Introduction

Hyperdiverse plant communities thrive in south-western Australia’s ancient landscapes, which rank among the most nutrient-impoverished habitats worldwide (McArthur 2004; Sauquet et al. 2009; Cook et al. 2015). In highly weathered soils, the majority of phosphorus (P) is present in occluded inorganic or in organic forms (Turner et al. 2018). Plants adapted to nutrient scarcity have evolved a variety of specialized root structures, which can mobilise sparingly available P and micronutrients from the surrounding soil (Lamont 2003; Shane and Lambers 2005). Species of the genus Banksia in south-western Australia form compound cluster roots during the wet winter season; cluster roots release carboxylates, protons, phenolic compounds and acid phosphatases into the surrounding soil (Lamont 1982; Lambers et al. 2006). These root structures are confined to the relative P-rich uppermost soil layers (Pate and Watt 2002) and the spatiotemporally concentrated secretion of these root exudates results in the mobilisation of P, manganese (Mn), iron and zinc from sorbed inorganic and organic nutrient pools in the soil (Gardner and Boundy 1983; Grierson and Attiwill 1989; Dinkelaker et al. 1989; Teste et al. 2020).

Several plant species growing on nutrient-poor soils lack specialised root structures capable of acquiring nutrients from poorly-available resource pools (i.e. “nutrient mining”). Nutrient facilitation is considered a plant nutrient-acquisition strategy on highly-weathered soils (Li et al. 2007; Lambers et al. 2022; Shen et al. 2023). Indeed, experiments under controlled conditions indicate that nutrient facilitation may occur among native Australian species that share a long co-evolutionary history (Muler et al. 2014). The non-cluster root forming shrub Hibbertia racemosa preferentially allocates root biomass towards the cluster-root forming Banksia attenuata, and shows decreased root biomass allocation to the neighbour when grown next to another conspecific (de Britto Costa et al. 2021). Furthermore, imaging of root dynamics in a Banksia woodland using minirhizotrons indicated that the likelihood of observing a root of a non-cluster root forming neighbour in the vicinity to a cluster root was substantially greater than observing a non-cluster root next to a non-cluster root (Teste et al. 2020). Taken together, these findings suggest that nutrients mobilised by cluster-root activity of Banksia spp. can be shared with non-cluster-root forming neighbours and that interspecific root intermingling with active cluster roots takes place in nutrient-poor soils. However, whether nutrient facilitation occurs in natural habitats still remains to be determined.

In contrast to iron, zinc and copper, plant uptake of Mn is poorly regulated and can therefore result in large variations in leaf Mn concentrations ([Mn]) within a single species (Rengel 2000; Lambers et al. 2021). Roots take up Mn as the divalent cation Mn2+. The availability of Mn2+ in soils is determined by pH and redox potential of the soil solution. Small decreases in pH can lead to substantial increases in plant-available Mn, especially in moderately acidic soils (Sanders 1983) and wetter, reducing conditions, which can be found in seasonally water-logged soils, can also induce increases in Mn availability (Kirk et al. 2014; Lambers et al. 2021). Pioneer studies in barley and sugarcane grown in solution culture indicate that plant uptake rates of Mn are linearly related to external [Mn] over a wide concentration range of 1–100 µM Mn2+ (Maas et al. 1968; Bowen 1981). Banksia woodlands are found on highly weathered acidic sandy soils with total [Mn] below 3 mg kg−1 (Ritchie et al. 2021) and extremely low exchangeable [Mn] (0.002 mg kg−1, values reported for Bassendean sand (Turner and Laliberté 2015; Birnbaum et al. 2018)). As the sandy soils of Banksia woodlands are well aerated, reducing conditions due to water-logging are unlikely to occur, even in areas lower in the landscape (Lambers et al. 2021). However, the performance of Banksia trees could depend on the access to water from deep sand layers and might therefore be influenced by the elevation at which the trees are growing (Groom 2004; Gao et al. 2020).

In addition to its role as an activator of diverse enzymes (reviewed by Hänsch and Mendel 2009), Mn is involved in photosynthesis in a component of the water-splitting complex of the PSII reaction centre (Schmidt et al. 2016). After root uptake, Mn2+ is transported through the xylem to transpiring leaves and shows little remobilisation during leaf senescence due to low phloem mobility (Loneragan 1988). Low remobilisation rates have been shown for temperate woody perennials, crop plants such as Zea mays, Brassica napus and several native Australian species, including Banksia trees and non-cluster root forming neighbouring shrubs (Hayes et al. 2014; Maillard et al. 2015). Given the low soil [Mn] in Banksia woodlands, the linear relationship between root uptake rate and external [Mn] (until approximately 100 µM Mn) and the low remobilisation of leaf Mn during leaf senescence, the variation of leaf [Mn] within one plant community can be used as an indicator of rhizosphere modifications such as enhanced exudation of carboxylates (Lambers et al. 2021; Zhou et al. 2022).

