Characterising fine-scale variation in plant species richness and endemism across topographically complex, semi-arid landscapes

Journal of Arid Environments xxx (xxxx) xxx–xxx

Contents lists available at ScienceDirect

Journal of Arid Environments journal homepage: www.elsevier.com/locate/jaridenv

Characterising fine-scale variation in plant species richness and endemism across topographically complex, semi-arid landscapes G. Di Virgilioa,b,∗, G.W. Wardell-Johnsona, T.P. Robinsonc, D. Temple-Smithd, J. Hesforde a

School of Molecular and Life Sciences, Curtin University, GPO Box U1987, Perth, 6845 WA, Australia Climate Change Research Centre, School of Biological Earth and Environmental Sciences, University of New South Wales, Sydney, 2052, Australia c School of Earth and Planetary Sciences, Curtin University, GPO Box U1987, Perth, 6845 WA, Australia d Mineral Resources Limited, 1 Sleat Rd, Applecross, 6153 WA, Australia e Tetris Environmental Pty Ltd, PO Box 3103, Myaree, WA 6154, Australia b

A R T I C LE I N FO

A B S T R A C T

Keywords: Conservation Digital elevation model Environmental heterogeneity Inselbergs LiDAR

The banded ironstone formation (BIF) ranges of south-western Australia are prominent landforms in a flat landscape and host a diverse flora. Plant diversity is expected to have a positive relationship with environmental heterogeneity in these ranges. However, there has been a lack of high-resolution data to assess how fine-scale environmental variation structures changes in plant communities across these ranges. We calculated species richness and corrected weighted endemism over 659 quadrats (each 400 m2) and investigated their spatial distribution across BIFs in relation to 1 m resolution variables (microtopographic heterogeneity, solar radiation and topographic wetness) using geographically weighted regression. Microtopographic heterogeneity was most strongly related to richness and endemism, but this association was spatially variable at short distances across BIFs: relationships were negative or weakly positive on north-eastern range sections, whereas positive associations became progressively stronger further south and west on central, western and southern sections. Negative solar radiation-plant associations were reduced in these areas, likely because metre-scale surface variation moderates insolation. Topographic wetness-plant associations were negative on BIFs, but positive on the surrounding plains. The presence of fine-scale, geographically variant heterogeneity-diversity relationships in other locations would be difficult to detect if high-resolution environmental data are not used, with the implication that conservation decision-making may be compromised. Given climatic warming predicted for south-western Australia and other regions globally, a similar approach to that applied here can contribute to conservation by identifying locations likely to act as micro-buffers against warming.

1. Introduction The link between environmental heterogeneity and plant species diversity is a long-standing fundamental ecological hypothesis (Ricklefs, 1977; Schimper, 1903) and supported by a raft of studies as reviewed by Tews et al. (2004). It assumes that heterogeneous environments have a greater number of niches within them, which are expected to support a greater range of plant species through reduced competition for the same space (Hutchinson, 1959; Rocchini et al., 2010). These environments may also serve as refuges from harsh extremes, promoting persistence and, potentially, speciation (Stein et al., 2014). These theoretical relationships are expected to hold across spatial scales (Ettema and Wardle, 2002; Tamme et al., 2010). However, at finer spatial scales the extent to which environmental

heterogeneity explains plant diversity is less clear, in part because there has been a paucity of high-resolution environmental data with which to explore this link empirically (Moeslund et al., 2013). Greater testing of the effect of environmental heterogeneity on diversity using field data is therefore needed, particularly at metre scales and in a variety of different environments (Dufour et al., 2006). Airborne LiDAR (light detection and ranging) data can now be acquired at very high resolution (e.g. 1 m) offering extensive coverage (e.g. 1000 km2) of terrain, enabling the link between environmental heterogeneity, in the form of microtopographic heterogeneity, and field data to be explored empirically at the landscape scale. In addition to exploring microtopographic heterogeneity, various other metrics of microtopography (Brubaker et al., 2013) can also be derived from high resolution digital elevation models (DEMs) such as slope, aspect,

∗ Corresponding author. Present address: Climate Change Research Centre, School of Biological Earth and Environmental Sciences, University of New South Wales, Sydney, 2052, Australia. E-mail address: [email protected] (G. Di Virgilio).

https://doi.org/10.1016/j.jaridenv.2018.04.005 Received 18 December 2017; Received in revised form 7 April 2018; Accepted 13 April 2018 0140-1963/ © 2018 Published by Elsevier Ltd.

