Chemical">n ). Liclass="Chemical">nes show the outer hull arouclass="Chemical">nd all plots withiclass="Chemical">n a treatmeclass="Chemical">nt. Arrows iclass="Chemical">ndicate the progressive chaclass="Chemical">nge with iclass="Chemical">ncreased class="Chemical">n class="Species">P. contorta invasion along Axis 1, and the return of community composition on Axis 1 but strongly shifted plant community along Axis 2 following P. contorta removal.

Soil chemistry and biota

Soil pH was significantly lower under class="Chemical">No-Removal thaclass="Chemical">n uclass="Chemical">nder Seedliclass="Chemical">ng-Removal or Sapliclass="Chemical">ng-Removal treatmeclass="Chemical">nts, but soil C, total class="Chemical">n class="Chemical">N and C : N did not differ among treatments (Table 2). Despite the lack of change in total N, there was a 14- to 17-fold increase in NO3-N in the Tree-Removal treatment compared with the other treatments (Table 2). Total P was 1.5× higher in the Tree-Removal treatment than in the Sapling-Removal treatment, but no other contrasts were significant (Table 2). Available P, in contrast, showed up to a 3.2× increase in Tree-Removal relative to Seedling-Removal or Sapling-Removal treatment and was also 2.5× higher in the No-Removal relative to Seedling-Removal treatment (Table 2).

Table 2.

Soil abiotic and biotic responses to pine management strategy, means and standard errors. Significant responses shown in bold, values sharing the same superscript letter are not significantly different (P < 0.05, Tukey's HSD test). P value for NO3-N, total P and available P based on log-transformation to reduce inequality of variances. †Both PLFA fungal dominance and nematode feeding fungal channel dominance calculated as F/(F+ B). A value of 1 would indicate complete fungal dominance, and a value of 0 complete bacterial dominance. Mean separation and P value are based on arcsine square-root transformation.

	Pine removal treatment
	Seedling-Removal	Sapling-Removal	No-Removal	Tree-Removal	P3,20
Soil chemistry
pH	5.2±0.031a	5.1±0.024a	4.9±0.026b	5.0±0.12ab	0.0058
C (%)	6.2 ± 0.22	5.8 ± 0.26	5.5 ± 0.34	6.0 ± 0.37	0.46
Total N (%)	0.35 ± 0.014	0.31 ± 0.017	0.30 ± 0.025	0.33 ± 0.029	0.34
C : N ratio	18 ± 0.35	19 ± 0.54	19 ± 0.66	18 ± 0.66	0.45
NH4–N (mg kg−1)	13 ± 3.1	10 ± 2.2	12 ± 1.0	19 ± 7.0	0.41
NO3-N (mg kg−1)	0.34±0.03a	0.26±0.01a	0.41±0.13a	5.8±4.5b	0.0064
Total P (mg kg−1)	770±67ab	560±23a	720±120ab	840±59b	0.035
Available P (mg kg−1)	5.9±1.3a	6.1±0.92ab	15±4.2bc	19±4.4c	0.0018
Soil microbial community (PLFA, relative C19)
Total PLFA (nmol g−1)	268 ± 16	277 ± 7.2	292 ± 14	266 ± 19	0.55
Fungi (nmol g−1)	44±4.0a	48±2.3a	44±4.0a	26±5.2b	0.0051
Bacteria (nmol g−1)	76 ± 3.9	76 ± 2.3	88 ± 5.6	82 ± 2.8	0.11
Fungal dominance†	0.36±0.016a	0.38±0.014a	0.33±0.027a	0.24±0.031b	0.0010
AMF (nmol g−1)	11 ± 0.61	11 ± 0.70	10 ± 1.4	12 ± 1.3	0.44
Soil invertebrate community
Enchytraeids (g−1)	0.24±0.12a	0.36±0.07a	0.90±0.22a	2.4±0.79b	0.0012
Nematodes (g−1)	4.1 ± 1.0	5.0 ± 1.3	5.3 ± 1.1	8.8 ± 2.1	0.15
Bacterial-feeding nematodes (g−1)	1.3±0.34a	1.5±0.19a	3.3±0.71ab	4.7±1.6b	0.014
Fungal-feeding nematodes (g−1)	0.87 ± 0.34	0.60 ± 0.12	0.15 ± 0.04	0.41 ± 0.25	0.14
Fungal channel dominance†	0.36±0.05a	0.29±0.03ab	0.05±0.02c	0.15±0.11bc	0.0019
Plant feeding + associated (g−1)	1.2 ± 0.3	2.2 ± 1.2	0.98 ± 0.16	1.6 ± 0.47	0.70
Omnivorous nematodes (g−1)	0.51 ± 0.04	0.49 ± 0.1	0.68 ± 0.28	1.5 ± 0.59	0.058
Predacious nematodes (g−1)	0.20 ± 0.043	0.24 ± 0.10	0.25 ± 0.06	0.50 ± 0.19	0.21

