Abstract
The inshore reefs of the Great Barrier Reef (GBR) have undergone significant declines in water quality following European settlement (approx. 1870 AD). However, direct evidence of impacts on coral assemblages is limited by a lack of historical baselines prior to the onset of modern monitoring programmes in the early 1980s. Through palaeoecological reconstructions, we report a previously undocumented historical collapse of Acropora assemblages at Pelorus Island (central GBR). High-precision U-series dating of dead Acropora fragments indicates that this collapse occurred between 1920 and 1955, with few dates obtained after 1980. Prior to this event, our results indicate remarkable long-term stability in coral community structure over centennial scales. We suggest that chronic increases in sediment flux and nutrient loading following European settlement acted as the ultimate cause for the lack of recovery of Acropora assemblages following a series of acute disturbance events (SST anomalies, cyclones and flood events). Evidence for major degradation in reef condition owing to human impacts prior to modern ecological surveys indicates that current monitoring of inshore reefs on the GBR may be predicated on a significantly shifted baseline.
1. Introduction
Marine ecosystems are showing evidence of global decline, resulting in long-term losses of abundance, diversity and habitat structure [1,2]. Humans are known to alter both the temporal and spatial scales of natural disturbance regimes, impacting upon the potential of marine ecosystems to recover following perturbation [3,4]. Trajectories of decline have been observed in coral reefs throughout the Caribbean and Indo-Pacific, attributable to the synergistic effects of anthropogenic disturbances [5,6]. In the Caribbean region, palaeoecological reconstructions of coral reefs have highlighted the collapse of coral communities at a regional level, which is unprecedented within the Holocene [7,8] and Pleistocene [9,10] records. Despite clear evidence of anthropogenic impacts on Indo-Pacific reefs [5], there is a general lack of historical data or palaeoecological baselines, particularly from the Great Barrier Reef (GBR).
Recently, several authors have highlighted the importance of the ‘shifting baseline syndrome’ [2,11], whereby our current perception of ecosystem health is biased by a lack of long-term baselines prior to anthropogenic impacts. This phenomenon strongly implies that no reef can be considered ‘pristine’ [2]. Using historical evidence, Pandolfi et al. [1] argued that altered anthropogenic disturbance regimes resulted in regional decline of the GBR long before the onset of modern monitoring programmes (approx. 1980), and prior to more recent threats such as ocean acidification and coral bleaching. Such studies provide an invaluable starting point for reversing long-term trajectories of decline by placing marine ecosystems in context of their historical baselines [12].
Following European settlement of the Queensland coastline in the late nineteenth century, extensive catchment clearing for livestock and agriculture (e.g. sugarcane) has resulted in substantial increases in sediment, nutrients and herbicides to the inshore GBR [11,13,14]. Since the onset of monitoring of the inshore GBR in the early 1980s [15], declines in coral cover [11,16–19], local removal of coral species [20] and persistent phase shifts to macroalgal-dominated ecosystems [21,22] have been reported across the inshore reefs. This decline has largely been attributed to altered disturbance regimes associated with coral bleaching events in the last few decades [17,23]. However, few historical baselines of coral assemblages exist prior to the 1980s [11], and the relationship between water quality and decline of the inshore GBR remains a controversy [11,19,24–27].
We present the results of palaeoecological reconstructions of coral assemblages from Pelorus Island, an inshore reef from the Palm Islands region of the central GBR. We surveyed three sites along the leeward side of the island (figure 1). To examine temporal changes in coral assemblages, we compared the community structure of life assemblages (living scleractinian corals) and death assemblages (dead coral fragments directly adjacent to modern corals), an approach commonly used to identify signals of anthropogenic disturbance and shifts in benthic community structure in the absence of historical baselines [28,29]. We obtained high-precision U-series dates [30] from dead coral samples within death assemblages to estimate the timing of mortality and to explore the age structure of the death assemblages. Finally, we extracted core samples from the reef matrix and quantified long-term (centennial to millennial) trajectories in coral community structure to determine historical baselines of coral assemblages. Our results imply a shifting baseline and a previously undetected historical collapse in coral communities coinciding with increased sediment and nutrient loading following European settlement of the Queensland coastline.
Map of the Palm Islands region (central Great barrier Reef) and study sites at Pelorus Island (green boundaries indicate land, yellow boundaries indicate reef outlines).