Here, our specific aim was to test whether Banksia size and proximity to non-cluster rooted woody shrubs influenced leaf [Mn] of non-cluster rooted neighbours. We hypothesised that neighbouring shrubs are facilitated in terms of Mn uptake depending on distance to surrounding cluster-root-forming Banksia trees which results in higher leaf [Mn] in shrubs with more and larger nearby Banksia trees. For that purpose, we mapped all Banksia attenuata and B. menziesii trees and selected neighbouring shrubs within a study site located on top of a sand dune in the Swan Coastal Plain in Western Australia (ca. 7800 m2). Bulk soil samples (0–7 cm depth) were collected and analysed for physical properties and nutrient concentrations to test for major heterogeneities within the study site. To assesses the effect of Banksia tree size and proximity to non-cluster-rooted woody shrubs and shrub elevation on leaf [Mn], samples of similarly-aged leaves were taken, digested and analysed using ICP-OES.

Materials and methods

Study area and study site

The study site was based within Alison Baird Reserve, which is located approximately 15 km southeast of Perth CBD, Western Australia (Baird 1984; Tauss 2019). The reserve is situated on the Swan Coastal Plain, which is a geographic feature composed of marine dune deposits and the Swan River. The reserve comprises an up to 2-million years old dune of the Bassendean sand system (Turner et al. 2018) of about 4 m height whose dune crest is vegetated by Banksia woodland dominated by Banksia attenuata and B. menziesii accompanied by a species-rich community mainly composed of woody shrubs, several perennial herbs and some winter-green herbs (Speck and Baird 1984; Tauss 2019; Smith et al. 2023; for a review on contemporary Banksia woodlands see Ritchie et al. (2021)). A site was chosen on the upper part of the dune (7877 m2), where B. attenuata and B. menziesii are the dominating cluster-root bearing species in terms of biomass and where gradients of high and low Banksia density and biomass (here termed “Banksia effect”) were present (Fig. 1). The study site presented a strongly weathered podsol with acidic humus and a sesquioxide layer at around 5 m. In the topsoil, Colwell-P concentrations are low, typically below 1 mg kg−1 (Gao et al. 2020).

Fig. 1
figure 1

Overview of the Alison Baird Reserve study site. Geographical location of the reserve on the Swan Coastal Plain, Western Australia and boundaries of the study site (7877 m2) (A). The plant community found on top of a dune is an open sclerophyll Banksia woodland with Banksia attenuata and B. menziesii as the dominant species. Neighbouring understorey woody shrubs within the Banksia woodland were chosen randomly and leaves were sampled: Bossiaea eriocarpa (n = 61), Hibbertia hypericoides (n = 32). Differences in stem diameter of Banksia individuals are indicated by different symbol sizes (B). Elevation of the study site(C)

Full size image

Location mapping, plant species and biomass estimation

Plot boundaries and the positions of Banksia trees, neighbouring woody shrubs and soil sampling sites were surveyed using a Trimble R10 GNSS survey system, with base processed to Geoscience Australia AUSPOS (horizontal accuracy 16 ± 4 mm, vertical accuracy 35 ± 6 mm, mean ± SD). Each Banksia tree and sapling located within the study site was mapped and individuals of Bossiaea eriocarpa Benth. (Fabaceae) and Hibbertia hypericoides Benth. (Dilleniaceae) were chosen randomly. Both shrub species showed a large variation of leaf [Mn] in a preliminary screening of various non-cluster rooted species (Wasaki et al., unpublished). Regeneration after fire is mainly achieved through resprouting in all species assessed in this study (Pate and Bell 1999) For plant biomass estimation, Banksia tree and sapling stem diameter were measured 1 cm above the lignotuber. Canopy area of B. eriocarpa and H. hypericoides shrubs were estimated (as a proxy for shrub biomass) using the standard equation for the area of an ellipse A = π*a*b/4, where a and b represent the maximum canopy diameter and the canopy diameter perpendicular to the maximum. Leaf litter cover and depth under Banksia canopies was scored at a scale from 0–10. On a plant biomass basis, B. attenuata and B. menziesii were the dominant species within the study site. Both species form cluster roots which are continuously produced when soil is moist, from May to October, a period during which approximately 80% of the mean annual precipitation falls (Veneklaas and Poot 2003; Lamont 2003; https://weather.agric.wa.gov.au/station/SP). Banksia cluster roots have an average lifespan of three weeks (Teste et al. 2018).