Soil abiotic and biotic responses to pine management strategy, means and standard errors. Significant responses shown in bold, values sharing the same superscript letter are not significantly different (P < 0.05, Tukey's class="Disease">HSD test). P value for class="Chemical">n class="Chemical">NO3-N, total P and available P based on log-transformation to reduce inequality of variances. †Both PLFA fungal dominance and nematode feeding fungal channel dominance calculated as F/(F+ B). A value of 1 would indicate complete fungal dominance, and a value of 0 complete bacterial dominance. Mean separation and P value are based on arcsine square-root transformation.

Total microbial class="Chemical">PLFA did class="Chemical">not chaclass="Chemical">nge with treatmeclass="Chemical">nt. There were, however, large shifts iclass="Chemical">n microbial commuclass="Chemical">nity compositioclass="Chemical">n iclass="Chemical">n the Tree-Removal treatmeclass="Chemical">nt relative to other treatmeclass="Chemical">nts (Table 2), with a decliclass="Chemical">ne iclass="Chemical">n the quaclass="Chemical">ntity of fuclass="Chemical">ngal class="Chemical">n class="Chemical">PLFA markers, a marginally significant increase in bacterial PLFA markers and a shift towards increasing bacterial dominance.

Soil invertebrate communities showed a shift to increased bacterial-feeding dominance under n class="Chemical">No-Removal aclass="Chemical">nd Tree-Removal (Table 2). This was driveclass="Chemical">n largely by a 2.5–3.6× iclass="Chemical">ncrease iclass="Chemical">n bacterial-feediclass="Chemical">ng class="Chemical">nematodes relative to Seedliclass="Chemical">ng-Removal. Eclass="Chemical">nchytraeids iclass="Chemical">ncreased 10× iclass="Chemical">n abuclass="Chemical">ndaclass="Chemical">nce iclass="Chemical">n the Tree-Removal treatmeclass="Chemical">nt relative to the Seedliclass="Chemical">ng-Removal treatmeclass="Chemical">nt aclass="Chemical">nd there was a margiclass="Chemical">nally sigclass="Chemical">nificaclass="Chemical">nt iclass="Chemical">ncrease iclass="Chemical">n class="Chemical">n class="Disease">omnivorous nematodes.

Ectomycorrhizal infection of seedlings

In the ex situ bioassay of soils from all plots, ectomycorrhizas were found on 22 of the 24 Pinus seedlings, with the two exceptions being in the Seedling-Removal treatment (no significant treatment effect; χ2 test, P = 0.15). The percentage of n class="Disease">root tips infected (29.6 ± 4.4 %) iclass="Chemical">n Piclass="Chemical">nus with ectomycorrhizas showed class="Chemical">no sigclass="Chemical">nificaclass="Chemical">nt differeclass="Chemical">nce amoclass="Chemical">ng treatmeclass="Chemical">nts.

class="Chemical">No class="Chemical">n class="Species">Pseudotsuga seedlings were ectomycorrhizal in soils from the Seedling-Removal or Sapling-Removal treatments (n = 6 for both), three out of six Pseudotsuga were ectomycorrhizal in the No-Removal treatment and two out of five Pseudotsuga were ectomycorrhizal in the Tree-Removal treatment (marginally significant difference in the χ2 test for differences across all four treatments, P = 0.057). The marginal significance may have reflected low sample size. When the two treatments where pines were actively controlled (i.e. Continuous and Sapling-Removal; 0 out of 12 were ectomycorrhizal) and the two treatments where pines had formed closed-canopy forest (i.e. No-Removal and Tree-Removal; 5 out of 11 were ectomycorrhizal) are contrasted, the difference is highly significant (χ2 test, P = 0.0055). When ectomycorrhizal, Pseudotsuga seedlings had 50.6 ± 15.2 % of root tips infected, with no significant treatment difference between the No-Removal and Tree-Removal treatments. Ectomycorrhizas were not found on any Kunzea seedlings in the experiment.

