Get 20M+ Full-Text Papers For Less Than $1.50/day. Subscribe now for You or Your Team.

Learn More →

Biogeographical regions and phytogeography of the eucalypts

Biogeographical regions and phytogeography of the eucalypts Introduction The eucalypts, the most abundant group of canopy woody plants in the Australian continent, are comprised of three genera: Angophora (Smith) , Corymbia (K.D. Hill & L.A.S. Johnson) and Eucalyptus (L'Hér.) (Brooker et al ., ). The eucalypts consist of around 800 species (894 taxa) with a high level of taxonomic (Nicolle, , ; Brooker, ; Dorothy et al ., ) and phylogenetic diversity (Brooker & Kleining, ; Ladiges et al ., , ; Ladiges, ). The most recent study of distribution patterns (Gill et al ., ) mapped species richness, endemism and general bioregions for Eucalyptus and Corymbia excluding Angophora . Since then an additional 250–300 species have been named and Corymbia has been elevated to a genus (Hill & Johnson, ; Brooker, ; Brooker et al ., ). There has also been a considerable increase in the availability of digitized herbarium records with geographical coordinates and allied environmental datasets, which can be used to explore species distribution patterns. All these elements combined provide an important opportunity to produce a comprehensive and updated analysis of their phytogeography. Bioregionalizations are an useful approach to the understanding of patterns of biodiversity (Wallace, ). For example, they have been successfully applied to delineate zoogeographical regions of the world (Kreft & Jetz, ; Holt et al ., ). Bioregionalizations provide a way to identify units of ecological, evolutionary and historical relevance, enhancing our understanding of the spatial distribution of a biota (Morrone & Crisci, ). Also, by grouping the bioregions into clusters of meaningful geographical units, relationships among regions can be revealed. This is essential to reconstruct the natural history of a continent. Consequently, these aspects are important for future management of biodiversity and ultimately conservation planning. The Interim Biogeographic Regionalisation of Australia (IBRA; Thackway & Cresswell, ) is a comprehensive bioregionalization that integrates many data types, including expert opinion and political boundaries, making it a challenge to interpret biologically. In contrast, climate centred approaches, based on environmental variables, gross primary productivity and remote sensing data, but no biological factors, were explored by Mackey et al . ( ) and Hutchinson et al . ( ). A consistent taxonomy of bioregionalization frameworks would greatly help the research field (Ebach, ). We use Simpson's beta dissimilarity metric to generate a bioregionalization of the eucalypt phytogeographical regions in Australia and Malesia, and then assess the relationships of the phytogeographical regions to the environment. Specifically our goals were to: (1) map spatial patterns of species richness, species endemism and species turnover, (2) propose a biogeographical regionalization of eucalypts based on species turnover and (3) assess the relationship of the phytogeographical regions to environment. These data were compared with other spatial analyses of the Australian flora and together allow the testing of previous biogeographical classifications. Methods Taxonomic and spatial dataset Angophora, Corymbia and Eucalyptus species records were extracted from the Australian Virtual Herbarium database (AVH; CHAH, ). The point records are given Appendix S5. Australian collections were curated to the accepted taxonomy of the Australian Plant Census (Lepschi & Monro, ). We excluded hybrids and used varieties and subspecies data at the species level to reduce taxonomic uncertainty. Species distributions outside of Australia and Malesia were not included if they were not available in the AVH. Spatial errors were identified, and taxonomic corrections were applied using the ArcMap and Google Refine software, respectively. Records without a geographical location were deleted, as were records located in the ocean or outside continental Australia and the Malesian region. The geographical range of each species was corrected manually according to the species distributions in the Euclid database (Brooker et al ., ) to exclude cultivated or naturalized collections. The geographical coordinates of the remaining 219,799 records were projected into a Lambert's conic conformal coordinate system to avoid the latitudinal biases of geographical coordinate systems, and then aggregated to 100 km × 100 km grid cells (906 in total) using the Biodiverse software, version 0.18 (Laffan et al ., ). This grid cell size was chosen to be consistent with Gill et al . ( ) and with González‐Orozco et al . ( ). Environmental dataset A correlation matrix, available through the spatial portal of the Atlas of Living Australia (ALA; htpp:// www.ala.org.au/ ), was used to select 11 environmental variables, which represented different environmental traits and contained minimal correlation (Table ). We also selected some environmental variables based on previous hypotheses of their effectiveness in predicting the distribution of species in Australia (Austin et al ., ; González‐Orozco et al ., ). The climatic variables are described in Houlder ( ) and Hutchinson et al . ( ), who developed the layers using ANUCLIM 6.1. The soil layers were obtained from the National Land & Water Resources Audit (ALWARA, ; Johnston et al ., ). The spatial resolution of the layers was 1 km ( c . 