Exploring rock fissures: does a specialized root morphology explain endemism on granite outcrops?

Abstract

Background and Aims

Worldwide, many plant species are confined to open, shallow-soil, rocky habitats. Although several hypotheses have been proposed to explain this habitat specificity, none has been convincing. We suggest that the high level of endemism on shallow soils is related to the edaphic specialization needed to survive in these often extremely drought-prone habitats. Previous research has shown that species endemic to ironstone communities in SW Australia have a specialized root morphology that enhances their chance to access fissures in the underlying rock. Here we test the generality of these findings for species that are confined to a shallow-soil habitat that is of much greater global significance: granite outcrops.

Methods

We compared temporal and spatial root growth and allocation of three endemic woody perennials of SW Australian granite outcrop communities with those of congeners occurring on nearby deeper soils. Seedlings of all species were grown in 1·2 m long custom-made containers with a transparent bottom that allowed monitoring of root growth over time.

Key Results

The granite outcrop endemics mostly differed in a predictable way from their congeners from deeper soils. They generally invested a larger portion of their biomass in roots, distributed their roots faster and more evenly over the container and had a lower specific root length. In different species pairs the outcrop endemics achieved their apparent advantage by a different combination of the aforementioned traits.

Conclusions

Our results are consistent with earlier work, indicating that species restricted to different types of drought-prone shallow-soil communities have undergone similar selection pressures. Although adaptive in their own habitat in terms of obtaining access to fissures in the underlying rock, these root system traits are likely to be maladaptive in deeper soil habitats. Therefore, our results may provide an explanation for the narrow endemism of many shallow-soil endemics.

INTRODUCTION

Worldwide, a large number of plant species are confined to relatively open, shallow-soil, rocky habitats (Kruckeberg and Rabinowitz, 1985; Baskin and Baskin, 1988; Pate and Hopper, 1993; Medail and Verlaque, 1997; Porembski and Barthlott, 2000; also see Figs 1 and 2 for some SW Australian species). Hypotheses invoked to explain such narrow endemism range from requirements for high light levels, specialized chemicals or biota in soils derived from bedrock, to limited genetic variability and poor competitive ability (Baskin and Baskin, 1988; Lavergne et al., 2004). Experimental work thus far supports shade intolerance as the most plausible hypothesis (Baskin and Baskin, 1988). Here, we suggest that the strong habitat specificity of many shallow-soil endemics is related to the degree of edaphic specialization that is needed to establish and survive in these often harsh habitats, and the incompatibility of these adaptations with deeper soil environments. The thin soils on which many shallow-soil endemics occur provide a limited water supply which is especially challenging for perennial species in seasonally dry or otherwise water-limited environments.

Fig. 1.

Species included in this study: (A) Calothamnus rupestris (left, 1·5 m tall) and Eucalyptus caesia subsp. caesia (4 m), Boyagin; (B) Allocasuarina huegeliana (7 m), Boyagin; (C) Calothamnus quadrifidus (2·5 m), King Rocks; (D) Eucalyptus drummondii (5 m), Boyagin; (E, F) Allocasuarina humilis (0·8 m), Little Mt Lindesay. Photographs: S. D. Hopper.

Fig. 2.

Mortality on granite outcrops in the Southwest Australian Floristic Region and adjacent Arid Zone associated with the record-breaking heat and drought during the summer of 2010/2011. Examples are from three rainfall zones [(A) high rainfall, 900 mm p.a.; (B, C) semi-arid, 300–500 mm p.a.; (D) arid, <300 mm p.a.]. (A) Dead Eucalyptus marginata and associated vegetation in the Northern Jarrah Forest; (B) dead and live Acacia lasiocalyx, Chiddarcooping; (C) dead and live Thryptomene australis and Dodonaea viscosa, with taller Eucalyptus caesia, Hakea petiolaris and Acacia lasiocalyx alive on deeper soils, NE of Pingaring; (D) dead and live Thryptomene australis and Allocasuarina campestris, Geelabbing (Mt Churchman). Photographs (A) Mike Pez, (B–D) S. D. Hopper.