Seedling biomass was positively correlated with ectomycorrhizal infection for both Pinus (log biomass, P1,22 = 0.036, R2adj = 0.15, no significant treatment effect or interaction) and class="Species">Pseudotsuga (liclass="Chemical">near, P1,21 = 0.006, R2adj = 0.48, class="Chemical">no sigclass="Chemical">nificaclass="Chemical">nt treatmeclass="Chemical">nt effect or iclass="Chemical">nteractioclass="Chemical">n). class="Chemical">n class="Chemical">Phosphorus concentrations of both seedlings were also positively correlated with ectomycorrhizal infection for both Pinus (P1,20 = 0.019, R2adj = 0.21, no significant treatment effect or interaction) and Pseudotsuga (P1,21 = 0.011, R2adj = 0.23, no significant treatment effect or interaction). Nitrogen was positively correlated with ectomycorrhizal infection for Pinus (P1,14 = 0.015) with a significant effect of tree-removal treatment (P3,14 = 0.0067), discussed below, and interaction term (P3,14 = 0.044; model R2adj = 0.59). The significant interaction was driven by a more positive effect of ectomycorrhizal infection in the No-Removal compared with other treatments. Nitrogen was not correlated with ectomycorrhizal infection for Pseudotsuga (P1,18 = 0.72).

Seedling biomass and nutrient responses

The only significant effects of soil origin on seedling biomass in the ex situ bioassay were for Carex and Poa, both of which had increased biomass in the Tree-Removal compared with the Seedling-Removal soils (Fig. 4).

Figure 4.

Biomass, percent N and percent P change in bioassay seedlings of six species planted in soils from Sapling-Removal, No-Removal and Tree-Removal relative to Seedling-Removal. Scale bars show mean and standard error of (treatment—Seedling-Removal)/maximum (treatment, Seedling-Removal). Values significantly different from zero are indicated by asterisks. The significant increase in Kunzea P in the Tree-Removal treatment was only marginally significant as a log-response ratio (log (treatment/control)). All other results were qualitatively similar.

Biomass, percent class="Chemical">N aclass="Chemical">nd perceclass="Chemical">nt P chaclass="Chemical">nge iclass="Chemical">n bioassay seedliclass="Chemical">ngs of six species placlass="Chemical">nted iclass="Chemical">n soils from Sapliclass="Chemical">ng-Removal, class="Chemical">n class="Chemical">No-Removal and Tree-Removal relative to Seedling-Removal. Scale bars show mean and standard error of (treatment—Seedling-Removal)/maximum (treatment, Seedling-Removal). Values significantly different from zero are indicated by asterisks. The significant increase in Kunzea P in the Tree-Removal treatment was only marginally significant as a log-response ratio (log (treatment/control)). All other results were qualitatively similar.

All six species had significantly reduced class="Chemical">N coclass="Chemical">nceclass="Chemical">ntratioclass="Chemical">ns iclass="Chemical">n the Tree-Removal treatmeclass="Chemical">nt compared with Seedliclass="Chemical">ng-Removal. Poa aclass="Chemical">nd Carex also had lower class="Chemical">n class="Chemical">N concentrations in the No-Removal treatment, and Poa and Kunzea had lower N concentrations in the Sapling-Removal treatment relative to the Seedling-Removal. Vector analysis diagrams (Fig. 5) suggest that the lower N concentrations are due to biomass increases being substantially greater than total N increases, resulting in decreased concentrations of N despite increasing total N.

Figure 5.

Vector analysis of foliar percent N and P as a function of seedling mass of each species in the ex situ bioassay showing means and standard errors for each treatment. Grey isoclines indicate constant nutrient content, and arrows indicate change relative to the Seedling-Removal treatment. Declines in nutrient concentration occur despite increased total nutrient content due to the disproportionate increase in plant mass.

class="Chemical">Phosphorous coclass="Chemical">nceclass="Chemical">ntratioclass="Chemical">ns of seedliclass="Chemical">ngs were reduced iclass="Chemical">n Piclass="Chemical">nus, Carex aclass="Chemical">nd Poa iclass="Chemical">n Sapliclass="Chemical">ng-Removal, aclass="Chemical">nd also iclass="Chemical">n Carex iclass="Chemical">n class="Chemical">n class="Chemical">No-Removal and Tree-Removal soils relative to Seedling-Removal. Only Kunzea in the Tree-Removal treatment had significantly increased P concentration. Vector analysis diagrams (Fig. 5) showed that the significant decreases in P shown in Fig. 4 were mostly due to increased biomass rather than decreased total P.

Vector analysis of foliar percent n class="Chemical">N aclass="Chemical">nd P as a fuclass="Chemical">nctioclass="Chemical">n of seedliclass="Chemical">ng mass of each species iclass="Chemical">n the ex situ bioassay showiclass="Chemical">ng meaclass="Chemical">ns aclass="Chemical">nd staclass="Chemical">ndard errors for each treatmeclass="Chemical">nt. Grey isocliclass="Chemical">nes iclass="Chemical">ndicate coclass="Chemical">nstaclass="Chemical">nt class="Chemical">nutrieclass="Chemical">nt coclass="Chemical">nteclass="Chemical">nt, aclass="Chemical">nd arrows iclass="Chemical">ndicate chaclass="Chemical">nge relative to the Seedliclass="Chemical">ng-Removal treatmeclass="Chemical">nt. Decliclass="Chemical">nes iclass="Chemical">n class="Chemical">nutrieclass="Chemical">nt coclass="Chemical">nceclass="Chemical">ntratioclass="Chemical">n occur despite iclass="Chemical">ncreased total class="Chemical">nutrieclass="Chemical">nt coclass="Chemical">nteclass="Chemical">nt due to the disproportioclass="Chemical">nate iclass="Chemical">ncrease iclass="Chemical">n placlass="Chemical">nt mass.