0.01°). The environmental layers were reprojected into the same Lambert's conic conformal coordinate system as the species data and also aggregated to 100 km × 100 km grid cells using Biodiverse . For each environmental variable, we calculated the mean, standard deviation, coefficient of variation, minimum and maximum values of the set of 1 km grid cells within each 100 km grid cell, again using Biodiverse . Environmental variables used in our analyses Environmental variable Description Annual precipitation Monthly precipitation estimates (mm) Annual mean temperature The mean of the week's maximum and minimum temperature (°C) Annual mean radiation The mean of all the weekly radiation estimates (Mj m −2 day −1 ) Precipitation of coldest quarter Total precipitation over the coldest period of the year (mm) Radiation seasonality Standard deviation of the weekly radiation estimates expressed as a percentage of the annual mean (Mj m −2 day −1 ) Precipitation seasonality Standard deviation of the weekly precipitation estimates expressed as a percentage of the annual mean (mm) Temperature seasonality Standard deviation of the weekly mean temperatures estimates expressed as a percentage of the annual mean (°C) Ridge top flatness Metric of the topographic flatness derived from a 9 arc‐second resolution raster digital elevation model (dimensionless; Gallant & Dowling, ) Rock grain size Lithological property of the bedrocks related to the mean grain size (0–10 units) Percentage of sand Content of sand (%) in the top 30 cm of soil layer estimated from soil maps at a resolution of 1 km (%) Percentage of clay Content of clay (%) in the top 30 cm of soil layer estimated from soil maps at a resolution of 1 km (%) Species richness and endemism Species richness (SR), weighted endemism (WE) and corrected weighted endemism (CWE) were calculated in the Biodiverse software version 0.18 (Crisp et al ., ; Laffan et al ., , ) for each 100 km × 100 km grid cell. CWE is a relative measure of endemism and is essentially a function of range restriction. It can be interpreted as the degree to which ranges of species found in the grid cell are, on average, restricted to that grid cell (Laffan & Crisp, ). Once the species richness and endemism scores were calculated for all grid cells, we defined centres of species richness and endemism by selecting those grid cells with the highest 1% of scores (Orme et al ., ; González‐Orozco et al ., ). Species turnover A matrix of Simpson's beta (β sim ) species turnover was generated for all pairwise grid cell combinations (Tuomisto, ). Simpson's beta was used because it reduces the effect of any species richness imbalance between locations. β sim i , j = 1 − a a + min ( b , c ) . where a refers to the number of species common to cells i and j, b is the number found in cell i but not cell j, and c is the number found in cell j but not cell i. A low β sim value indicates that many taxa are shared between two grid cells (low dissimilarity) and a high β sim means a small number of shared taxa (high dissimilarity). Cluster analysis The β sim pairwise distance matrix was used in an agglomerative cluster analysis to generate a WPGMA (weighted pair‐group method using arithmetic averages) hierarchical cluster in Biodiverse . WPGMA weights the contributions of clusters by the number of terminal nodes (data set cells) they contain, ensuring each cell contributes equally to each merger of which it is a part. Its performance was determined to be as successful as UPGMA in Kreft & Jetz ( ). We implemented a tie breaker approach such that, when more than one pair of clusters had the minimum turnover score and thus could be merged, the algorithm selected the pair that maximized the corrected weighted endemism score in the cluster (Crisp et al ., ; Laffan & Crisp, ). This approach guarantees a stable solution that is replicated each time the analysis is run, as well as increasing the degree of endemism and thus the spatial compactness of the resultant bioregions. We identified the phytogeographical regions from the clusters based on two criteria: (1) a phytogeographical region is preferably represented by a group of contiguous, or near‐contiguous, grid cells, (2) each cluster that represents a phytogeographical region needs to be clearly separated from its children or parent. Relative Environmental Turnover Relative environmental turnover (RET) was applied to understand the relationship of