2. Material and methods
Pelorus Island (18°33 S, 146°29 E) is situated in the Palm Islands region, central GBR (figure 1) between the Dry Tropics and Wet Tropics regions, and is surrounded by Holocene fringing reefs. During peak flood years, fine sediment discharge from the Burdekin River (approx. 165 km south of the study site) reaches the fringing reefs of the Palm Islands [13,31], in addition to flood plumes from the Herbert River to the north, and adjacent local rivers [31].
(a) Comparisons of life and death assemblages
We surveyed three leeward reef sites at Pelorus Island (figure 1). At each site, four belt transects (1 m width, 20 m length) were surveyed on SCUBA at 4–6 m depth. Life assemblages were defined as living scleractinian corals, and death assemblages defined as surficial dead corals adjacent to modern life assemblages [28]. Transects were photographed at intervals using a 1 m2 quadrat (n = 20, 20 m2 per transect) and the per cent cover of life and death assemblages determined using Coral Point Count with Excel extensions (CPCe [32]). The taxonomic structure of life and death assemblages was determined at the genus level and compared between assemblages and sites using hierarchical single linkage cluster analysis and a three-factor permutation multivariate analysis of variance (PERMANOVA [33]) repeated measures model based on Bray–Curtis similarities, where assemblage (life and death assemblages as repeated measures) and site were considered fixed factors, and transect as a random factor. Growth forms of Acropora were defined following Wallace [34]. To provide an estimate of morphological differences between living and dead Acropora, 10 live colonies and 10 dead fragments of Acropora were selected at random from each 20 m transect, and the primary growth form and branch thickness (measured by the broadest diameter of the branch) was determined (total n = 40 per site).
(b) Age structure of death assemblages
Surficial death assemblages were collected by hand on SCUBA at eight points (n = ∼40 fragments per point) at random across a single 20 m transect at each site. From these collections, fragments of the dominant genus/growth form present in death assemblages were identified, and samples were selected for U-series dating haphazardly (i.e. with no bias for taphonomic state of fragments). Sample fragments were sectioned laterally, and a sub-sample (2–3 g) of skeleton was taken from the cleanest section (i.e. unaffected by internal bioerosion) in closest proximity to the growth margin. Approximately 1 g of carefully cleaned material from each sub-sample was used for thermal ionization mass spectrometry (TIMS) and multi-collector inductively coupled plasma mass spectrometer (MC-ICPMS) U-series dating (see the electronic supplementary information for detailed methodology). Ages are reported with 2σ error bars, which represent two standard deviations from the mean true age. Averages (Tukey's biweighted means [35]) and relative probability distributions of U-series ages were determined using Isoplot v. 3.0 [36].
(c) Long-term trajectories in coral community structure
Four cores were extracted from the reef matrix at random along a single 20 m transect at each site (5 m water depth, total n = 12) using a standard percussion coring method (10 cm wide aluminium pipe, 2 m length). Cores were sectioned lengthways using a circular saw and imaged using a multi-slice computed axial tomography (CAT) scanning instrument (Lightspeed VCT, General Electric Healthcare, Milwaukee, WI, USA, 625 μm slice thickness at 400 mA and 120 kV). Longitudinal sections of CAT scans were imported into Adobe Illustrator CS3 (Adobe systems, San Jose, CA, USA), and the outlines of each coral genus were traced to form taxonomic core logs.
(d) Palaeoenvironmental baselines
To understand historical environmental baselines, we combined available continuous palaeoenvironmental records from a range of published and sclerochronological databases within the Palm Islands region, including cyclone tracks (1906–present [37]), HADISST 1.1 sea surface temperature (1870–present [38]), observed coral bleaching events (approx. 1980–present [17,23]), Pacific Decadal Oscillation records (PDO, 1900–2000 [39]), Burdekin River discharge (1921–1999 [13]) and Ba/Ca proxies of sediment flux from the Burdekin River derived from coral cores (1758–1998 [13]). No historical baselines exist for other biotic disturbance events such as disease or crown-of-thorns starfish outbreaks within the Palm Islands region prior to the onset of monitoring programmes in 1987.