In the Southwest Australian Floristic Region (SWAFR sensu Hopper and Gioia, 2004), which experiences a Mediterranean climate, and undoubtedly in many other seasonally dry climates, the most severe drought stress is experienced by plants in shallow soil overlying rocky hills and ridges (e.g. Mishio, 1992; Groom and Lamont, 1995). This has been exemplified by the recent episode of plant deaths on and around SW Australian granite outcrops (G. Matusick, Murdoch University, Western Australia, pers. comm.; S. D. Hopper, UWA, Western Australia, unpubl. res.; see Fig. 2) in association with one of the driest winters on record followed by one of the driest and hottest summers on record (Bureau of Meteorology, 2011). These observations would suggest that local shallow-soil habitats could actually function as the ‘bush canaries’ for sensing warming and/or drying impacts through time.

There is growing evidence that, during the dry season, water held within the underlying bedrock is essential for meeting the transpiration demands of shrubs and trees (e.g. Rose et al., 2003; Querejeta et al., 2007). Therefore, obtaining root or mycorrhizal (Bornyasz et al., 2005) access to the water stored in the weathered bedrock via fissures and micropores is probably essential for the establishment and survival of many rock endemics. Indeed there is ample evidence of roots of woody species growing through the bedrock (e.g. Zwieniecki and Newton, 1994, 1995; Canadell et al., 1996; Sternberg et al., 1996; Witty et al., 2003; Bornyasz et al., 2005; Schenk, 2008). However, very little is known about how shallow-soil endemics locate suitable cracks and whether they have evolved any specific root system adaptations to increase their success.

In a previous study, we showed that two rare Hakea species, confined to shallow-soil ironstone communities in the SWAFR, differ from four of their common congeners in several root system characteristics that probably confer an advantage in locating cracks and fissures in the underlying rock (Poot and Lambers, 2003a, b, 2008; Poot et al., 2008). As very young seedlings, the rare species invest more of their biomass in roots, as was also found by Walck et al. (1999) in a comparison of a narrowly endemic rock outcrop Solidago species with a common congener. Within their root systems the rare Hakea species allocated more biomass to deep roots and produced longer laterals. As a consequence, they explored the bottom of their 1·8 m long and 0·15 m deep containers much more thoroughly than their congeners from deeper soil. Subsequently, in a reciprocal transplant experiment, we showed that first-summer survival rates of seedlings in the shallow-soil ironstone habitats was much lower than those in any of the other habitats, with the shallow-soil endemics having much higher survival rates than their common congeners (Poot and Lambers, 2008).

Shallow-soil ironstone communities are relatively rare and occur on a geologically distinct substrate (e.g. Johnstone et al., 1973; Gibson et al., 2000). Therefore, the observed differences in root system morphology may have evolved in response to the particular pedogenesis of these habitats, rather than the shallow-soil condition per se. In the present study we test the generality of our earlier findings on ironstone endemics by investigating perennial woody species confined to a shallow-soil habitat that is of much greater global significance: granite outcrops.

The SWAFR is recognized as one of 25 global biodiversity hotspots, i.e. regions on Earth richest in endemic species under threat (Myers et al., 2000). Approximately 2000 of a total of 8000 plant species in SW Australia have been estimated to occur on and around granite outcrops (Hopper et al., 1997; Hopper and Gioia, 2004), including many rare species that are confined to these shallow-soil habitats. The richness of outcrops is undoubtedly related to the myriad of different habitats that are found within close proximity, ranging from bare rock, cryptogamic crusts, rich herbfields on shallow soils, gnammas or rockpools, to areas with taller perennial vegetation in soil-filled fissures, and in deeper soils surrounding the fringes of the rock (Figs 1 and 2). Thus, apart from containing habitats that are considered among the most drought-prone and harsh environments (i.e. shallow soil pockets on the rock), the down-slope fringes of outcrops that collect the runoff and the permanent gnammas are often considered mesic refuges.