Please cite this article as: Di Virgilio, G., Journal of Arid Environments (2018), https://doi.org/10.1016/j.jaridenv.2018.04.005

Journal of Arid Environments xxx (xxxx) xxx–xxx

G. Di Virgilio et al.

Fig. 1. (A) The low mountain ranges comprising BIFs in the study region. (B) Many BIFs have a high degree of microtopographic variation. Several plant species grow on surfaces with skeletal soil cover; the examples shown here are (C) Tetratheca aphylla subsp. aphylla (Elaeocarpaceae) and (D) Banksia arborea (Proteaceae).

the region. We hypothesise that environmental heterogeneity of BIFs is likely to be a key factor influencing plant establishment and hence patterns of plant species diversity and endemism. Moreover, given the complex variation of BIF surfaces, the relationship between environmental heterogeneity and patterns of plant biodiversity may vary at short distances across hillsides, i.e. at the scale of proximate microhabitats. We have three aims: 1) identify centres of high plant species richness and endemism on BIF ranges and the surrounding landscape; 2) identify the microtopographic correlates of species richness and endemism; and 3) determine the role of environmental heterogeneity in explaining patterns of plant species diversity.

curvature, topographic wetness and solar radiation. For example, given that moisture is vital for plant function, and the variety of adaptations to moisture stress (e.g. Chaves et al., 2016; Poot and Lambers, 2008), any fine-scale variation in surface relief and aspect that moderates incident solar radiation could cause evapotranspiration variability at metre scales (Bennie et al., 2008). Such microhabitats may be important in plant establishment and survival, particularly in semi-arid environments where surfaces can experience intense insolation and rapid moisture loss. The banded ironstone formations (BIFs) of semi-arid, south-western Australia are island-like outcrops that constitute a highly suitable system to study the impacts of environmental heterogeneity on species diversity (Fig. 1A). They comprise a matrix of fractured rock surfaces, fissures and depressions (Fig. 1B–Mucina and Wardell-Johnson, 2011) and host a diverse flora (Hopper and Gioia, 2004). Many plant species are endemic to individual BIF ranges (Fig. 1C) or are ‘BIF specialists’ (Fig. 1D) in this region (Gibson et al., 2012). Plant species composition varies between adjacent BIF ranges within relatively short distances in this area, i.e. 25–60 km (Butcher et al., 2007; Gibson et al., 2010). Hence, these patterns cannot be attributed to mesoscale climatic gradients, or to geology. While floristic patterns can be partly attributed to topographic variation (Gibson et al., 2010), data resolution has not been sufficient to analyse vegetation patterns in relation to environmental heterogeneity across an entire BIF range empirically. Understanding the potential for environmental heterogeneity to provide buffering from projected climatic warming (Delworth and Zeng, 2014) may inform plant conservation strategies in

2. Material and methods 2.1. Study area The study area is adjacent to the Southwest Australian Floristic Region (SWAFR; Hopper and Gioia, 2004) and is comprised of the entire Helena and Aurora Range (HAR), Mt. Jackson Range, neighbouring, small rock formations and surrounding plains. The areas studied are delineated in red and blue (Fig. 2) and have a total area of 1593 km2. The 52 km2 HAR is the largest range in this area and consists of Precambrian BIF and talus slopes surrounded by outwash and sand plains (Hocking et al., 2007). It comprises a range of low mountains, with the highest peak 702 m above sea level (ASL), approximately 200 m above the surrounding plains. Mt. Jackson (613 m ASL) is comprised of similar 2

Journal of Arid Environments xxx (xxxx) xxx–xxx

G. Di Virgilio et al.

Fig. 2. Study area in south-western Australia encompassing the Helena-Aurora and Mt. Jackson Banded Ironstone Formation Ranges, and surrounding areas. The variation in elevation within the area delimited red (1524 km2) was surveyed by airborne light detection and ranging (LiDAR). The variation in elevation within the area delimited blue (69 km2) was surveyed by aerial radar. Inset: the location of the study area (red) in the Coolgardie Biogeographic Region (dark green) which borders the Southwest Australian Floristic Region (SWAFR; lightgreen). (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

2.3. Estimating plant species richness and endemism

geology. They are located in the Coolgardie Biogeographic Region (Thackway and Cresswell, 1997) and experience a semi-arid Mediterranean climate. Average annual rainfall ranges from 270 to 290 mm with sporadic summer rainfall. Monthly mean maximum temperature ranges from 31.0 to 37.3 °C in the warmest month (January), and mean minimum temperature from 1.4 to 6.9 °C in the coldest month (July). Several conservation-significant plant species (DPaW, 2015) occur on these ranges.

Plant species richness was calculated as the number of vascular plant species per 20 m × 20 m quadrat (Colwell, 2009). Geographic centres of endemism were identified using Biodiverse software, version 0.19 (Laffan et al., 2010). This approach was used to identify local centres of endemism and assess how endemism patterns vary with location (Laffan and Crisp, 2003). It is based on a moving circular window with a radius calibrated as half the range, defined as the distance at which species turnover is no longer spatially dependent (Webster and Oliver (2001) - 7 km in this case - Supplementary Materials, Fig. S1). The moving window was iterated over each quadrat in turn. At each iteration, an endemism score was calculated for a target quadrat by counting the species within this quadrat and weighting each species by its representation across the study area and its number of occurrences captured within a circular neighbourhood around this quadrat. This technique was used to identify centres of endemism using weighted endemism (WE) that was corrected for the strong relationship between WE and species richness (r = 0.79, P < 0.001), i.e. corrected weighted endemism (CWE), as proposed by Crisp et al. (2001). CWE is calculated as WE divided by species richness.