Discussion

Plant–soil feedbacks are widely studied, yet relatively few studies have demonstrated both the mechanistic basis of these feedbacks and their consequences for aboveground communities (Bever ; Suding ). Our study of tree invasions into native grassland demonstrates two important plant–soil feedback mechanisms at work: increased ectomycorrhizal inoculum and altered soil biogeochemical class="Chemical">cycliclass="Chemical">ng. Altered soil properties subsequeclass="Chemical">ntly caused species-specific legacies for placlass="Chemical">nt commuclass="Chemical">nities, resulticlass="Chemical">ng iclass="Chemical">n a major shift iclass="Chemical">n domiclass="Chemical">naclass="Chemical">nce amoclass="Chemical">ng differeclass="Chemical">nt placlass="Chemical">nt growth forms. The alteratioclass="Chemical">n of soil biogeochemical class="Chemical">n class="Chemical">cycling is particularly interesting because most prior studies have focussed on the effects of invasive N-fixing plants in N-limited systems (e.g. Vitousek ; Heneghan ; Liao ; Nuñez and Dickie 2014), whereas Pinus is not N fixing. As one of the most invasive trees globally (Richardson and Rejmánek 2004; Procheş ), understanding Pinus effects is inherently important, but we suggest that our results can also be generalized to other ectomycorrhizal invasive trees (Nuñez and Dickie 2014).

We expected that the establishment of invasive Pinus would cause major shifts in soil nutrient class="Chemical">cycliclass="Chemical">ng, with iclass="Chemical">ncreased P availability but also iclass="Chemical">ncreased fuclass="Chemical">ngal domiclass="Chemical">naclass="Chemical">nce aclass="Chemical">nd class="Chemical">n class="Chemical">N limitation relative to the predominantly native plant community present when Pinus was excluded (i.e. Seedling-Removal treatment). Invasive Pinus increased P availability, consistent with our expectation, but there were also unexpected large increases in labile N (NO3-N) following pine removal. Although based on a single point-in-time measurement, the increase in available P and NO3-N was reflected in increased seedling total N and total P (Fig. 5) and hence indicates an increase in total flux, not just pool size. Soil microbial trophic channels did not shift towards increased fungal dominance, but rather to increased dominance by bacteria, bacterial-feeding nematodes and enchytraeids. Unlike N-fixing invaders, invasive Pinus did not substantially add resources. Rather, Pinus appears to have switched soil function from a slow-cycling, more nutrient-conservative system to a fast-cycling, potentially more ‘leaky’ system (Hobbie 1992), with the same quantity of nutrients cycling more rapidly. This fundamental biogeochemical shift also contributes to the strong belowground legacies generated during invasion by Pinus, with subsequent effects on performance of co-occurring plant species. In particular, we observed large increases in the growth of graminoids (an arbuscular mycorrhizal grass and a non-mycotrophic sedge Carex) in soils from which Pinus had been removed. This belowground legacy observed in the greenhouse is consistent with field observations of increased dominance by non-native grasses and non-native herbaceous plants, species that are likely adapted to high-nutrient turnover, in plots where Pinus had been removed. More generally, these findings support the view that biogeochemical disruptions can facilitate plant invasions (e.g. Davis ).

We also found ectomycorrhizal facilitation by invasive Pinus, but this effect differed among plant species. The establishment of closed-canopy class="Species">P. contorta forest, whether iclass="Chemical">ntact (class="Chemical">n class="Chemical">No-Removal) or removed (Tree-Removal), facilitated ectomycorrhizal infection of the con-familial invasive tree species P. menziesii, but not conspecific Pinus or the native shrub K. ericoides. This suggests a selective biotic feedback, whereby early plant invaders differentially facilitate the establishment or subsequent performance of later plant species through their effects on soil populations of symbionts. Shared symbionts may be particularly the case for con-familial trees that originate from the same geographical region, such as P. contorta and P. menziesii, which are known to naturally share fungal associates (Molina and Trappe 1992). Pseudotsuga was not a major component of current plant communities at the site (comprising <1 % cover) but is of particular conservation concern as the most shade tolerant of invasive conifers (Froude 2011).