Leaf sampling and elemental composition

Leaf material was collected in late October 2019, during the onset of the dry summer season. For each individual shrub, we collected mature undamaged leaves from the cohort of the previous year which could be visually differentiated based on the differences in lignification of the stem. The leaf material was then oven-dried at 70 °C for 48 h, and ground to a fine powder using a mill and zirconium beads. The ground material was digested using ultra-pure acids (HNO3:HClO4, 3:1; v/v) and subsequently analysed for P and Mn via inductively coupled plasma optical-emission spectroscopy using Yttrium as an internal standard (Optima 5300DV ICP-OES; PerkinElmer Inc., Waltham, MA, USA).

Soil sampling, chemical analyses, and texture characterisation

To derive a map of the total soil elemental concentrations, pH and soil particle size distribution within the study site, we collected topsoil samples from 20 locations in late summer, i.e. end of January 2020. Samples were collected from the topsoil after removal of the litter layer, using a bulk density cylinder (7.2 cm diameter, 7.2 cm height, ca. 293 cm3 inner volume) dried to constant weight at 40 °C and sieved (< 2 mm). Before further analyses, roots with a diameter above 1 mm were removed.

Electrical conductivity (EC) was determined in DI-water suspension (1:5, w/v). Soil pH was determined in DI-water suspension and CaCl2 solution (1:5, w/v, 0.01 M CaCl2). Total N and C was determined by Dumas dry combustion using an Elementar Vario Macro CNS (Hanau, Germany). Elements in soil were extracted with aqua regia digestion (3:1 mixture of HNO3: HCl, v/v, 1 h at 130 °C) and determined by inductively coupled plasma optical emission spectrometry (ICP-OES; 5300DV spectrometer, Perkin Elmer, Shelton, CT, USA) (Rayment and Lyons 2011). Organic matter was removed from an aliquot of the soil samples prior to aqua regia digestion following the method described by (Kunze and Dixon 1986) using hydrogen peroxide according to standard protocols. For the determination of plant available P, soil samples were extracted for 4 h in 0.5 M NaHCO3, at pH 8.5 (1:10, w/v, soil solution ratio), phosphate was subsequently determined by the molybdate blue method (Murphy and Riley 1962). The inorganic P in the extracts was analysed by the Dick and Tabatabai (1977) method, the absorbance was recorded at 850 nm (He and Honeycutt 2005).

Particle size distribution of the soil samples was measured by laser diffraction using a Mastersizer 2000 with 200 G wet dispersion accessory (Malvern Panalytical, Malvern, UK). Fractions were split according six fraction sizes from clay to coarse sand at < 2, 63, 200, 630, 2000 μm. No particles with sizes < 2 µm were recorded in the samples analysed.

Statistical analyses and modelling

Using nearest neighbour interpolation, we obtained estimates for all soil variables for the locations of all shrubs studied. We then tested for correlations between interpolated soil variables and shrub leaf [Mn] using linear regressions.

To test for factors affecting leaf [Mn], we used multiple linear models with leaf [Mn] of the shrub as the dependent variable. In all models we included shrub size (i.e. canopy area) and elevation of the shrub as possible explanatory variables, but we also included different explanatory variables that represented the effect of nearby Banksia trees with varying levels of detail. We tried three different modelling approaches for each shrub species. First, we included distance to nearest Banksia and size of nearest Banksia as possible independent/explanatory variables, with and without transformations of dependent variable. Shrubs less than 5 m from the boundary of our study area were excluded to avoid edge effects, as they may have had nearer Banksia trees outside the study area. Second, we explored models with a binary True/False variable representing whether there was any Banksia within a certain threshold distance of the shrub. Third, we explored including variables that represented ‘short-range Banksia effect’ and ‘long-range Banksia effect’. For a given shrub, the short-range Banksia effect was defined as the sum of the radii of any Banksia trees or saplings within a certain fixed threshold distance independent of the size of the Banksia. The long-range Banksia effect was defined as the sum of the stem diameters of any Banksia trees or saplings within a threshold distance dependent on the size of the Banksia; specifically, the threshold distance for this effect was assumed to be a linear function of the stem diameter of the Banksia individual. For the second and third approach, we explored different threshold distances and chose distances that optimised the model fit. Shrubs close to the boundary of the study area (close enough that they may have been affected by unrecorded Banksia trees beyond the boundaries) were excluded to avoid edge effects. Other Banksia effects were also tested, such as an effect declining with increasing distance, but the threshold models explained the data considerably better. The linear models were simplified using stepwise simplification based on AIC (Akaike Information Criteria), to avoid overfitting. All statistical analyses were carried out in R; the analysis code used for modelling is available in the supplements.