2.2. Plant data The study area was surveyed using 659 (20 m × 20 m) quadrats resulting in the collection of 10,562 vascular plant occurrences, comprising 448 species, 188 genera and 55 families. Quadrats were sampled following the Environmental Protection Authority (EPA) and Department of Parks and Wildlife (DPaW) protocols for a Level 2 flora survey (EPA & DPaW, 2015). Sampling was conducted during spring (October–November) 2012, autumn (May) 2013, spring (September–October) 2013 and autumn (April) 2014, with each quadrat visited at least once in spring. Nomenclature follows Florabase - the database for the Flora of the State of Western Australia managed by the State Land Management Agency, DPaW. Vouchers of all taxa recorded during the surveys were collected and lodged at the Western Australian Herbarium in Perth. 3

Journal of Arid Environments xxx (xxxx) xxx–xxx

G. Di Virgilio et al.

for each variable; and c) local design matrix condition numbers – defined as the ratio of largest to smallest eigenvectors (Di Battista et al., 2016). We aimed to minimise local correlations to < 0.6, local VIFs to < 7.5, and to minimise multicollinearity condition numbers to < 30 (Wheeler and Calder, 2007). This multicollinearity-screening reduced the initial set of predictive variables to three: a) microtopographic heterogeneity; b) TWI; and c) solar radiation. All predictive variables were normalised to a 0–1 scale prior to statistical analyses. Once the spatial patterns inherent in the relationships between the microtopographic predictors and dependent variables had been characterised by the above analyses, we delineated the HAR in into two geographic sections: a central/southern/western section and northeastern section. We then created micro-topographic profiles for each section by classifying their 1 × 1 m grid cells according to their magnitude of annual solar radiation (watt hours per m2) and their aspect (polar degrees). Chi-square tests were applied to these profiles under the null hypothesis that there is no difference between the microtopographic profiles of the two sub-sections of the HAR.

2.4. Topographic data Elevation data for most of the study area (96% - ca. 1524 km2, Fig. 2) were derived from fixed-wing aerial surveys performed by AAM Group. This survey used a Leica ALS60 Airborne laser sensor with a 120.1 kHz pulse rate frequency, a footprint size of 0.25 m, and captured data points with an average 1.2 m point separation. Laser strikes were classified as ground and non-ground returns; only the former were used in this study. These point data had vertical and horizontal accuracies of 0.2 m and 0.4 m respectively and used the GDA94 MGA Zone 50 projected coordinate system. A separate fixed-wing aerial survey was flown over the much smaller, north-western sub-section of the study area (4% - ca. 69 km2, Fig. 2) and measured spot heights at 0.1 m resolution using a radar altimeter with vertical accuracy of 0.3 m. Both datasets were interpolated into digital elevation models (DEMs) using a natural neighbours interpolation algorithm (Beutel et al., 2010) and mosaicked together. 2.5. Predictive variables

3. Results Nine predictive variables were initially derived from the DEM mosaic using ArcGIS 10.3 (ESRI, 2014) and QGIS 2.14.3 (QGIS Development Team, 2017). These were slope (first derivative of elevation), curvature (the slope of the slope), profile curvature (the direction of the maximum slope), plan curvature (perpendicular to the direction of the maximum slope), annual solar radiation (watt hours per m2) for 2014, topographic position as calculated by Weiss (2001), the topographic wetness index (TWI) as calculated by Moore et al. (1993), and microtopographic heterogeneity (or roughness). This involved applying a circular moving window to each cell within the DEM and its adjacent cells within a 10 m radius, and calculating the standard deviation (SD) of elevation for the target cell. We used microtopographic heterogeneity as a proxy for environmental heterogeneity and so these terms are henceforth used interchangeably. All variables were aggregated into the 20 m × 20 m quadrats using the mean to produce a fine-scale average of the physical environment within the quadrat.

3.1. Quantification of species richness and CWE Quadrats with higher plant species richness occur mainly on mountains in the central and south-western sections of the HAR and near Mt. Jackson (Fig. 3A). There were prominent concentrations of high CWE on the low mountains in the centre of the HAR and its southwest (Fig. 3B) and near Mt. Jackson. Small clusters of relatively high endemism are also located near the north-western edge of the HAR (ca. 12 km from the range centre) and at several locations along the eastwest transect on the plains further south. See Supplementary Materials, Figs. S2–3 for per-quadrat numeric values for species richness and CWE. 3.2. Microtopographic determinants of richness and CWE 3.2.1. GLM Plant species richness and CWE were positively associated with microtopographic heterogeneity, (annual) solar radiation and TWI (Table 1A). Microtopographic heterogeneity had the strongest relationship with CWE and species richness (Table 1A). All coefficients were statistically significant (α = 0.01) except for the solar radiationspecies richness association. The GLM regressed against CWE had the best fit to the three predictive variables used in the model (Table 1A).