Effects of Pinus on soils and soil communities

The common expectation is that forest ecosystems, and particularly ectomycorrhizal forests, will have fungal-dominated, slow nutrient class="Chemical">cycliclass="Chemical">ng soils (Imberger aclass="Chemical">nd Chiu 2001; Harris 2009). Our observatioclass="Chemical">n that iclass="Chemical">nvasive Piclass="Chemical">nus iclass="Chemical">ncreased soil class="Chemical">nutrieclass="Chemical">nt availability aclass="Chemical">nd the importaclass="Chemical">nce of the bacterial eclass="Chemical">nergy chaclass="Chemical">nclass="Chemical">nel ruclass="Chemical">n couclass="Chemical">nter to this expectatioclass="Chemical">n, although it is coclass="Chemical">nsisteclass="Chemical">nt with earlier observatioclass="Chemical">ns of ectomycorrhizal class="Chemical">n class="Species">P. nigra and K. ericoides successions (Dickie ). The concept of forests as fungal dominated and grasslands as bacterial dominated is based primarily on comparisons of more mature forests with post-agricultural grasslands, but our results suggest that this may not apply more generally to early stages of woody succession or invasion into grass and shrub-dominated communities. Our results also suggest that Pinus invasion has increased energy flows to soil and a shift towards bacterial dominance, with concomitant increases in nutrient cycling rates (Hobbie 1992).

Increased bacterial dominance was expressed most strongly in increased bacterial predator densities rather than bacterial class="Chemical">PLFA levels, reflecticlass="Chemical">ng the top-dowclass="Chemical">n regulatioclass="Chemical">n of the bacterial eclass="Chemical">nergy chaclass="Chemical">nclass="Chemical">nel (Wardle 2002). Shifts towards domiclass="Chemical">naclass="Chemical">nce of the bacterial eclass="Chemical">nergy chaclass="Chemical">nclass="Chemical">nel iclass="Chemical">n soil food webs have beeclass="Chemical">n frequeclass="Chemical">ntly associated with iclass="Chemical">ncreased ecosystem productivity, e.g. iclass="Chemical">n the preseclass="Chemical">nce of class="Chemical">n class="Chemical">N-fixing plants (St John ; de Vries ). An increase in ecosystem productivity also commonly occurs in plant invasions, particularly invasions by woody plants (Liao ). More rapidly cycling systems usually result in higher soil nutrient availability, consistent with our observation of large increases in NO3-N and available P. The dominance of bacterial-feeding nematodes and enchytraeids was particularly high in the Tree-Removal treatment. This may reflect a response to the large pulse of C and nutrients released after felling of pines 3 years prior to this study. Nonetheless, not all of the observed effects can be attributed to a post-removal nutrient flush as the nematode fungal channel dominance reached its nadir under intact pine (No-Removal). There was also no observed increase in soil C in the Tree-Removal treatment relative to other treatments, suggesting that any possible flush of organic C following removal was no longer present in the soil after 3 years. Rapid mineralization of C is consistent with the observed bacterial trophic dominance.

Increased availability of P has been observed in both planted and invasive Pinus (Chen ; Dickie ), as has a release of soil C (Chapela ; Dickie ). These changes may reflect the unique enzymatic capabilities of some ectomycorrhizal fungi, particularly the ability to acquire class="Chemical">N aclass="Chemical">nd P from orgaclass="Chemical">nic sources while obtaiclass="Chemical">niclass="Chemical">ng C from their placlass="Chemical">nt symbioclass="Chemical">nts (Dickie ). However, the domiclass="Chemical">naclass="Chemical">nce of class="Chemical">native shrubs associated with ericoid mycorrhizal fuclass="Chemical">ngi iclass="Chemical">n this system prior to iclass="Chemical">nvasioclass="Chemical">n suggests that the ability to utilize orgaclass="Chemical">nic class="Chemical">nutrieclass="Chemical">nt sources is class="Chemical">not a class="Chemical">novel ecosystem fuclass="Chemical">nctioclass="Chemical">n (Read aclass="Chemical">nd Perez-Moreclass="Chemical">no 2003). Rather, the relatively high growth rates of iclass="Chemical">nvasive Piclass="Chemical">nus may be associated with greater C allocatioclass="Chemical">n belowgrouclass="Chemical">nd, which may iclass="Chemical">ncrease the rate of orgaclass="Chemical">nic class="Chemical">nutrieclass="Chemical">nt miclass="Chemical">neralizatioclass="Chemical">n by microbes, but more detailed iclass="Chemical">nvestigatioclass="Chemical">n is required. Our results, iclass="Chemical">n combiclass="Chemical">natioclass="Chemical">n, democlass="Chemical">nstrate that iclass="Chemical">nvasioclass="Chemical">n by Piclass="Chemical">nus causes biogeochemical disruptioclass="Chemical">n through chaclass="Chemical">nges iclass="Chemical">n macroclass="Chemical">nutrieclass="Chemical">nts, as well as shifts iclass="Chemical">n the soil class="Chemical">n class="Species">biota controlling biogeochemical processes. The long-term persistence of changes in soil nutrient cycling remains uncertain, particularly given that pines increased the rate of nutrient cycling but did not increase nutrient stocks. Indeed, the initial state of the ecosystem is somewhat perplexing in having quite high levels of total N and P (C : N : P ratio 83 : 4.6 : 1) and only moderately acidic soils (pH 5.2), yet being dominated by relatively slow-growing Ericaceous shrubs. This may suggest that once the system has been disrupted into a high-nutrient cycling state, it may not quickly recover.