For the current study, three species were chosen that are common in fissures and smaller soil pockets on the higher slopes of a typical Western Australian granite outcrop (Boyagin Rock). These are the microhabitats where plant establishment and survival may be highly dependent on the root foraging strategy of a species. In a glasshouse experiment, 4- to 6-month-old seedlings of the outcrop species were compared with those of three congeneric species (i.e. there were three species pairs) that occurred nearby on deeper soil. Seedlings were grown on long (1·2 m) and shallow (0·15 m) custom-made containers that had a transparent bottom to allow the recording of horizontal root spread over time. We hypothesized that the granite outcrop endemics have a root foraging strategy that increases their chance of finding cracks in the underlying rocks (sensu Poot and Lambers, 2003a, 2008), by (1) investing relatively more biomass in their roots (i.e. they will have a higher root mass ratio); (2) spreading their roots faster over the bottom of the containers; resulting in (3) a more homogeneous horizontal distribution of roots at the final harvest (i.e. they will have a smaller percentage of their root system in the centre of the containers).

MATERIALS AND METHODS

Experimental species

All species included in the study were chosen from Boyagin Nature Reserve (32 °28′23′′S, 116 °52′43′′E), approx. 110 km ESE of Perth, Western Australia. This 6700 ha reserve is situated on the western edge of the West Australian wheatbelt and contains a variety of habitats, including open eucalypt woodlands, heathlands and granite outcrops (Fig. 1). We chose three locally common, woody outcrop species [Calothamnus rupestris Schauer, Eucalyptus caesia Benth. and Allocasuarina huegeliana (Miq.) L.A.S. Johnson] on the basis of (1) their commonness on shallow-soil patches on one of the twin Boyagin Rock inselbergs (referred to as ‘Boyagin Rock’); and (2) the availability of seeds. The latter restricted species choice to those that retain seeds for multiple years in their canopies (i.e. species that are bradysporous), a feature that is common in the SW Australian flora (e.g. Cowling et al., 1996). Both C. rupestris and E. caesia are relatively rare species that are restricted to shallow soil on granite outcrops. On Boyagin Rock, both species occur on shallow-soil patches near the top of one of the dome-shaped inselbergs (i.e. bornhardts; Twidale, 2007). Allocasuarina huegeliana (rock sheoak) is a much more common and widespread species in SW Australia which is associated with granite outcrops and typically occurs in dense stands near the base of the rock. On Boyagin Rock this species also occurs more sparingly on shallow-soil patches closer to the top.

The three granite-outcrop species were contrasted with congeners that (1) occurred nearby on deeper soils; and (2) that had a similar life form: C. quadrifidus R.Br., E. drummondii Benth. and A. humilis (Otto & A. Dietr.) L.A.S. Johnson. Only for the Allocasuarina comparison did we not find a congeneric species with the same life form, as the granite outcrop species A. huegeliana generally grows as a tree up to 10 m tall, whereas A. humilis is a shrub growing to a maximum height of 2 m. All three species are widespread and occur in a variety of soil types and habitats. In Boyagin Nature Reserve, they grow in an area <1 km SW of Boyagin Rock. Habitats consisted of open Eucalyptus wandoo woodland (for C. quadrifidus and A. humilis), grading into tall open heath (for E. drummondii). For an overview of characteristics of the experimental species, see Table 1 and illustrations in Fig. 1.

Table 1.

Selected characteristics of the experimental species used in this study

Species	Family	Distribution	Life form	Fire strategy
Calothamnus rupestris	Myrtaceae	Restricted range outcrop endemic	Shrub (0·9–4 m)	Resprouter (Bell, 1994)
Calothamnus quadrifidus	Myrtaceae	Widespread	Shrub (0·5–3 m)	Resprouter (George and Gibson, 2010)
Eucalyptus caesia	Myrtaceae	Restricted range outcrop endemic	Mallee (up to 14 m)	Resprouter (Hopper, 2000)
Eucalyptus drummondii	Myrtaceae	Widespread	Mallee (up to 8 m)	Resprouter (Brooker and Kleinig, 1990)
Allocasuarina huegeliana	Casuarinaceae	Widespread, associated with granite	Tree (4–10 m)	Seeder (Hopper, 2000)
Allocasuarina humilis	Casuarinaceae	Widespread	Shrub (0·2–2 m)	Resprouter (Herath and Lamont, 2009)

Experimental design

Woody follicles with seeds were collected for all species in June and September 2005 from Boyagin Nature Reserve. For each species, seeds were collected from at least five plants. Seeds of all species were placed just under the surface of native low-nutrient potting mix in germination trays on 13 October 2005. The mix was amended with 1·5 g L⁻¹ slow-release fertilizer. Trays were placed in a glasshouse at the University of Western Australia and were watered several times every day during the germination period. All species, apart from A. humilis, reached 50 % germination after approx. 12 d, with overall germination levels reaching 80–100 %. Allocasuarina humilis reached 50 % germination after 27 d and had an overall germination of 60 %.