2.6. Statistical analyses and model selection Species richness and CWE were first regressed against the predictive variables in separate multivariate, generalised linear models (GLM) assuming a Gaussian error distribution and an identity link function (Nelder and Wedderburn, 1972). A key assumption of global regression analyses, namely that relationships between covariates are constant across geographic space (i.e. spatially stationary), is often violated for spatial data (Fotheringham et al., 2002). One solution is to create local estimates of model parameters and return local indicators of goodness of fit, which can be achieved by using Geographically Weighted Regression (GWR) and a roaming kernel that identifies and creates local estimates of model variation (Fotheringham et al., 2002). Thus, we also used GWR to regress richness and CWE against microtopographic predictors in two separate multivariate models using adaptive Gaussian kernels with bandwidths determined by the small sample size corrected Akaike Information Criterion (AICc) bandwidth selection method. All GWR calibrations and analyses were performed using the GWmodel package in R (Gollini et al., 2013). Monte Carlo randomisation (iterations = 5000) was used to test the significance of the spatial variability of each GWR model's local parameter estimates. We tested for global multicollinearity among the metrics of microtopography using ordinary least squares (OLS) linear regression models, excluding variables with a variance inflation factor (VIF) > 7.5. This VIF threshold was selected because it is within the recommended thresholds of 3 and 10 stipulated in Zuur et al. (2010). Local variable redundancy can also be problematic for local correlative approaches, such as GWR. Consequently, we used GWmodel to assess independent variables for a) local correlations amongst variable pairs; b) local VIFs

3.2.2. GWR Relative to the GLMs, the goodness of fit of all models increased significantly (α = 0.01) and AICc scores were lower (Table 1B). The model regressed against CWE performed best, as per the GLM results. Microtopographic heterogeneity had a positive relationship with CWE and richness with median regression coefficients (β1) ranging from 0.045 to 0.206 (Table 1B), whereas TWI had a negative association (Table 1B). Solar radiation had a negative relationship with species richness only. All predictors had a statistically significant relationship with the spatial variation in dependent measures, except for TWIrichness. 3.2.3. Spatial variation of microtopographic correlates The relationship between microtopographic heterogeneity is positive for both species richness and CWE across the HAR and immediately to its south (Fig. 4). Notably, the magnitudes of the regression coefficients for microtopographic heterogeneity regressed against species richness and CWE follow a spatial gradient, i.e. an east-west progression. That is, their magnitudes are smaller on the mountains in the north-eastern HAR, but become progressively larger on mountains in the centre of the range, and then larger again further southward and westward on the range (Fig. 5). Microtopographic heterogeneity also 4

Journal of Arid Environments xxx (xxxx) xxx–xxx

G. Di Virgilio et al.

Fig. 3. (A) Spatial variation in 20 m × 20 m quadrats of species richness estimates on the Helena-Aurora and Mt. Jackson Banded Ironstone Formation Ranges and surrounding area. (B) Spatial variation of corrected weighted endemism (CWE) across the Helena-Aurora and Mt. Jackson Banded Ironstone Formation Ranges.

westward from the HAR. Solar radiation has a negative relationship with species richness and CWE in the north-eastern HAR (Fig. 4). However, as is the case for plant-microtopographic heterogeneity associations, there is an east-

has a moderate, positive relationship with CWE on the lower elevation landscape ca. 12 km to the north-west of the HAR. Although this landscape is comparatively flat, there is a ridge of higher elevation terrain that is part of a chain of disjointed, small outcrops which extend 5

Journal of Arid Environments xxx (xxxx) xxx–xxx

G. Di Virgilio et al.

Table 1 Coefficient (β1) estimates (normalised) generated by regressing predictive variables against species richness and endemism on the Helena-Aurora and Mt. Jackson Banded Ironstone Formation Ranges and surrounding area using A. Generalised Linear Modelling (GLM) and B. Geographically Weighted Regression (GWR). Coefficients for GWR are the median of all local estimates. CWE = corrected weighted endemism. TWI = topographic wetness index. Measure

Microtopographic heterogeneity

TWI

Solar radiation

Accuracy Statistics

β1

β1

β1

R2 Adj.

AICc

∗∗

∗

A

Species Richness CWE

0.354 0.455∗∗

0.127 0.126∗∗

0.066 0.163∗

0.059 0.189

−410.768 −975.368

B

Species Richness CWE

0.045∗ 0.206∗∗

−0.223 −0.120∗∗

−0.065∗∗ 0.039∗∗

0.328 0.624

−593.207 −1436.79

**p < 0.001, *p < 0.01.

underlie the spatial gradients described above. The mountains in the centre and south-western HAR (i.e. those showing positive CWE-microtopographic heterogeneity associations) have fewer 1 m2 cells exposed to intense solar radiation than the mountains in the north-eastern HAR, which is where CWE-microtopographic heterogeneity associations were negative (Fig. 6A). Similarly, a greater number of cells have a southern or south-western orientation on mountains in the centre and south-western HAR (Fig. 6B), whereas the north-eastern areas have more 1 m2 cells orientated northwards (i.e. equator-ward). A statistically significant difference was found between the orientations of the 1 m2 cells on the central-western-southern subsection versus the northeastern section: χ2 (1, N = 2650) = 13, p < 0.001.