3. Results
(a) Comparisons of life and death assemblages
Life assemblages at sites A and C were composed of isolated colonies of massive Porites, and individual Montipora and Acropora colonies (figure 2a). Both sites were characterized by low coral cover (4.4±1%, and 3.5±1%, respectively, see the electronic supplementary material, table S1) compared with the GBR average (21.7% [19]) and other inshore reefs within the region (26.9% [21]). In contrast to sites A and C, life assemblages at site B were dominated by Pavona, Millepora and Echinopora (figure 2a), characteristic of modern inshore reefs in the region [15], and coral cover was considerably higher than the average of the inshore GBR (36.3±8%) and 8–10 times higher than sites A and C.
Life and death assemblages at Pelorus Island, Great Barrier Reef. (a) Relative abundance of the six most common coral genera at each site (±s.e.), (b) relative abundance of Acropora growth forms (pooled between sites), (c) branch thickness of Acropora corals (±s.e.).
The community composition of death assemblages at all sites was dominated by a framework of Acropora (53–97% composition, figure 2a). Low densities of Acropora were present in life assemblages across all sites, yet were fundamentally different in growth form (χ2 = 278.7, p < 0.01, figure 2b) and branch thickness (ANOVA, F249, p < 0.01, figure 2c) to those found in death assemblages. Living Acropora assemblages were composed of thinly branched determinate growth forms of Acropora (corymbose and caespitose morphologies, figure 2b), while death assemblages were formed by expanses of thick arborescent Acropora spp. (see the electronic supplementary material, figure S1), which were largely absent from life assemblages (figure 2b).
The relative generic abundance of coral assemblages varied significantly among sites (PERMANOVA, F2 = 2.736, p < 0.05). Hierarchical cluster analysis of life and death assemblages indicated that death assemblages across all sites formed a single cluster at the 50 per cent similarity level, while life assemblages formed distinct clusters largely separate from the death assemblages (figure 3). Post hoc pairwise tests revealed significant differences between sites A and B (p < 0.05), and between sites B and C (p < 0.01), yet no significant difference between sites A and C (p > 0.5), consistent with the results from the cluster analysis (figure 3). The discordance between life and death assemblages was significant across all sites (PERMANOVA, F1 = 25.367, p < 0.001), yet no interaction was observed between the site and assemblage, indicating that the difference between life and death assemblages was consistent across all sites (table 1).
PERMANOVA results for differences of coral community structure between sites (A,B,C) and assemblages (life and death assemblages). SS = sum of squares, MS = mean square.
Single linkage cluster analysis of life and death assemblages at transect level across sites A–C. Nearest neighbour (single linkage) clustering indicates a clear difference between life and death assemblages. Six separate clusters (shaded) are evident at the 50% similarity level, with more than 90% of death assemblages within sites clustering in a single group (open circle, life assemblage; filled circle, death assemblage).
(b) Age structure of death assemblages
A total of 48 high-precision U-series dates were obtained from dead arborescent Acropora branches (sites A, B and C) and a further 15 U-series dates from dead Pavona fragments at site B (figure 4a). U-series calendar ages for all samples ranged between 1778±10 and 2007±1 (table 2), demonstrating the potential for considerable analytical precision obtained by our method of dating young corals (see the electronic supplementary material, table S2). Relative probability distributions of Acropora ages (figure 4b) revealed distinct peaks of mortality in the mid-1940s at site A (1945±6) and late 1930s at site C (1938±4), with a less defined peak of mortality occurring at site B (1917±41). The mortality of Acropora across all sites was largely constrained to a single time period, with 79.1 per cent of U-series ages falling between 1920 and 1955 (figure 4a). Few U-series ages obtained from Acropora assemblages occurred after 1980 (8.3%), while the majority of ages obtained from Pavona assemblages occurred after 1980 (80%), and centred around the late 1990s (average 1998±4).
Average ages (AD±2σ errors) for death assemblages and reef matrix cores. Weighted averages calculated from U-series analytical error using Tukey's biweight.
U-series ages from death assemblages at Pelorus Island, Great Barrier Reef (a) U-series ages (AD±2σ errors) from arborescent Acropora (n = 48) and Pavona (n = 15) assemblages (ordered from oldest to youngest age), (b) relative probability curves (Gaussian fit) derived from U-series ages (AD±2σ errors) from 1900 to 2010 and average ages (AD±2σ errors) as determined by the Tukey's biweight method.