After 6 weeks of growth in the germination trays, two randomly selected seedlings were transplanted to the centre of custom-made free-draining containers [1·2 m (length) × 0·2 m (width) × 0·15 m (height); transparent PVC], with five containers per species. The sides of the containers were first painted black to prevent light penetration and then white to prevent excessive heat absorption, whereas the bottom of the containers was covered by removable black plastic sheets. Containers were filled to a height of 0·1 m with a soil mixture containing 50 % Gin Gin loam, 40 % river sand and 10 % jarrah sawdust (Richgro, Jandakot, Western Australia), which was amended with 1·5 g L⁻¹ slow-release fertilizer. After 3 weeks, the smallest and/or least healthy-looking seedling per container was removed. All containers were placed randomly on four neighbouring benches in a glasshouse. Seedlings of the late-germinating species A. humilis were transplanted 2 weeks after the other species.

The rate of horizontal spread of the root system over the bottom of the container was monitored weekly (visible through the transparent bottom of the container) by measuring the longest distance from one far side of the container to the other in which roots were visible (i.e. maximally 1·2 m). Within each species pair, the date of final harvest was based on most replicates of one species reaching the far sides of the container. After plants of this species were harvested, the other species of the pair, which for all pairs had a visibly smaller aboveground biomass, was harvested approx. 18 d later, to be able to compare the species within a pair at a similar biomass. For the Calothamnus comparison, the outcrop-endemic (C. rupestris) was harvested after the non-outcrop species (C. quadrifidus), whereas the opposite was the case for the other two genera. The total growing period in the containers was 78–96 d for the Eucalyptus comparison, 99–117 d for the Calothamnus comparison and 111–131 d for the Allocasuarina comparison (i.e. numbers indicate the starting date of harvests for either of the species within a pair, with harvests for each species completed within several days).

At the final harvest, fresh and dry masses (48 h at 70 °C) of stems and leaves were determined. To ascertain spatial root placement, roots were harvested separately for 14 different compartments (two vertical and seven horizontal). The vertical compartments were 0·01 m (lower) and 0·09 m deep (upper), whereas the horizontal compartments were all 0·17 m long. Containers were accessed by first removing one of the far sides, after which a vertically sharpened stainless steel plate was driven in to separate the first horizontal compartment (i.e. 0·17 m in from the far side of the container). Thereafter, a horizontal steel plate (0·01 m deep) was driven in from the bottom far side of the container to separate the first two vertical compartments. Subsequently roots of the upper compartment were washed out, followed by roots of the lower compartment (by removing the horizontal steel plate), after which roots were rinsed with water to remove all remaining soil. These procedures were repeated until all compartments were washed out. In the upper compartment, just below the base of the plant, the thickened root system just below the root–shoot junction was harvested separately from the rest. All roots were stored at 4 °C for a maximum of 2 weeks awaiting further determination of root characteristics. Roots were computer scanned and analysed using Win Rhizo V3·9 software (Regent Instruments, Quebec, Canada). Total root length, average root diameter and number of tips were determined for each compartment. After scanning, oven-dry masses of roots were determined (after 48 h at 70 °C). In the calculation of average values of species' root system characteristics (Table 2), the thickened root section below the root–shoot junction was not included for average root diameter, specific root length (SRL) and root dry mass percentage. The exclusion of this section ensured that the values presented are a good reflection of the majority of each plant's root system.

Table 2.