west progression in the magnitudes of regression coefficients for solar radiation regressed against both response variables. These coefficients become stronger and positive further southward and westward on the HAR, though the pattern is less pronounced. TWI has a negative relationship with both measures, particularly on the HAR and Mt. Jackson (Fig. 4). These areas show strong, negative plant-TWI relationships. However, the negative sign of TWI-plant relationships is also spatially variable, for instance, being less pronounced towards the periphery of the HAR, especially in the area to its south. TWI regression coefficients are also negative on the low-elevation ridges 12 km to the north-west of the HAR. The surfaces of the different HAR summits show geographic differences in their microtopographic characteristics which may partially

Fig. 4. Spatial variation in the magnitude of regression coefficients (β1) for predictors regressed against species richness and corrected weighted endemism (CWE) in the Helena and Aurora and Mt. Jackson Banded Ironstone Formation Ranges, and the surrounding area. 6

Journal of Arid Environments xxx (xxxx) xxx–xxx

G. Di Virgilio et al.

Fig. 5. Spatial variation in the magnitude of regression coefficients (β1) for microtopographic heterogeneity regressed against plant species richness in the Helena and Aurora Banded Ironstone Formation Range and surrounding plains.

4. Discussion

variation at comparatively small (metre-scale) distances across the BIF ranges. Specifically, there was a fine-scale spatial gradient in microtopographic heterogeneity-plant associations, i.e. an east-west gradient across adjacent hillslopes in the range. Thus, the weak/negative association between microtopographic heterogeneity and the dependent measures on mountains in the north-eastern section of the HAR became positive and progressively stronger on sections further westward and southward along the breadth of the range. Whilst previous studies in this region have documented centres of plant species richness and endemism (e.g. Gibson et al., 2010), these patterns were only inferred to be related to topographic variables or to environmental heterogeneity. Studies conducted elsewhere but focusing specifically on the spatial relationships between microtopographic heterogeneity and plants have found either geographically uniform effects (e.g. Rose and Malanson, 2012), or the absence of any spatial gradient (e.g. Kuntz and Larson, 2006). Moreover, a solar radiation-richness and CWE association that was predominantly negative was less pronounced on the central, southern and western areas of the HAR, likely because micro-sites at these locations provide greater protection from insolation. As a result, these surfaces may experience lower temperatures and aridity, conditions favourable for plant establishment and survival. Ultimately this could be a factor influencing the high species diversity observed in these areas. Grid cells on the central and south-western summits also tended to have higher levels of microtopographic heterogeneity. High microtopographic heterogeneity at local scales suggests that nutrients, moisture and other resources are also spatially heterogeneous; resource

We first summarise the geographic variation in plant species richness and endemism and their spatially variant associations with environmental heterogeneity. We then explore the contribution made by these findings and aspects of our method to the wider debate on the extent and generality of positive environmental heterogeneity-species diversity relationships along with implications for conservation. 4.1. Quantification of plant species richness and CWE We identified geographic centres of high CWE and species richness in the study region. Quadrats with high species richness were dispersed widely across the HAR. Overall, a comparatively small number of quadrats on the HAR did not score highly for this measure. The smaller Mt. Jackson BIF range showed concentrations of high species richness both on and around the range. CWE estimates at Mt. Jackson showed a similar spatial pattern. In contrast to species richness, quadrats with high CWE were clustered on the peaks of the HAR. As expected, there were comparatively fewer high-CWE quadrats on the plains, the main exceptions were lower areas with higher topographic complexity. 4.2. Correlates of species richness and CWE Microtopographic heterogeneity is the major correlate of plant species richness and CWE on the BIF ranges. However, its associations with species richness and endemism showed a systematic spatial 7

Journal of Arid Environments xxx (xxxx) xxx–xxx

G. Di Virgilio et al.

Fig. 6. (A) Classification of 1 × 1 m grid cells according to their incident solar radiation (watt hours per m2) on ironstone summits in the central, southern and western sections of the Helena Aurora Range (HAR), versus summits in the north-eastern HAR. (B) Classification of grid cells according to their aspect (polar degrees) on summits in central, southern and western sections of the HAR (panel B) versus summits in its north-east.

environmental heterogeneity as a buffer against the effects of climate change on cooler (north-facing) slopes. On these BIFs however, we observed similar effects not only at the landform-scale of either north versus south-facing slopes, but also at micro-site scales on both northward and southward facing slopes at different positions along the BIF landform.

availability may thus follow a similar east-west gradient. This supports observations elsewhere that spatial heterogeneity of key resource variables contributes to species richness if these resources are limiting (e.g. Cramer and Verboom, 2017). Metre-scale differences in the surface aspect and relief on BIF ranges may partially explain the fine-scale spatial gradients observed by this study. Hillslopes in the north-eastern HAR have more 1 m2 grid cells with northward (equatorward) orientations, whereas hillslopes in the centre, south and west of the range have more grid cells with southerly (cooler) aspects. Thus, surfaces in the north-eastern HAR appear to provide less protection from solar radiation at metre scales, potentially rendering the micro-climate less favourable for plant establishment and persistence. Several studies have demonstrated that broader scale features such as the aspect of an entire slope can have significant influences on its microclimate and therefore affect vegetation patterns (Bennie et al., 2006). For instance, Maclean et al. (2015) reported that plant species in the northern hemisphere exploited fine-scale