Ectomycorrhizal facilitation

The establishment of closed-canopy invasive Pinus facilitated the ectomycorrhizal infection of class="Species">Pseudotsuga, but class="Chemical">not of Piclass="Chemical">nus or Kuclass="Chemical">nzea. Despite haviclass="Chemical">ng a broad receptivity to differeclass="Chemical">nt species of ectomycorrhizal fuclass="Chemical">ngal symbioclass="Chemical">nts, class="Chemical">n class="Species">Pseudotsuga seedlings can be relatively slow to establish ectomycorrhizas in early succession (Borchers and Perry 1990; Davis ; Horton ; Ledgard 2004; Kazantseva ). In its native range, Pseudotsuga appears to benefit from prior establishment of pioneering ectomycorrhizal plants (Borchers and Perry 1990; Horton ). Our observation of increased mycorrhizal infection of Pseudotsuga in soils where closed-canopy Pinus had been present suggests that as an invasion process, the prior co-invasion of Pinus and its mycorrhizal fungi (Dickie ) can result in greater ecosystem vulnerability to Pseudotsuga invasion. This is an important and potentially widespread example of sequential invasional meltdown, whereby one invasive species greatly increases the subsequent success or performance of other invaders, in this case through biotic interactions with mutualists (Simberloff and Von Holle 1999; Nuñez ).

We did not observe strong effects of invasive Pinus on the establishment of ectomycorrhizal mutualists on either Pinus or Kunzea seedlings, with most Pinus and no Kunzea seedlings having ectomycorrhizas. The differences in mycorrhizal infection between Pinus, n class="Species">Pseudotsuga aclass="Chemical">nd Kuclass="Chemical">nzea may reflect fuclass="Chemical">ngal host specificity. Iclass="Chemical">noculum of the ectomycorrhizal fuclass="Chemical">ngus Rhizopogoclass="Chemical">n is likely widespread at this site from adjaceclass="Chemical">nt established Piclass="Chemical">nus placlass="Chemical">ntatioclass="Chemical">ns aclass="Chemical">nd is uclass="Chemical">nusual amoclass="Chemical">ng ectomycorrhizal fuclass="Chemical">ngi iclass="Chemical">n haviclass="Chemical">ng spore loclass="Chemical">ngevity of at least several years iclass="Chemical">n soils (Bruclass="Chemical">ns ). The preseclass="Chemical">nce of a widespread spore baclass="Chemical">nk may explaiclass="Chemical">n the widespread iclass="Chemical">nfectioclass="Chemical">n of Piclass="Chemical">nus, but as Rhizopogoclass="Chemical">n has very high levels of host specificity (Moliclass="Chemical">na aclass="Chemical">nd Trappe 1994) it may oclass="Chemical">nly beclass="Chemical">nefit Piclass="Chemical">nus. Kuclass="Chemical">nzea, iclass="Chemical">n coclass="Chemical">ntrast, was class="Chemical">never ectomycorrhizal iclass="Chemical">n our greeclass="Chemical">nhouse bioassay despite its kclass="Chemical">nowclass="Chemical">n capability to form ectomycorrhizas (Orlovich aclass="Chemical">nd Cairclass="Chemical">ney 2004). This may reflect aclass="Chemical">n iclass="Chemical">ncompatibility with ectomycorrhizal fuclass="Chemical">ngi co-iclass="Chemical">nvadiclass="Chemical">ng with piclass="Chemical">ne. Oclass="Chemical">n the other haclass="Chemical">nd, Kuclass="Chemical">nzea has the ability to form arbuscular mycorrhizas as well as ectomycorrhizas (Moyersoeclass="Chemical">n aclass="Chemical">nd Beever 2004), aclass="Chemical">nd it may be that greeclass="Chemical">nhouse coclass="Chemical">nditioclass="Chemical">ns favoured this alterclass="Chemical">native mycorrhizal type as other greeclass="Chemical">nhouse studies have also failed to establish ectomycorrhizas oclass="Chemical">n Kuclass="Chemical">nzea (Davis ).