environmental variables to the phytogeographical regions estimated using species turnover, as first tested with the Australian genus Acacia (González‐Orozco et al ., ). Previous studies used the term environmental turnover to explore rates of change of dissimilarity in vertebrates and their relationship to environment depending on the geographical distance (Buckley & Jetz, ). RET differs from other studies because it contains two types of analyses: ordinations of the β sim with the environmental variables and the grid cell‐based Getis‐Ord Gi* hotspot statistic. For the ordination analyses, a non‐metric multidimensional scaling (NMDS) ordination was generated using the β sim values. We used NMDS because it is an unconstrained metric that assumes neither normally distributed variables nor linear relationships between variables. This approach allows us to obtain a dimensionless ordination diagram with axes that can be interpreted in terms of an environmental gradient. More traditional methods such as principal component analyses (PCA) were not applied because they assume the data follow a linear normal distribution. Mean values for each of the grid cells were extracted from the 100 km environmental layers. We used the R statistical software (R Development Core Team, ) to produce the β sim values in a matrix format. We then matched the records between the β sim dissimilarity matrix and the mean per cell values of the environmental datasets (see Appendix S2 for heat maps of the environmental variables). The function ‘metaMDS’ of the vegan package (Oksanen et al ., ) was then used to generate the ordination. The β sim values were overlaid onto the ordination and fitted with the environmental variable matrix using the vector fitting of the envfit function from the vegan package. The environmental variables that best explained the patterns of turnover were then displayed as vectors only for those cases with high predictability ( P < 0.001), assessed using 999 permutations. The stress values of the envfit results were used to estimate the efficiency of the NMDS ordinations. The environmental analyses excluded the Malesian region because most environmental variables were not available at an appropriate resolution for that region. The values of grid cells along the first and second axes of the NMDS ordination of β sim distances were extracted and mapped in ArcMap. The geographical rate of change of each axis was calculated using the slope tool in ArcGIS. Those locations with steep slope values correspond with rapid changes (breaks) in the geographical distributions of turnover. The Getis‐Ord Gi* hotspot statistic (Laffan, ; Külheim et al ., ) was calculated using Biodiverse to assess whether the environmental values within each phytogeographical region were significantly different from those for the Australian continent as a whole, where each region was represented by its set of 100 km × 100 km cells. The Gi* statistic is expressed as a z‐score indicating the degree to which the values of a subset of grid cells, in this case the cells comprising a cluster, are greater or less than the mean of the dataset. Those clusters with Gi* values >2 or <−2 represent sets of cells that have environmental values significantly different from expected ( P < 0.05). Results Species richness and endemism The most species rich and endemic regions of eucalypts were located south of the Tropic of Capricorn. We identified three main centres of species richness (Fig. a). Six of the ten grid cells with the highest species richness scores (86–116 species per grid cell) were located in the south‐east coastal province of the Southwest Floristic Region of Western Australia (location 1 in Fig. a). The remaining four of the ten centres (richness scores from 86 to 92) were located in south‐eastern Australia. The New South Wales and Queensland border (location 3 in Fig. a) has not been previously identified as a centre of eucalypt species richness, although the general region was recognized as a centre of species richness for the Australian flora (Crisp et al ., ). Maps of species richness (a) and endemism (b) for eucalypts in A ustralia and M alesia. Numbers in the maps are referred to in the text. Fourteen centres of high eucalypt endemism were identified (Fig b), with the highest scores (between 0.1 and 0.5) being located in Western Australia and East Timor. Five of these endemism centres are in eastern Australia (6, 9, 10, 11 and 12 in Fig. b) and were not identified by Crisp et al . ( ) as being among the 12 endemic areas of the Australian flora. Phytogeographical regions Seven eucalypt phytogeographical regions are proposed, of which six are in continental Australia and one comprises the islands to its north (Fig. a). There are two small clusters in the Malesian islands: one in East Timor and the second further east including part of Indonesia, East Timor and Papua New Guinea, all of which have endemic eucalypt species. The remainder of the Malesian region