4.3. The nature of fine-scale heterogeneity-diversity relationships The nature and generality of positive environmental heterogeneity–diversity relationships are widely debated. In a meta-analysis of 192 studies worldwide, Stein et al. (2014) found support for the generality of positive heterogeneity–diversity associations from landscape to global extents and highlighted the important role of spatial scale in heterogeneity–richness studies. Recently, Keppel et al. (2016) also found that habitat heterogeneity (as indicated by climatic and topographic variables) was predictive of orchid diversity on islands in the 8

Journal of Arid Environments xxx (xxxx) xxx–xxx

G. Di Virgilio et al.

south-west Pacific. However, none of these studies recorded fine-scale geographically variant heterogeneity-diversity relationships. Our findings thus provide a different perspective on the generality of environmental heterogeneity as a driver of plant diversity: in the case of plant species on the BIF ranges of Coolgardie, the nature and sign of this relationship is not uniform across fine spatial scales, but instead changes systematically over very short distances. Whilst this finding currently applies only to the BIF ranges of this region, we believe it could apply on environmentally heterogeneous rock formations elsewhere in Australia and the world (e.g. granite inselbergs), and perhaps other landforms. We suggest that the use of data of sufficiently high resolution along with spatially explicit analysis by future studies focusing on other global locations could also reveal a spatial variation in environmental heterogeneity-richness relationships. For example, in the study referred to above by Keppel et al. (2016), the resolution of the digital elevation model from which topographic variables were derived was 90 m, which is suited to investigating heterogeneity-diversity relationships across broad spatial extents or at landform-scales, but too coarse to resolve the fine-scale variations in surface configuration which appear to play an important role in the present study. Although high-resolution data like LiDAR have been historically difficult/expensive to obtain and computationally demanding to analyse, these and similar data sets are becoming more readily available, along with increases in computational power.

5. Conclusions Exploring the role of fine-scale environmental heterogeneity in influencing plant diversity patterns has often been hampered by a lack of environmental data of sufficiently high resolution and broad coverage. Here, we used laser-scanned elevation data to derive very high-resolution predictive variables to identify the correlates of plant species diversity and endemism at metre-scales on BIF ranges. Plant species richness and endemism generally increased in microhabitats whose microtopographic variation reduced the intensity of incident solar radiation. We confirmed microtopographic heterogeneity to be a significant, yet spatially variable, correlate of plant species richness and endemism at fine spatial scales on the BIF ranges studied. Studies conducted in other locations globally that investigate the nature of environmental heterogeneity-diversity relationships may not detect the potential presence of fine-scale spatial variability in these relationships if they do not use environmental data of sufficiently high resolution, with the implication that conservation decision-making may be compromised. Given predictions of climatic warming in the SWAFR and other regions globally, a similar approach to that applied here could contribute to the conservation of plant diversity by identifying the finescale areas likely to act as micro-buffers against warming. Acknowledgements This research was funded by Mineral Resources Limited and Curtin University. Photographs were taken by the authors. Botanical surveys were undertaken by ecologia Environment. We thank two anonymous reviewers for their helpful feedback on this research.

4.4. Implications for conservation BIFs have a patchy distribution in the Coolgardie Biogeographic Region, but provide important habitat heterogeneity in this relatively flat, resource-poor landscape, enabling the diversification and maintenance of regional species richness. BIFs are also widespread globally (Jacobi and do Carmo, 2008; Porembski and Barthlott, 2012), as are other rock outcrops like the inselbergs of the Brazilian Atlantic Forest, Guinean Forests of West Africa and South Africa (Porembski et al., 2016). These landforms make substantial contributions to generating and maintaining regional diversity in these biodiversity hotspots. For instance, the BIF ranges in the Brazilian Espinhaço Range and elsewhere in the SWAFR are centres for plant species diversity, as well as providing important ecosystem services (Gibson et al., 2012; Jacobi et al., 2007). However, if not managed appropriately, a variety of processes such as weed invasion, tourism, mining and urbanisation may damage BIF ranges, inselbergs and their biodiversity (Porembski et al., 2016). Improving understanding of the ecology of these landforms is critical to their conservation, but if decision-making does not take account of potential fine-scale, spatial variability in the nature of environmental heterogeneity-diversity relationships, decisions may be based on unreliable assessments. High-resolution environmental data have clear utility for this purpose. For example, climate change is another process that may impact several of these global biodiversity hotspots with increased water stress being projected in the SWAFR (Delworth and Zeng, 2014). However, the spatial gradient in the apparent influence of microtopographic heterogeneity on habitat suitability suggests that the effects of climatic drying in this region will not manifest uniformly across the HAR or potentially other BIF ranges. Given our findings, we envisage that such effects will vary at metrescales across very short distances on BIF surfaces. The impacts of climate change are often emphasised at broad scales. Yet this finding shows that the potential localised effects of climatic changes are pertinent, especially from the viewpoint of conserving endemic flora. Characterising the geographic variability in plant biodiversity on BIFs, inselbergs and other rock formations in relation to high spatial resolution environmental correlates can therefore help identify fine-scale priority areas for plant conservation.