(c) Long-term trajectories in coral community structure
Computed axial tomography (CAT) scans of reef matrix cores revealed an Acropora-dominated framework at sites A and C, interspersed with minor genera (mainly Pocillopora and Montipora, figure 5), similar to the composition of surficial arborescent Acropora death assemblages. U-series ages obtained from the base of each core show evidence of accretion of reef matrix since the early sixteenth century (1513±5) at site A, and mid-ninth century (846±8) at site C. Cores similarly extracted from site B show a comparable pattern of Acropora-dominated matrix since the mid-third century (266±12). An abrupt transition from arborescent Acropora assemblages to Pavona assemblages in the upper layers (0–30 cm) was observed in all four cores at site B (figure 5). The U-series ages obtained from the uppermost Acropora assemblages at the Acropora–Pavona transition in the sediment cores (figure 5) place the timing of the underlying Acropora mortality at site B to between 1928 and 1944.
Core logs of fossil coral assemblages at Pelorus Island, Great Barrier Reef. Colours represent individual coral genera within the reef matrix (see the inset key), depth of each core is represented in cm, and U-series ages (AD±2σ errors) from the cores are shown adjacent to where they were taken from each core.
(d) Palaeoenvironmental baselines
Consistent with previous analysis [40], five phases were identified in the PDO record (figure 6): 1905–1915 (cool phase), 1916–1943 (warm phase), 1944–1976 (cool phase), 1977–1998 (warm phase) and 1998–2010 (cool phase). Meteorological records indicate 27 cyclones passed within 100 km of Pelorus Island between 1910 and 2010 (figure 6), including three category 2 (1915, 2000), three category 3 (1940, 1956, 1986) and three category 4 cyclones (1955, 1971, 2006). Annual average SST anomalies (calculated relative to 1900–2010) from HADISST [38] indicate that more than 2.5°C thermal anomalies occurred in 8 out of 110 years (1935, 1962, 1970, 1973, 1977, 1987, 1998 and 2004, figure 6). Hydrographic records from the Burdekin River indicate that outflow varied between approximately 10 × 101 to 27.5 × 107 l per month, with monthly flood peaks exceeding 10 × 107 l per week occurring in 1946, 1950, 1951, 1954, 1955, 1958, 1968, 1974 and 1991 (figure 6). Ba/Ca proxies of sediment flux from the Burdekin River derived from coral cores between 1760 and 1998 have previously identified a significant (5–10-fold) increase in sediment flux following 1870, with large peaks occurring in 1927, 1936, 1968, 1970 and 1988 (figure 6) following periods of drought [13].
Relative probability curves from U-series ages (Acropora = red lines, all sites, Pavona = blue lines, site B) and regional palaeoenvironmental records for cyclone activity (T = tropical storm, 1–5 = tropical cyclone intensity scale), SST anomalies, Pacific Decadal Oscillation (PDO) index (LOWESS-smoothing at t = 10 points), sediment flux (Ba/Ca proxy) and Burdekin River discharge between 1900–2010.
4. Discussion
In the absence of historical baselines, comparisons of life and death assemblages provide a valuable tool for determining ecological change and signals of anthropogenic disturbance in benthic communities [28,29]. Our analysis of assemblages from leeward reef sites at Pelorus Island indicates a historical shift in community structure from once dominant arborescent Acropora assemblages at all sites, to either novel low coral cover assemblages (sites A and C), or in the case of site B, high live coral cover composed of non-Acropora assemblages characteristic of modern inshore reefs in the region [15]. While Acropora was present in our surveys of modern assemblages at low cover, in contrast to thick arborescent (branching) Acropora dominant in death assemblages, life assemblages were largely composed of small individual colonies of caespitose and corymbose colonies of Acropora, typical of early stage succession following disturbance [41].
High-precision U-series dating of historical Acropora death assemblages revealed that the majority of ages (92.2%, n = 48 samples across all sites) occurred prior to the onset of broad-scale surveys of the GBR in the 1980s [15,23,42]. Despite high cover of arborescent Acropora from windward reef flat and slope environments throughout the Palm Islands region (including Pelorus) prior to the 1998 mass bleaching event [43], only a single Acropora age was obtained after 1990 from Pelorus Island. Indeed, the majority of U-series ages (79.1%) obtained from Acropora death assemblages occurred between 1920 and 1955. Critically, this historical mortality occurred prior to the mass bleaching events of 1998 and 2002 [44,45], and prior to the onset of monitoring of the GBR in the early to mid 1980s. The absence of previously dominant arborescent Acropora growth forms in the living coral assemblages, and the low number of recent Acropora U-series ages (above 1980) indicates a lack of recovery of arborescent Acropora populations at Pelorus Island following mortality in the early to mid-twentieth century.