Results

Alison Baird Reserve study site

We mapped 217 B. attenuata and 298 B. menziesii trees or saplings using a GNSS survey system with high vertical and horizontal accuracy (Fig. 1). Stem diameters of B. attenuata and B. menziesii ranged from 5 to 380 mm and 3 to 330 mm within the site (mean ± SE was 75 ± 5 mm and 69 ± 2 mm, respectively). The coordinates of 61 B. eriocarpa individuals were recorded and leaves were sampled. A lower number of H. hypericoides shrubs were present at the study site. In total, 32 samples were taken of the latter species.

Leaf [Mn] and shrub canopy area

Within the study site, the leaf [Mn] in one-year old leaves ranged from approximately 27 to 474 mg kg−1 in B. eriocarpa and from 43 to 287 mg kg−1 in H. hypericoides (Fig. S1A). We found no significant correlations between leaf [Mn] and leaf [P], neither in B. eriocarpa nor in H. hypericoides (data not shown).

The canopy area of randomly chosen B. eriocarpa individuals ranged from 71 to 5140 cm2. We observed a similar range of canopy areas in the 32 H. hypericoides individuals (75 to 6704 cm2, Fig. S1B).

Soil properties and litter score

The well-developed podzol was characterised by medium-sized sand as the main textural component, an acidic soil pH and extremely low nutrient concentrations (Table 1). The total soil [Mn] was 6.9 ± 0.6 mg kg−1 of which approximately 61 ± 10% was associated with organic matter. Total soil [P] was 13.1 ± 0.7 mg kg−1, with 70.3 ± 3.7% in organic form. Bicarbonate extractable soil P was determined spectrophotometrically as an estimate of plant-available P; varying from 1.4 to 4.2 mg kg−1.

Table 1 Soil properties of the topsoil layer
Full size table

Leaf [Mn] in understorey shrubs did not show a significant correlation with any of the assessed bulk soil parameters (Supp. Table 1). When we performed a regression analysis for B. eriocarpa alone, we found a significant correlation between leaf [Mn] and soil pH(CaCl2) (R2 = 0.08, F(1,62) = 5.15, P < 0.027). However, the correlation between leaf [Mn] of B. eriocarpa and soil pH(DI) was not significant (Supp. Table S1). We tested whether the litter score recorded for the nearest Banksia affected leaf [Mn] in understorey shrubs and found no significant effect in either shrub.

Analysis of factors affecting leaf [Mn] of B. eriocarpa

For B. eriocarpa, using our first approach to quantify a potential Banksia effect (distance to and size of the nearest Banksia individual), we found that leaf [Mn] declined with increasing distance to the nearest Banksia tree, increased with greater elevation and was especially high when large Banksia trees were within a radius of 2 m (R2 = 0.2788, model P = 0.003, Table 2, Fig. 2).

Table 2 Coefficient estimates and P-values for model predicting leaf manganese concentration ([Mn]) of Bossiaea eriocarpa with elevation of shrub, distance to nearest Banksia, and size of nearest Banksia a
Full size table
Fig. 2
figure 2

Observations (circles) and model predictions (lines) of leaf [Mn] of Bossiaea eriocarpa based on elevation of shrubs, distance to nearest Banksia (x-axis), and size of nearest Banksia (thick line, larger points: larger Banksia stem area, thin lines, smaller points: smaller Banksia stem area). Lines are model predictions for Banksia trees of size 100 cm2 (smaller) and 500 cm2 (larger). For values and significance of model coefficients, see Table 2