clusters with larger Australian regions. One cluster differentiates the south‐eastern Australia region from the rest of the continent (Fig b). (a) Map of the seven eucalypt phytogeographical regions, (b) corresponding cluster dendogram and (c) the biomes of A ustralia according to Crisp et al . ( ). Map regions are coloured based on relationship in the corresponding dendogram. For example the blue regions for a higher level cluster as do the red colours. Figure a contains two green clusters in M alesia. These two clusters subtend the rest of the dendogram (2b) and are shown in black and grey for clarity. The dendogram branch lengths (Table ) for each of the geographical clusters (Fig. ) show Australia is subdivided into six major phytogeographical regions (A–F). The Eremaean north region is subdivided into four subregions, where the Pilbara (a) is most similar to western coastal (b) whereas central (c) is most similar to the Eastern (d) (See [(a,b) (c,d)] in Fig. A). The monsoon phytogeographical region is composed of mainland Australia and Malesian subregions, where the Top End/Malesia subregion (a) is most dissimilar to the rest of the subregions (See [(a) (b,c,d)] in Fig. B). The inland monsoon subregions are divided east‐west in three subregions: east (b), central (c) and west (d). The tropical/subtropical region is an independent cluster related to the northern regions of the monsoon and the Eremaean north (See [(a) (b,c)] in Fig. C). This region has a strong north to south pattern with Cape York (a) to the north and a southern coastal cluster, which can be further separated into Queensland Central (b) and New South Wales south (c) subregions. Values of branch length for clusters of floristic regions and subregions of A ustralian eucalypts Eucalypt phytogeographical regions Geographic subdivisions Branch length to parent (brackets refer to subregions labels in Fig. ) Malesia Top End/Malesia 0.04 Monsoonal Belt 0.08 South‐east (Major split) 0.23 Alpine 0.21 (a) Tasmanian/southern Victoria 0.15 Tasmanian 0.27 (b) South‐eastern mainland 0.03 (c) Eremaean south/south‐west (Major split) 0.01/0.02 Eremaean south: 0.03 Southern MDB 0.14 (a) Nullarbor Plain 0.18 (b) Northern 0.14 (c) Southern 0.14 (d) South‐west: 0.18 Eastern/Central 0.40 (a) Western (Southern) 0.21 (b) Western (Northern) 0.53 (c) Tropical/Subtropical (Major split) 0.03/0.08 Tropical: 0.20 (a) Cape York 0.17 (b) QLD North 0.24 Subtropical: 0.34 (c) NSW South Monsoon/Eremaean north (Major split) 0.06 Monsoon: 0.11 Malesia 0.16 (a) East 0.16 (b) Central 0.06 (c) West 0.12 (d) Eremaean north: 0.04 West 0.06 Western‐ Pilbara 0.13 (a) Western‐ Coastal 0.15 (b) Central eastern 0.10 Central 0.07 (c) Eastern 0.04 (d) Map of the eucalypt phytogeographical regions and their respective subregions. Regions are denoted in upper case letters: A = Eremaean north; B = monsoon; C = Tropical/subtropical; D = south‐west; E = Eremaean south; F = south‐east. Subregions are denoted in the parenthetical notation using lower case (see Table for corresponding names) and are coloured based on relationship in the corresponding dendogram. Eucalypt phytogeographical regions and subregions defined by analysis of S impson's beta (β sim ) species turnover (letters in brackets refer to regions and subregions shown in Fig ) Eucalypt phytogeographical regions Geographic subdivisions Subregions Monsoon (B) Top End/Malesia Top End/Malesia (a) Monsoonal Belt East (b) Central (c) West (d) Eremaean north (A) West Pilbara (a) Coastal (b) Central/eastern Central (c) East (d) Eremaean south (E) Eastern/Central Southern MDB (a) Nullarbor Plain (b) Western South (d) North (c) Tropical/Subtropical (C) Tropical Cape York (a) Subtropical QLD North (b) NSW South (c) South‐east (F) Alpine Alpine (a) Tasmanian/southern Victoria Tasmania (b) South‐eastern Victoria (c) South‐west (D) South‐eastern South‐eastern (a) Western North (c) South (b) The Eremaean south and the south‐west phytogeographical regions form a single cluster separate from the monsoon, Eremaean north and tropical/subtropical regions. While the branch combining the two areas is short, the branches supporting Eremaean south and the south‐west are longer. We recognize these two eucalypt phytogeographical regions based on long nested branches. The south‐west region is divided into south and west subregions (see [(a) (b,c)] in Fig. D). The Eremaean south region is subdivided into two subregions that stretch from east to west (See [(a,b) (c,d)] in Fig. E). The most eastern/central subregion further subdivides into a southern Murray‐Darling Basin subregion (a) and a subregion along the Nullarbor coastal plain (b) whereas western subregions (c–d) of inland and coastal areas connect with the south‐west region. The south‐east region is subdivided into an alpine (a) subregion and a Tasmanian/southern Victorian subregion (See [(a) (b,c)] in Fig. F). The southern portion subdivides into Tasmanian (b) and south‐eastern mainland (c) subregions. Relative