Root characteristics (± s.e.) at the final harvest for three outcrop/non-outcrop species pairs

	Calothamnus rupestris/quadrifidus	Eucalyptus caesia/drummondi	Allocasuarina huegeliana/humilis
Total fresh mass (g)	45·7 ± 4·1/28·4 ± 6·0*	13·0 ± 2·9/5·8 ± 0·79*	21·7 ± 2·2/15·7 ± 1·9
% fresh mass bottom	11·7 ± 1·3/26·1 ± 3·9**	29·3 ± 1·7/31·6 ± 3·3	38·6 ± 4·3/32·9 ± 3·4
% fresh mass outer	54·3 ± 5·5/22·7 ± 8·3*	25·0 ± 11·4/8·9 ± 3·5	43·4 ± 2·1/16·2 ± 4·8***
Total length (m)	267 ± 26/224 ± 34	83·9 ± 15·6/52·3 ± 4·1	138 ± 12/120 ± 14
Number of tips	105 ± 11/94·2 ± 8·8	40·4 ± 7·5/20·8 ± 1·1*	59·6 ± 4·1/47·4 ± 7·4
Root diameter (mm)	0·48 ± 0·04/0·33 ± 0·02**	0·39 ± 0·02/0·33 ± 0·02*	0·38 ± 0·01/0·33 ± 0·01*
SRL (m g−1)	90·2 ± 5·7/169 ± 24**	137 ± 13/195 ± 21*	86·4 ± 5·6/108 ± 7·4*
Dry mass (%)	6·81 ± 0·37/5·34 ± 0·41*	5·2 ± 0·39/5·5 ± 0·29	8·21 ± 0·76/8·04 ± 0·45

Asterisks indicate significant differences within an outcrop/non-outcrop species comparison (P < 0·05, after one-way ANOVA of ln-transformed values; * 0·01 < P < 0·05, ** 0·001 < P < 0·01).

Statistical analysis

Differences in biomass allocation and general root characteristics amongst species within each pair were tested with one-way analysis of variance (ANOVA; Genstat ver. 12·1, VSN International, Oxford UK). Variables were ln-transformed if residual plots indicated increasing variances at larger mean values. Differences in spatial distribution pattern were tested by calculating mean values for the outer and inner root compartments (i.e. compartments 1, 2, 6 and 7, vs. 3, 4 and 5, respectively, in Fig. 4) and for the upper and lower root compartments and testing differences among species within a pair by one-way ANOVA. Differences in temporal root growth along the bottom of the containers were tested using repeated measurement ANOVA in the same statistical package.

Fig. 4.

Percentage of root fresh mass allocated to the bottom 10 mm (‘lower’) and top 90 mm (‘upper’) layers of 1·20 m long containers as well as allocation to the lignified root section just below the root–shoot junction (top root), of three outcrop (left panels) and non-outcrop (right panels) congeneric species pairs (Calothamnus, Eucalyptus and Allocasuarina). Note that the seedlings were planted in the centre of the containers (compartment 4).

RESULTS

Whole-root characteristics

Although we attempted to harvest the species within each congeneric pair at a similar biomass to decrease possible allometric effects, this was not fully achieved within the Allocasuarina comparison. The outcrop-endemic A. huegeliana attained a significantly larger biomass than its non-outcrop counterpart (Fig. 3A). In terms of biomass allocation, the two outcrop endemics, C. rupestris and E. caesia, allocated a greater proportion of their total biomass to roots (Fig. 3B). As a consequence, both species had considerably more root fresh mass than their deeper soil congeners at the end of the experiment (Table 2). Despite significant differences in total root mass, none of the species pairs differed significantly in total root length. The lack of difference in total root length was mainly caused by outcrop endemics producing thicker roots (i.e. with a larger average root diameter) with a lower SRL (i.e. they produced less root length per unit root mass; Table 2). In all three species pairs, the thicker roots of outcrop endemics tended to be associated with the upper layers (results not shown). No consistent differences in root-tip number or root dry mass percentage were observed amongst the species from the different habitat groups.

Fig. 3.

(A) Total plant dry mass and (B) percentage of dry mass allocated to roots (root mass ratio) ± s.e., at the final harvest for three outcrop and non-outcrop species pairs of Calothamnus, Eucalyptus and Allocasuarina. Asterisks indicate significant differences within an outcrop/non-outcrop species comparison (one-way ANOVA of ln-transformed values; * 0·05 > P >0·01, ** 0·01> P >0·001).