Belowground legacy effects on plant growth

The changes in plant community composition observed following the removal of Pinus represent a major functional shift from shrub-dominance to dominance by exotic herbs and grasses. Grasses were present in the Seedling-Removal treatment, but even within this broad functional category, class="Species">A. capillaris is disticlass="Chemical">nct as a rhizomatous grass while the domiclass="Chemical">naclass="Chemical">nt class="Chemical">native grasses of the area are tussock formiclass="Chemical">ng. Oclass="Chemical">nce established, rhizomatous grasses have the poteclass="Chemical">ntial to form declass="Chemical">nse swards that are more resistaclass="Chemical">nt to subsequeclass="Chemical">nt establishmeclass="Chemical">nt of woody placlass="Chemical">nts thaclass="Chemical">n tussock-formiclass="Chemical">ng grasses (D'Aclass="Chemical">ntoclass="Chemical">nio aclass="Chemical">nd Vitousek 1992). class="Chemical">n class="Species">Hieracium pilosella, the main invasive herb, also forms dense mats. A study of the invasive con-generic H. lepidulum in adjacent ecosystems suggested that the addition of Hieracium to the established native vegetation did not result in a loss of plant diversity (Meffin ), but other studies have suggested a strong ability of H. pilosella to suppress native plant growth, particularly under high fertility and high light conditions (Moen and Meurk 2001). In part these discrepancies may reflect priority effects: an established native community may be resilient to Hieracium impacts, but an established Hieracium mat may be resistant to native re-establishment following disturbance.

The observed rapid increase in the abundance of non-native grasses and herbs following Pinus removal in the field could have been driven by multiple factors, including seed availability and an ability to respond to disturbance. class="Chemical">Noclass="Chemical">netheless, the large iclass="Chemical">ncrease iclass="Chemical">n class="Chemical">n class="Chemical">graminoid growth in the greenhouse bioassay suggests that belowground biogeochemical legacies of Pinus are an important contributing factor. Graminoids and herbs tend to have a much greater plasticity in response to increased nutrient availability compared with woody plants, and this may be particularly the case for invasive grasses and herbs (Funk 2008). The increased availability of N and P following Pinus invasion may therefore be particularly beneficial to these functional groups.

Removal versus species effect

Some of the treatment effects we observed were associated with both the class="Chemical">No-Removal aclass="Chemical">nd Tree-Removal treatmeclass="Chemical">nts, while others were observed oclass="Chemical">nly followiclass="Chemical">ng Tree-Removal. For example, available P aclass="Chemical">nd bacterial domiclass="Chemical">naclass="Chemical">nce of class="Chemical">nematode-feediclass="Chemical">ng ratios were sigclass="Chemical">nificaclass="Chemical">ntly higher iclass="Chemical">n both the class="Chemical">n class="Chemical">No-Removal and Tree-Removal treatments compared with the Seedling-Removal treatment. This suggests that these effects can be attributed to effects of the shift in plant community from native dominance to Pinus rather than Pinus removal activity or residual biomass left after removal. In contrast, the increase in available N and bacterial dominance measured by PLFA were seen only in the removal treatment. Effects seen only in the removal treatment may represent non-species-specific removal effects (nutrient pulses, disturbance) rather than Pinus-specific species effects. The increased bacterial dominance between No-Removal and Tree-Removal, which is driven by a reduction in fungi, may reflect a loss of Pinus-associated ectomycorrhizal fungal biomass following tree removal.

Treatment effects on plant communities were complex. Increased class="Chemical">graminoid biomass was oclass="Chemical">nly statistically sigclass="Chemical">nificaclass="Chemical">nt iclass="Chemical">n the Tree-Removal treatmeclass="Chemical">nt. class="Chemical">n class="Chemical">Nonetheless, Poa spp. in particular had an increase in biomass almost as large under the intact Pinus No-Removal treatment as under the Tree-Removal treatment (Figs 4 and 5). Our interpretation is that increased P availability, bacterial dominance of nematode trophic channels and graminoid growth are likely driven by direct effects of Pinus, but that these effects may be further augmented by a large influx of leaf litter, fine roots and woody biomass following tree removal. This would be consistent with the Pinus uptake of both total P and recalcitrant P sources, incorporation into foliage and recycling through litter fall (all presence effects; Chen ; Dickie ). Tree removal may further augment these effects by creating a sudden influx of litter and fine roots. In contrast, increased N availability and decreased fungal biomass as measured by PLFA appear likely to be a removal effect, as neither was observed in the other treatments. Soil pH appeared to be the only Pinus effect that was reversible within three years of removal. Taken as a whole, the results suggest that both direct effects of Pinus and removal effects are important belowground legacies. Where effects can be attributed specifically to removal (e.g. increased NO3-N), there may be potential to mitigate effects through different removal strategies.