U-series ages obtained from the Pavona death assemblage at site B provide evidence of more recent mortality, with 80 per cent of U-series dates occurring after 1980. In contrast, and similar to results from sites A and C, 75 per cent of U-series ages obtained from Acropora death assemblages at site B occurred prior to 1980. We interpret the timing of mortality at site B to indicate a transition in recent decades from historically dominant arborescent Acropora to modern Pavona assemblages, consistent with the high cover and abundance of Pavona in the life assemblages at site B. The timing of the Pavona transition detected in reef matrix cores places the timing of the preceding Acropora mortality at site B to between 1928 and 1944, which is strikingly consistent with the U-series ages obtained from the Acropora death assemblages. Several Pavona ages clustered closely around 1998 (1996±8, 1997±7, 1998±3, 1998±5, 1999±8), and probably represent mortality associated with the 1998/1998 mass coral bleaching event, which was particularly severe in the Palm Islands region [44].
Mismatches between living and dead assemblages can be used to detect signs of anthropogenic ecological change [28], but are not without caveats. Despite the clear potential for time averaging (e.g. the mixing of temporally distinct cohorts [8]) and taphonomic bias (e.g. differential preservation among coral growth forms [46]) in palaeoecological analysis of coral death assemblages, the clustering of U-series ages obtained from historical Acropora indicates a narrow constraint in the timing of mortality, and suggests that Acropora death assemblages are largely autochthonous (i.e. relatively unaltered by taphonomic alteration, vertical mixing of surface layers and fragment transportation [28,29,46]), as expected in low-energy leeward reef habitats [47].
In the Caribbean region, widespread losses of Acropora in the early 1980s have been followed by transitions to coral assemblages dominated by Agaricia sp. [48] and Porites sp. [29] that are unprecedented in the Holocene and Pleistocene records [9]. Our core records indicate that persistent Acropora-dominated framework existed across all sites at Pelorus Island throughout the late Holocene. The timing of the transition from historically dominant arborescent Acropora to modern Pavona assemblages at site B in the core records (1935±7, 1941±3, 1936±6, figure 5) with one exception (1624±7) are consistent with ages obtained from arborescent Acropora death assemblages, and further supports the systematic collapse of arborescent Acropora across all sites between 1920 and 1955. Importantly, this transition is without precedence over the past 1700 years of record, despite a high frequency of ‘super-cyclones’ [49] and climatic fluctuations [50] during this period. Such an abrupt transition over relatively recent timescales precludes an explanation of natural trajectories of geomorphic decline [51], and strongly implicates extrinsic (i.e. anthropogenic) forcing in the early–mid-twentieth century.
To explore the timing and potential causes of historical coral mortality, we compiled palaeoenvironmental records from the twentieth century for the Palm Islands region. Visual assessment of these records indicates considerable climatic variability (SST, cyclone activity and river discharge) coinciding with U-series ages obtained from Acropora death assemblages (approx. 1920–1955, figure 6). The two decades prior to the Acropora mortality (1902–1921) were characterized by a continuous period of negative SST anomalies associated with a PDO cool phase [40]. The subsequent change to the warm phase of the PDO resulted in an increase in positive SST anomalies (1922–1948), and a period of extreme drought [40,52], which coincides with the onset of Acropora mortality within death assemblages. A shift towards a PDO cool phase in the late 1940s [40] resulted in an increase in the frequency and intensity of cyclones (notably category 4 and 3 cyclones in 1955 and 1956) and higher frequency of high flow events (more than10 × 106 l per week) from the Burdekin River (1946, 1950, 1951, 1954, 1955), coinciding with the later stages of Acropora mortality. Considering the relatively broad range of U-series ages obtained from Acropora assemblages (1920–1955), we suggest that multiple disturbance events (e.g. a period of extended SST anomalies followed by increased cyclone activity and flooding, figure 6) may be implicated as a proximal cause of mortality, rather than a single disturbance event.