Spatial and temporal root distribution

At the final harvest, the spatial root distribution differed considerably amongst species (Fig. 4). The rock outcrop endemics in the Calothamnus and Allocasuarina comparisons had a much more homogeneous root system distribution than their deeper soil congeners, whereas in the Eucalyptus comparison the difference was non-significant. The rock outcrop-endemic C. rupestris allocated significantly more root mass to the sides of the containers (i.e. compartments 1, 2, 6 and 7; Fig. 4, Table 2) and to the upper compartments (i.e. the top 90 mm; Fig. 4, Table 2), than its deeper soil congener C. quadrifidus. Whereas the deeper soil C. quadrifidus still had 40 % of its root mass in the centre compartment of the pot and <2 % on one of the sides, C. rupestris distributed its roots homogeneously in the container, with no differences amongst compartments (Fig. 4). Similarly, the rock outcrop species A. huegeliana allocated much more root mass to the sides of the container (Fig. 4, Table 2) than its deeper soil congener, A. humilis. In this comparison there was no difference between the species in their relative vertical root allocation. In the Eucalyptus comparison the granite outcrop-endemic E. caesia invested less root mass in the top root (3·3 % vs. 10·4 %, P < 0·001) and there was no significant difference in allocation to either the sides of the container or the bottom vs. top compartments (Fig. 4, Table 2).

The transparent bottoms of the containers allowed us to evaluate the speed at which root systems of different species explored their length. For two out of the three comparisons (Eucalyptus and Allocasuarina) the outcrop endemics explored the bottom layer of the containers faster than their congeners from deeper soils (Fig. 5). Surprisingly, the species that had the most homogeneous root system distribution at the final harvest (C. rupestris) did not differ significantly (P = 0·19) from its deep-soil congener (C. quadrifidus).

Fig. 5.

Horizontal spread of the root system along the bottom of the container for three outcrop/non-outcrop congeneric species pairs: (A) Calothamnus, (B) Eucalyptus, (C) Allocasuarina. Asterisks indicate significant differences within an outcrop/non-outcrop species comparison (one-way repeated measurements ANOVA for the overlapping time period; * 0·01 < P < 0·05, ** 0·001 < P < 0·01).

DISCUSSION

Root system characteristics of shallow-soil species

Several studies have shown that a considerable fraction of soil water in shallow-soil rocky communities is located in the rock matrix (e.g. Graham et al., 1997; Duniway et al., 2007; Jacob et al., 2009) and it is likely that in seasonally dry habitats, obtaining access to this water is crucial for plant survival. As transport of water through micropores in the rock along water potential gradients would be extremely slow (Hubbert et al., 2001; Schwinning, 2010), access to this stored water seems to depend on coarse root access to fissures in the rock (i.e. to get close to the stored water) and subsequent fine root or mycorrhizal access to the local rock matrix (e.g. Bornyasz et al., 2005; Schwinning, 2010), or to a perched water table. In the present study we have shown that three SW Australian granite outcrop species differed from deeper soil congeners in at least two out of three root system characteristics that confer an advantage in terms of exploring a large rock surface area for fissures. They (1) generally invested more biomass in their roots; (2) explored the bottom of their containers faster and; (3) at the end of the experiment had allocated a larger proportion of their roots to the outer parts of the container. These differences are similar to those obtained in our earlier work in a comparison of shallow-soil ironstone endemics and their common congeners from deeper soil (Poot and Lambers, 2003a, 2008, also see Walck et al., 1999 for similar differences in root mass ratio). This suggests that the root system of many woody perennials ocurring in different types of seasonally dry, shallow-soil environments may have been shaped by selection forces increasing the chance to access fissures in the underlying rock.