Appendix A. Supplementary data Supplementary data related to this article can be found at http://dx. doi.org/10.1016/j.jaridenv.2018.04.005. References Bennie, J., Hill, M.O., Baxter, R., Huntley, B., 2006. Influence of slope and aspect on longterm vegetation change in British chalk grasslands. J. Ecol. 94, 355–368. Bennie, J., Huntley, B., Wiltshire, A., Hill, M.O., Baxter, R., 2008. Slope, aspect and climate: spatially explicit and implicit models of topographic microclimate in chalk grassland. Ecol. Model. 216, 47–59. Beutel, A., Molhave, T., Agarwal, P.K., 2010. Natural neighbor interpolation based grid DEM construction using a GPU. In: Proceedings of the 18th SIGSPATIAL International Conference on Advances in Geographic Information Systems. ACM, San Jose, California, pp. 172–181. Brubaker, K.M., Myers, W.L., Drohan, P.J., Miller, D.A., Boyer, E.W., 2013. The use of LiDAR terrain data in characterizing surface roughness and microtopography. Appl. Environ. Soil Sci. 2013, 13. Butcher, R., Byrne, M., Crayn, D.M., 2007. Evidence for convergent evolution among phylogenetically distant rare species of Tetratheca (Elaeocarpaceae, formerly Tremandraceae) from Western Australia. Aust. Syst. Bot. 20, 126–138. Chaves, M.M., Costa, J.M., Zarrouk, O., Pinheiro, C., Lopes, C.M., Pereira, J.S., 2016. Controlling stomatal aperture in semi-arid regions-The dilemma of saving water or being cool? Plant Sci. 251, 54–64. Colwell, R.K., 2009. Biodiversity: concepts, patterns and measurement. In: Levin, S.A. (Ed.), The Princeton Guide to Ecology. Princeton University Press, Princeton, pp. 257–263. Cramer, M.D., Verboom, G.A., 2017. Measures of biologically relevant environmental heterogeneity improve prediction of regional plant species richness. J. Biogeogr. 44, 579–591. Crisp, M.D., Laffan, S., Linder, H.P., Monro, A., 2001. Endemism in the australian flora. J. Biogeogr. 28, 183–198. Delworth, T.L., Zeng, F., 2014. Regional rainfall decline in Australia attributed to anthropogenic greenhouse gases and ozone levels. Nat. Geosci. 7, 583–587. Di Battista, T., Moreno, E., Racugno, W., 2016. Topics on Methodological and Applied Statistical Inference. Springer International Publishing, Switzerland. DPaW, 2015. Threatened Species and Communities. Department of Parks and Wildlife, Perth. Australia. Dufour, A., Gadallah, F., Wagner, H.H., Guisan, A., Buttler, A., 2006. Plant species richness and environmental heterogeneity in a mountain landscape: effects of variability and spatial configuration. Ecography 29, 573–584. EPA and DPaW, 2015. Technical Guide – Terrestrial Flora and Vegetation Surveys for Environmental Impact Assessment. Environmental Protection Authority and

9

Journal of Arid Environments xxx (xxxx) xxx–xxx

G. Di Virgilio et al.

in forested landscapes. J. Hydrol. 150, 717–747. Mucina, L., Wardell-Johnson, G.W., 2011. Landscape age and soil fertility, climatic stability, and fire regime predictability: beyond the OCBIL framework. Plant Soil 341, 1–23. Nelder, J.A., Wedderburn, R.W.M., 1972. Generalized linear models. J. Roy. Stat. Soc. 135, 370–384. Poot, P., Lambers, H., 2008. Shallow-soil endemics: adaptive advantages and constraints of a specialized root-system morphology. New Phytol. 178, 371–381. Porembski, S., Barthlott, W., 2012. Inselbergs: Biotic Diversity of Isolated Rock Outcrops in Tropical and Temperate Regions. Springer Berlin Heidelberg. Porembski, S., Silveira, F.A.O., Fiedler, P.L., Watve, A., Rabarimanarivo, M., Kouame, F., Hopper, S.D., 2016. Worldwide destruction of inselbergs and related rock outcrops threatens a unique ecosystem. Biodivers. Conserv. 25, 2827–2830. QGIS Development Team, 2017. Open source geospatial foundation project. https:// www.qgis.org, Accessed date: 12 March 2017. Ricklefs, R.E., 1977. Environmental heterogeneity and plant species diversity: a hypothesis. Am. Nat. 111, 376–381. Rocchini, D., Balkenhol, N., Carter, G.A., Foody, G.M., Gillespie, T.W., He, K.S., Kark, S., Levin, N., Lucas, K., Luoto, M., Nagendra, H., Oldeland, J., Ricotta, C., Southworth, J., Neteler, M., 2010. Remotely sensed spectral heterogeneity as a proxy of species diversity: recent advances and open challenges. Ecol. Inf. 5, 318–329. Rose, J.P., Malanson, G.P., 2012. Microtopographic heterogeneity constrain alpine plant diversity, Glacier National Park, MT. Plant Ecol. 213, 955–965. Schimper, A.F.W., 1903. Plant Geography upon a Physiological Basis. English translation by W.R. Fisher. Clarendon Press, Oxford. Stein, A., Gerstner, K., Kreft, H., 2014. Environmental heterogeneity as a universal driver of species richness across taxa, biomes and spatial scales. Ecol. Lett. 17, 866–880. Tamme, R., Hiiesalu, I., Laanisto, L., Szava-Kovats, R., Partel, M., 2010. Environmental heterogeneity, species diversity and co-existence at different spatial scales. J. Veg. Sci. 21, 796–801. Tews, J., Brose, U., Grimm, V., Tielbörger, K., Wichmann, M.C., Schwager, M., Jeltsch, F., 2004. Animal species diversity driven by habitat heterogeneity/diversity: the importance of keystone structures. J. Biogeogr. 31, 79–92. Thackway, R., Cresswell, I.D., 1997. A bioregional framework for planning the national system of protected areas in Australia. Nat. Area J. 17, 241–247. Webster, R., Oliver, M.A., 2001. Geostatistics for Environmental Scientists. Wiley, Hoboken, New Jersey. Weiss, A.D., 2001. Topographic position and landforms analysis. In: Poster Presentation. ESRI Users Conference, San Diego, CA. Wheeler, D., Calder, C., 2007. An assessment of coefficient accuracy in linear regression models with spatially varying coefficients. J. Geogr. Syst. 9, 145–166. Zuur, A.F., Ieno, E.N., Elphick, C.S., 2010. A protocol for data exploration to avoid common statistical problems. Meth. Ecol. Evol. 1, 3–14.