In the absence of underlying chronic stressors, coral reefs are capable of rapid recovery from acute disturbance events within relatively short timescales [53]. In peak flood years, the Burdekin River disperses large amounts of fine sediment (105 to 106 tonnes) in freshwater flood plumes, affecting inshore and mid-shelf reefs throughout the Palm Islands region [13,31]. Following European settlement of the Burdekin region in 1862, widespread catchment clearing and introduction of cattle and sheep in large numbers [52], resulted in a 5–10-fold increase in sediment flux from the Burdekin river during flood events from 1870 onwards [13]. In addition to land clearing activities in the late nineteenth and early twentieth century, historical records indicate that agricultural fertilizer usage became widespread throughout adjacent catchments from approximately 1930 onwards [54]. Modelling estimates suggest that in the catchments directly adjacent to the Palm Islands, exports of total suspended sediments, herbicides and nutrient loads have increased between 2.1- to 19.5-fold since European colonization [14], resulting in a significantly altered disturbance regime. We suggest that the increase in sediment flux and nutrient loading to the region following European settlement in the late nineteenth century may have acted as the ultimate cause for the lack of recovery in historical Acropora assemblages in the early-to-mid-twentieth century, following mortality owing to a series of acute disturbance events (SST anomalies, cyclones and flooding).
While palaeoecological reconstructions of near-shore reefal shoals on the central GBR suggests that some coral assemblages may be ‘preadapted’ to high turbidity environments [51], the apparent decline of arborescent Acropora assemblages in the earlier part of the twentieth century at Pelorus Island may suggest a level of sensitivity of sub-tidal inshore fringing reefs of the central GBR to increases in sediment flux and nutrient loading. Arborescent Acropora assemblages were dominant in fore-reef environments throughout the Palm Islands region until the 1998 bleaching event [16,23,43,55], yet by the early 1980s, recruitment rates of Acropora corals on inshore reefs of the Palm Islands were strongly suppressed owing to sub-optimal growth conditions [56], and modern studies on nearby inshore reefs indicate that coral recruitment is reduced under exposure to chronic terrestrial runoff [57]. While many growth forms of Acropora exhibit high rates of recruitment (e.g. caespitose and corymbose), arborescent Acropora exhibit sparse larval recruitment and reproduce predominantly by fragmentation [58], suggesting that they may be particularly susceptible to factors affecting recruitment following disturbance.
Collectively, our results indicate a historical change in coral community structure at Pelorus Reef, and suggest that this shift coincided with European settlement of the Queensland coastline. This implies that current monitoring of the inshore GBR may at some locations be predicated on a substantially shifted baseline [11]. On a global scale, our results are consistent with a recent report from the Caribbean region, where land use changes prior to 1960 were implicated in a significant decline in Acropora corals in near-shore reefs [59]. Yet, in contrast to the region-wide collapse of Acropora in the Caribbean, considering the abundance of Acropora within regions prior to the 1998 bleaching event [23,43], the extent of historical mortality on the inshore GBR is likely to vary on local scales. Finally, our results raise two concerns in the interpretation of modern coral assemblages and baseline assessments of coral cover. Firstly, living coral assemblages at site B had relatively high coral cover compared with the GBR average. Yet, modern surveys of live coral cover would significantly underestimate the previously ‘hidden’ shift in coral community structure at the site, resulting in a shifted baseline. Secondly, considering the disparity among Acropora growth forms between life and death assemblages and the near absence of previously dominant framework building Acropora in the present study, modern surveys should consider the dynamics and recovery of corals within genera [23] in addition to the generic level.
Acknowledgements
We thank Orpheus Island Research Station for their help and support during fieldwork, and S. G. Smithers, C. T. Perry, H. Lescinsky and M. Lybolt for critical discussions. We also thank John Bruno and two anonymous reviewers for greatly improving the quality of the manuscript. This project was partially funded by a Marine and Tropical Science Research Facility (MTSRF) Project 1.1.4 to J.X.Z. T.J.D. and J.M.P., Australian Research Council Centre of Excellence for Coral Reef Studies to J.M.P., Australian Research Council LIEF Project LE0989067 to J.X.Z., J.M.P., Y.F. and others, and the Mia J. Tegner and International Society for Reef Studies awards to G.R.
- Received September 7, 2012.
- Accepted October 17, 2012.
- © 2012 The Author(s) Published by the Royal Society. All rights reserved.
References