Apart from the above-mentioned traits, the shallow-soil endemics also consistently had a lower SRL which was mainly due to their larger average root diameter. This seems counterintuitive as a lower SRL, all else being equal, would lead to the exploration of a smaller rock surface area. Nonetheless, in two out of three species pairs the outcrop species (i.e. E. caesia and A. huegeliana) explored the length of the bottom of the containers faster (Fig. 5), whereas in the third pair, the outcrop species C. rupestris had allocated substantially more root mass to the sides of the container at the final harvest (Fig. 4). The latter finding strongly suggests that C. rupestris also explored space faster than its deeper soil congener and, given a larger container size, would have explored a larger bottom surface area as well. We hypothesize that the faster exploration of space of the shallow-soil endemics is likely to be a consequence of a less ramified root system with a high investment in few, predominantly lower order (i.e. large diameter) laterals, with a strong apical dominance, that mainly function to explore space and transport water. This contrasts with the dominant proliferation of laterals close to the base of the plant as observed for all deeper soil species in this study and in Poot and Lambers (2008). Low SRL, high diameter roots have been associated with greater longevity (e.g. Eissenstat, 1992), faster growth (e.g. Eissenstat, 1991, and references therein) and greater hydraulic conductance (Passioura, 1988), all characteristics that appear essential for obtaining permanent and fast access to the extremely heterogenous supply of water through fissures in the rock. However, these traits come at a cost as they are also associated with higher construction costs and a lower efficiency for water and nutrient uptake (e.g. Eissenstat, 1992). Thus, the lower SRL of shallow-soil endemics in this study and their higher average root diameter (see a similar trend for root diameter in Poot and Lambers, 2008) may be functional and possibly reflect a trade-off between local water and nutrient acquisition vs. space exploration and water transport.

Mechanisms for detection of fissures in the underlying rock

Although the observed root traits of the shallow-soil species are likely to increase the rock surface area that can be explored, the eventual detection of fissures may be a matter of gravitropic and/or hydrotropic responses, with roots following the pathways of infiltrating water as suggested by Schenk (2008). However, fissures filled with water during the wet season, when roots are actively growing, may quickly dry out during the dry season. Thus, investing roots in the ‘wrong’ fissure may be quite costly. This is especially relevant for establishing seedlings, and we hypothesize that seedlings of shallow-soil endemics, apart from having a specialized morphology or architecture, may have evolved mechanisms to discriminate between fissures.

Most fissures that lead to a reliable water source in a seasonally dry environment (e.g. a perched water table inside the rock) will already be occupied by plant roots, and would only be accessible after adult plant mortality such as after a fire or a prolonged drought. Thus, during a regeneration event, dead woody plant roots would be amongst the very few indicators of suitable water access points, and the ability to track them may be highly beneficial. However, the growth of most plant roots is thought to be inhibited when encountering an ‘obstacle’. They change direction to grow towards areas of less resistance (e.g. Clark et al., 2003), or divert resources to increase growth of other laterals or initiate lateral primordia (Misra and Gibbons, 1996; Thaler and Pages, 1999; Falik et al., 2005). Although, this general behaviour seems functional in deeper soils and prevents allocation to areas that are less likely to yield resources, it is likely to be detrimental on seasonally dry shallow soils as it would divert resources away from exploring a larger rock surface area. Indeed, in contrast to deeper soil congeners, Hakea seedlings from shallow-soil ironstone communities did not show this root foraging behaviour (Poot and Lambers, 2003a), and had much higher first summer survival rates when transplanted to their native shallow-soil habitat (Poot and Lambers, 2008). Thus, roots of shallow-soil endemics may have an altered sensing mechanism that ensures continued apical dominance of roots foraging for fissures and that may also allow for tracking of dead woody roots.

Advantages and disadvantages of the observed root traits as dependent on habitat

Although the observed root traits of shallow-soil endemics are likely to be adaptive, their selective advantage may strongly depend on local habitat features. These would include the frequency of fissures and crevices in the underlying rock, their shape and size (e.g. Zwieniecki and Newton, 1994), their spatial distribution and their reliability in terms of water supply during the dry season. For example, in rocky habitats with an abundance of fissures and a short dry season, there may not be an advantage. Conversely, in habitats where entry points into the rock are scarce and the dry season is long and hot, the observed root adaptations could be crucial. However, in the latter case, further drought tolerance mechanisms may also be required, especially if most of the water is locked up in the rock matrix (i.e. in micropores) as opposed to being more freely available through perched water tables. The limited water transport capacity of mycorrhizal hyphae and fine roots that acces the rock matrix would then necessitate further water loss regulation.