Full size image

Using our third approach, we found several feasible alternative models for B. eriocarpa based on short- and long-range Banksia effects, along with shrub area and/or elevation which had notably higher R2 (R2 > 0.6) and lower model P-values (P < 0.001) (See Supp. Table S2). These models had similar AIC values as the model presented in Fig. 2 and should thus be considered as similarly valid alternative explanations for the data. They were based on fewer data points (due to exclusion of shrubs to avoid edge effects) than the first simpler model and are more difficult to interpret due to significant interactions and correlation between explanatory variables, but, importantly, they all consistently indicate a strong positive effect of larger nearby Banksia trees (positive short-range Banksia effect). The effect of elevation was again positive, whenever it appeared in these models. These models also suggest that larger shrubs generally had lower leaf [Mn], especially if they did not have nearby Banksia trees. For the smaller shrubs, leaf [Mn] appeared to be higher in those shrubs without many Banksia trees in their wider vicinity (negative short-range Banksia effect for small shrubs). These models also indicated that the largest Banksia trees could affect B. eriocarpa shrubs up to 14 m away. We did not find a significant model for B. eriocarpa using our second approach (P = 0.056 for best model found).

Analysis of factors affecting leaf [Mn] of H. hypericoides

Using our second approach, we found a good model for H. hypericoides using a simple binary variable indicating whether a Banksia plant was within a threshold distance of the shrub, where this threshold distance was determined to be 7.3 m (R2 = 0.6115, P < 0.001, Table 3, Fig. 3). According to this model, small shrubs without Banksia individuals within 7.3 m had higher leaf [Mn] than small shrubs with Banksia trees within 7.3 m, but as shrub size increased, leaf [Mn] of shrubs without nearby Banksia trees decreased and leaf [Mn] of shrubs with Banksia trees increased, so that large shrubs without Banksia trees within 7.3 m had lower leaf [Mn] than large shrubs with Banksia trees within 7.3 m.

Table 3 Coefficient estimates and P-values for model predicting leaf manganese concentration ([Mn]) of Hibbertia hypericoides with area of shrub and simple binary variable indicating whether there was any Banksia within 7.3 m as predictors
Full size table
Fig. 3
figure 3

Predictions of leaf Mn concentrations ([Mn]) from best model for Hibbertia hypericoides. Predicted (lines) and observed (points) leaf [Mn] for H. hypericoides shrubs of different sizes (x-axis) that were or were not affected by Banksia neighbours (defined as presence of any Banksia within 7.3 m). For values and significance of model coefficients, see Table 3

Full size image

Using our third approach, we also found an alternative model for H. hypericoides based on short- and long-range Banksia effects which indicated similar patterns to the simpler binary variable model, also with a threshold distance of 7.3 m (long-range effect) (see Supp. Table S3). However, this model indicated that leaf [Mn] increased with shrub size even more steeply when there was a Banksia tree within 1 m (short-range effect). Both models for H. hypericoides showed similar patterns to the alternative models for B. eriocarpa, based on short- and long-range Banksia effects (Supp. Table S2). We did not find a significant model for H. hypericoides using our first approach (P = 0.12 for best model found).

Discussion

Belowground facilitation among plant species with different nutrient-acquisition strategies has been proposed as a mechanism that contributes to maintenance of hyperdiverse plant communities on highly-weathered soils (Lambers et al. 2015; Teste et al. 2017). Cluster-root activity increases Mn uptake, as well as leaf [Mn] in Proteaceae, including Banksia (Lambers et al. 2021). Neighbouring species without cluster roots show root intermingling and enhance nutrient uptake when grown next to Banksia plants in greenhouse experiments (Muler et al. 2014; de Britto Costa et al. 2021). In field surveys, species of the non-cluster root-forming genera Bossiaea and Hibbertia interestingly present high leaf [Mn], relative to other non-cluster-rooted species (Abrahão et al. 2018; Zhong et al. 2021; Lambers et al. 2022). However, whether the distance to Banksia trees explains variation in leaf [Mn] of neighbouring shrubs growing in natural environments has not been explored so far. We found evidence for higher leaf [Mn] in B. eriocarpa shrubs, when Banksia plants with a larger stem diameter were growing in proximity relative to shrubs that did not grow close to Banksia trees (Fig. 2). However, based on empirical data, our models also predict additional significant effects of elevation on leaf [Mn] in B. eriocarpa and of shrub biomass in both B. eriocarpa and H. hypericoides (estimated by shrub canopy size, Supp. Table S2, Fig. 3).