Department of Parks and Wildlife, Perth, Western Australia. ESRI, 2014. ArcGIS Desktop: Release 10.3. Environmental Systems Research Institute, Redlands, CA. Ettema, C.H., Wardle, D.A., 2002. Spatial soil ecology. Trends Ecol. Evol. 17, 177–183. Fotheringham, A.S., Brunsdon, C., Charlton, M., 2002. Geographically Weighted Regression: the Analysis of Apatially Varying Relationships. Wiley, Hoboken, New Jersey. Gibson, N., Meissner, R., Markey, A.S., Thompson, W.A., 2012. Patterns of plant diversity in ironstone ranges in arid south western Australia. J. Arid Environ. 77, 25–31. Gibson, N., Yates, C., Dillon, R., 2010. Plant communities of the ironstone ranges of South Western Australia: hotspots for plant diversity and mineral deposits. Biodivers. Conserv. 19, 3951–3962. Gollini, I., Lu, B., Charlton, M., Brunsdon, C., Harris, P., 2013. GWmodel: an R package for exploring spatial heterogeneity using geographically weighted models. J. Stat. Software 63, 1–50. Hocking, R.M., Langford, R.L., Thorne, A.M., Sanders, A.J., Morris, P.A., Strong, C.A., Gozzard, J.R., 2007. A Classification System For Regolith In Western Australia. Geological Survey of Western Australia. Hopper, S.D., Gioia, P., 2004. The Southwest Australian Floristic Region: evolution and conservation of a global hot spot of biodiversity. Annu. Rev. Ecol. Evol. Systemat. 35, 623–650. Hutchinson, G.E., 1959. Homage to santa-rosalia or why are there so many kinds of animals. Am. Nat. 93, 145–159. Jacobi, C.M., do Carmo, F.F., 2008. The contribution of ironstone outcrops to plant diversity in the Iron Quadrangle, a threatened Brazilian landscape. Ambio 37, 324–326. Jacobi, C.M., do Carmo, F.F., Vincent, R.C., Stehmann, J.R., 2007. Plant communities on ironstone outcrops: a diverse and endangered Brazilian ecosystem. Biodivers. Conserv. 16, 2185–2200. Keppel, G., Gillespie, T.W., Ormerod, P., Fricker, G.A., 2016. Habitat diversity predicts orchid diversity in the tropical south-west Pacific. J. Biogeogr. 43, 2332–2342. Kuntz, K.L., Larson, D.W., 2006. Microtopographic control of vascular plant, bryophyte and lichen communities on cliff faces. Plant Ecol. 185, 239–253. Laffan, S.W., Crisp, M.D., 2003. Assessing endemism at multiple spatial scales, with an example from the Australian vascular flora. J. Biogeogr. 30, 511–520. Laffan, S.W., Lubarsky, E., Rosauer, D.F., 2010. Biodiverse, a tool for the spatial analysis of biological and related diversity. Ecography 33, 643–647. Maclean, I.M.D., Hopkins, J.J., Bennie, J., Lawson, C.R., Wilson, R.J., 2015. Microclimates buffer the responses of plant communities to climate change. Global Ecol. Biogeogr. 24, 1340–1350. Moeslund, J.E., Arge, L., Bøcher, P.K., Dalgaard, T., Svenning, J.C., 2013. Topography as a driver of local terrestrial vascular plant diversity patterns. Nord. J. Bot. 31, 129–144. Moore, I.D., Norton, T.W., Williams, J.E., 1993. Modelling environmental heterogeneity

10