Compared with shallow-soil environments, deeper soil habitats generally are less open, support more standing biomass and have higher plant densities. Under these conditions the root traits of shallow-soil endemics are likely to be maladaptive for several reasons. Too much early investment in roots would be at the cost of aboveground competitiveness (see also Lavergne et al., 2004), which may be crucial for establishment and survival in deeper soil habitats with higher plant densities. Similarly, lateral root exploration may be greatly constrained due to stronger belowground competition. Also, the comparatively small investment of shallow-soil endemics in roots close to the base of the plant would compromise their ability to compete for nutrients and space. Finally, the potentially altered sensing mechanism of their roots would make them less able to avoid other ‘objects’ such as the roots of competing species (see Schenk et al., 1999). Thus, the high degree of endemism amongst species occurring on shallow soils (Kruckeberg and Rabinowitz, 1985; Baskin and Baskin, 1988; Pate and Hopper, 1993; Medail and Verlaque, 1997; Porembski and Barthlott, 2000), and their reduced competitive ability in other habitats (e.g. Hart, 1980; Baskin and Baskin, 1988; Walck et al., 1999; Lavergne et al., 2004), may be partly a consequence of the constraints of their specialized root system morphology (i.e. cost of specialization).

Do all species occurring in shallow-soil habitat need specialized root traits?

Although worldwide many perennial woody species in shallow-soil habitats are endemic to them, most communities have many species that also occur in the surrounding matrix (e.g. Cook et al., 2002). In the context of root system adaptations to shallow soils, this suggests that these species have different ecotypes in the different environments (see Chapman and Jones, 1975 for an example of differentiation in drought tolerance), are more plastic or do not require any special adaptations. The latter can easily be envisaged for species that commonly occur in the vicinity of outcrops and produce many, highly dispersable seeds. Inevitably, some of these seeds would germinate directly above a fissure in the underlying rock and are likely to establish without any specialized root traits.

Interestingly, Hunter (2003) showed that in eastern Australian granite outcrop communities most of the perennial woody species that also commonly occur in the matrix are resprouters, whereas most of the species that are confined to the outcrop are obligate seeders. Also, worldwide the proportion of obligate seeders is higher on outcrops than in the surrounding matrix (Ashton and Webb, 1977; Fuls et al., 1992; Gröger and Barthlott, 1996; Hopper et al., 1997). These observations suggest that obligate seeders have a relative advantage on outcrops when compared with resprouters, which has often been ascribed to the lower fire frequency in these habitats (e.g. Clarke, 2002). An alternative hypothesis would be that seeders, by not investing heavily in an underground lignotuber, can allocate more resources to an explorative root system. Together with their often better dispersable seeds and higher seed production compared with resprouters (e.g. Zammitt and Westoby, 1987), this would give them a relative advantage in the strong competition for the limited number of safe seedling sites (i.e. fissures and crevices) on outcrops. We realize that two of the shallow-soil endemics discussed in this paper are known as resprouters, but we believe that this does not necessarily weaken this argument. Both of these species are known to produce many seeds (personal observation, uncharacteristic for resprouters), highlighting that the dichotomous distinction between seeders and resprouters is in reality more a continuum. Also, the potential disadvantage of resprouters may to a large extent depend on the timing of the formation of the lignotuber, with species investing in it early at a disadvantage compared with those investing in it later. Although long-lived resprouters may be disadvantaged in establishing themselves on an outcrop for the above reasons, once they have established themselves they may persist there for a long time.

Conclusions and future directions

Our results have demonstrated that several species endemic to shallow-soil granite outcrop communities in SW Australia have specific root traits that probably increase their chance to access fissures in the underlying rock. They generally allocated more biomass to roots, spread them faster over the bottom of the long containers and allocated more roots to the sides of the container. These findings are consistent with our earlier work on shallow-soil ironstone endemics, suggesting that these adaptations may be common amongst woody species inhabiting seasonally dry shallow soils, and provide an additional explanation for the strong habitat specificity of many shallow-soil endemics worldwide. Future work, in detailed glasshouse studies, as well as in the field will have to elucidate (1) how common these specialized root traits are in other shallow-soil communities around the world; (2) whether species that posses these traits have the inferred competitive advantage on shallow soils and what specific habitat features this may depend on; and (3) how important these traits and their plasticity are in explaining the absence of shallow-soil endemics in surrounding deeper soil habitats. Apart from shallow-soil communities, our findings may also be highly relevant for plant communities growing on highly compacted soil or soil with high bulk density in which soil pores are often invoked as being crucial for resource supply.