Conclusions

The regime shift driven by Pinus invasion has the potential to represent a novel stable state (Folke ). Periodic removal of trees does not result in restoration of native plants, at least within 3 years, but rather dominance of non-native grasses and herbaceous plants, likely driven by increased nutrient availability. Even re-measuring the site in 2014, 9 years following Pinus removal, we found no native regeneration (data not shown). This community is, in turn, likely to be re-invaded by Pinus, given increased ectomycorrhizal inoculum of most fungal species will likely persist at least as long as Pinaceae (Pinus or class="Species">Pseudotsuga) are preseclass="Chemical">nt. Eveclass="Chemical">n followiclass="Chemical">ng removal of these trees, spores of Rhizopogoclass="Chemical">n aclass="Chemical">nd some other early-successioclass="Chemical">nal fuclass="Chemical">ngi caclass="Chemical">n persist for at least several aclass="Chemical">nd poteclass="Chemical">ntially maclass="Chemical">ny years (Bruclass="Chemical">ns ; class="Chemical">n class="Chemical">Nguyen ). At present there are no effective management options to remove invasive fungi once present.

From an applied perspective, removal of Pinus at the seedling or early sapling stage appears to be effective in preventing many of the negative effects of invasion and should be a management priority. Where Pinus trees dominate the site it may not be practical to restore and maintain the native tussock grassland and shrubland present prior to Pinus invasion. However, it is worth noting that the site was originally a forested ecosystem prior to class="Species">human iclass="Chemical">nduced fire starticlass="Chemical">ng with Māoris after the 13th ceclass="Chemical">ntury aclass="Chemical">nd iclass="Chemical">ncreasiclass="Chemical">ng with Europeaclass="Chemical">ns iclass="Chemical">n the 19th ceclass="Chemical">ntury. If the class="Chemical">native forest could be restored, poteclass="Chemical">ntially by uclass="Chemical">nder-placlass="Chemical">nticlass="Chemical">ng class="Chemical">native Kuclass="Chemical">nzea or class="Chemical">n class="Species">Nothofagus trees before complete Pinus tree removal, it would likely be resistant to both exotic grass and Pinus invasion. Our results suggest that any such efforts should not expect ectomycorrhizal Pinus to facilitate ectomycorrhizal infection of native trees, hence inoculation is likely to be necessary (Dickie ).

Our findings highlight that belowground legacies can be generated across multiple trophic levels, involving plants, fungi, nematodes and bacteria. In addition, these legacies represent potentially profound biogeochemical disruption with consequent effects on the success or performance of co-occurring native and non-native species. Understanding the persistence and importance of belowground legacies of biological invaders is needed to predict the assembly and function of these increasingly widespread novel ecosystems.

Sources of Funding

This research was supported by Core funding for Crown Research Institutes from the n class="Chemical">New Zealaclass="Chemical">nd Miclass="Chemical">nistry of Busiclass="Chemical">ness, Iclass="Chemical">nclass="Chemical">novatioclass="Chemical">n aclass="Chemical">nd Employmeclass="Chemical">nt's Scieclass="Chemical">nce aclass="Chemical">nd Iclass="Chemical">nclass="Chemical">novatioclass="Chemical">n Group, with additioclass="Chemical">nal support of class="Chemical">n class="Disease">IAD by the Bio-Protection Research Centre of Lincoln University.

Contributions by the Authors

I.A.D., M.G.S.J. and D.A.P. conceived the study; all authors contributed to experimental design: I.A.D., M.G.S.J., K.O., G.W.Y. and K.I.B. gathered soil data; K.I.B. and I.A.D. designed and executed the greenhouse component; D.A.P. and C.W.M. gathered most vegetation data; I.A.D. analysed the data and wrote the paper, with contributions from M.G.S.J., K.O. and D.A.P.; all authors except G.W.Y. helped to revise and improve the text.

Conflicts of Interest Statement

n class="Chemical">Noclass="Chemical">ne declared.

Supporting Information

The following Supporting Information is available in the online version of this article–

Table S1. Plant community cover (summed across height tiers) in each of the four treatments: Seedling-Removal, Sapling-Removal, class="Chemical">No-Removal aclass="Chemical">nd Tree-Removal of class="Chemical">n class="Species">P. contorta. A unique six-letter species code (used in Fig. 3), with species, family, growth form and native status in the alphabetical order by species code (National Vegetation Survey: nvs.landcareresearch.co.nz).