Variation in leaf [Mn] in two understorey shrub species

In both sympatric species, B. eriocarpa and H. hypericoides, we found a large variation in [Mn] of ~ 1-year old leaves which was not related to bulk soil [Mn] or any other topsoil parameters assessed, suggesting that the occurrence of a Banksia effect could be tested with these two species. Within our study site, we observed the highest leaf [Mn] in B. eriocarpa (> 400 mg kg−1), while maximum values in H. hypericoides were below 300 mg kg−1. Both shrub species share similar flowering phenology and responses to fire-prone environments, but differences in root system architecture have been reported (Dodd et al. 1984; Pate and Bell 1999; Veneklaas and Poot 2003; Clarke et al. 2015). While H. hypericoides produces a shallow main root with several lateral roots in the uppermost soil layers (Pate and Bell 1999), B. eriocarpa presents fewer lateral roots and produces a deeper taproot reaching ~ 2 m in depth (Dodd et al. 1984). A detailed analysis of water relations in a Banksia woodland showed that shrubs with deeper root systems present higher stomatal conductance during the onset of drier summer months, whereas shrubs with shallower root systems reduce stomatal conductance as early as late spring (Veneklaas and Poot 2003). If the deeper taproot of B. eriocarpa allowed the shrubs to access water for a longer time throughout the year, it is conceivable that faster cumulative transpiration rates contributed to the higher [Mn] of mature leaves relative to the leaves of shallow-rooted H. hypericoides shrubs.

The effect of elevation on leaf [Mn] in B. eriocarpa

To test whether the large variation in leaf [Mn] might be explained by a Banksia effect, we used a modelling approach using elevation, shrub canopy area and different definitions of a potential Banksia effect as explanatory variables. For B. eriocarpa, leaf [Mn] was higher on top of the dune, particularly when larger Banksia trees were present within 2 m. This positive effect of elevation could be related to differences in the extent of cluster-root formation and activity, as B. attenuata and B. menziesii depend on water uptake from deep sand layers (> 4 m depth), which are not present at the lower parts of the dune (Groom 2004; Gao et al. 2020). B. eriocarpa forms fewer lateral roots than H. hypericoides (Pate and Bell 1999) and might, therefore, be more sensitive to variation in Banksia performance.

Banksia effects and interactions with shrub biomass on leaf [Mn]

Our data show that leaf [Mn] was higher in B. eriocarpa shrubs that had very close Banksia neighbours than in conspecifics without Banksia individuals within approximately 2 m of the shrub’s base, partially confirming our initial hypothesis (Fig. 2).

In addition to the positive effect of close Banksia trees on shrub leaf [Mn], we found an interaction of Banksia effect with shrub canopy area in the more complex models for B. eriocarpa and in a simpler model for H. hypericoides (Supp. Table S2, Fig. 3). Larger shrubs had lower leaf [Mn] when there were no nearby Banksia trees present, when compared with larger shrubs growing close to Banksia individuals; and smaller shrubs had higher leaf [Mn] when no Banksias trees were present nearby. This result indicates that complex interactions exist between plant size and accumulation of Mn that tends to follow the transpiration stream.

If cluster-root activity facilitates Mn uptake in neighbouring shrubs, why are leaf [Mn] higher in small shrubs growing in areas with low Banksia density relative to small shrubs grown in areas with a high Banksia effect? One possible explanation is that the presence of Banksia neighbours alters root biomass accumulation at greater soil depth, as well as root system architecture. Neighbour effects on root-allocation patterns have been reported previously (Gersani et al. 2001; Li et al. 2006; de Britto Costa et al. 2021). In a semiarid ecosystem, such changes in root system architecture might result in relatively shorter or longer periods of transpiration throughout the year. Accordingly, small shrubs growing away from Banksia trees would produce deeper main roots, have faster annual cumulative transpiration rates and therefore accumulate more Mn in their leaves. Small shrubs growing among Banksia individuals could produce more lateral roots and shorter main roots. When grown in rhizoboxes, H. racemosa allocated more root biomass towards the roots of a B. attenuata neighbour than towards the roots of a conspecific (de Britto Costa et al. 2021). During the dry season, Banksia trees exhibit hydraulic lift of water from deeper to more shallow soil horizons (Burgess et al. 2000) which might influence rooting patterns in neighbouring shrubs. However, the idea of differential rooting depths of understorey shrubs in response to Banksia neighbours requires further investigation. Root allocation, together with seasonal patterns in plant water relations of shrubs growing close to or away from Banksia trees have to be assessed in the future.

In conclusion, our study indicates that Mn uptake of understorey shrubs can be facilitated by very close Banksia trees, especially if they are tall. Our model predictions reveal that complex interactions between Banksia effect and neighbouring shrub size exist which might be related to the rooting depths of understorey shrubs resulting in different annual transpiration rates and ultimately leaf [Mn].