environmental turnover (RET) The geographical surface of the first NMDS axis indicates a major geographical break oriented north‐west to east across the continent (Fig. a, see also the red line in Fig. c). This break is aligned with the summer–winter rainfall line defined by Burbidge ( ). The pattern produced by the second axis (Fig. b) matches the division of arid central Australia, which has less eucalypt diversity, and the mesic zone on the coastal regions of eastern Australia. Areas noted as having high slope values are regions with abrupt changes of species turnover and are inferred turnover breaks. For example, the major break we found is detected by NMDS1 (red line in Fig. c), and follows the subregion boundaries (Fig. ). No major pattern is found for NMDS2, but ten small inferred turnover break areas are scattered across the continent (Fig. d). Of these, area 1 is a known barrier for the Wet Tropics, area 2 corresponds to the MacPherson/MacLeay overlap zone between NSW and Queensland east coast, area 4 corresponds with the Murray and Gippsland Basin barriers, and area 10 reflects the known barrier of the Nullarbor Plain of southern Australia. Species turnover of A ustralian eucalypts measured with the non‐metric multidimensional scaling ( NMDS ) method for (a) axis 1 ( NMDS 1) and (b) axis 2 ( NMDS 2). The two right side panels plot the slopes of NMDS 1 (c) and NMDS 2 (d) and the inferred biogeographic breaks (red lines and polygons). A low stress value (0.1016) suggests a good fit of the environmental data to the β sim clusters in the ordination model (Fig. ). With the exception of clay content, all variables were significantly ( P < 0.001) correlated with β sim . For example, the values towards the northern areas of monsoonal Australia correspond with the points located in the top left corner of the plot (indicated by the North arrow). The vectors for annual precipitation and precipitation during winter vectors correlate with species turnover towards the east and north‐west of Australia. Vectors for temperature seasonality, annual mean radiation, topography, sand content and lithology correlate with species turnover towards the west and south‐west of Australia. Vectors for annual mean temperature and annual mean radiation vectors correlate with species turnover towards the western parts of the monsoon region as well as the central Australian deserts located in the eastern part of the Eremaean north region. Vectors for precipitation during the coldest quarter and radiation seasonality were highly correlated with the southern regions. Non‐metric multidimensional scaling (NMDS) of eleven environmental variables fitted to turnover (β sim ) of A ustralian eucalypts. Vector names representing seasonal related variables are bolded while aseasonal and non‐climatic variables are not. The N orth arrow in the top left corner represents the geographic orientation of the NMDS values. The results of the gridded approach, where the effects of 11 environmental variables were tested in each of the eucalypt bioregions, are shown in Table . A value >2 or <−2 indicates that the values of an environmental variable in that floristic region are, collectively, significantly different from the study region as a whole. Overall, climate variables such as precipitation and temperature appear more important for explaining the turnover patterns of the eucalypt phytogeographical regions than are non‐climatic variables such as soil and topography. However, there is some degree of variability from region to region. For example, precipitation seasonality is the main driver for the monsoon region, while precipitation during coldest quarter of the year is the main driver for the south‐eastern region. Temperature seasonality is the main driver for the Eremaean north region, meaning that monthly variation in temperature correlates well with species turnover in the drier and hotter areas north of the Tropic of Capricorn. Interestingly, precipitation seasonality represented as winter rainfall (south of the Tropic of Capricorn) becomes the main driver for the Eremaean south region, meaning that rainfall patterns between June and August play a determinant role. As expected, the main driver of the tropical and subtropical north‐eastern and eastern regions is annual mean precipitation. The species turnover in the temperate regions of south‐east and south‐west Australia is highly correlated with the amount of precipitation during winter months. The results of the gridded approach tested on each of the eucalypt subregions are given in Table S3 of Appendix S3. Gi* statistic scores for the cells comprising the six major phytogeographical regions of A ustralian eucalypts. Bold text signifies statistically significant (α = 0.05). N = number of grid cells per region Monsoon ( N = 127) Eremaean north ( N = 341) Eremaean south ( N = 204) Tropical ( N = 38) Subtropical ( N = 38) South‐east ( N = 65) South‐west ( N = 21) Annual mean radiation 6.46 9.83 −2.47 −0.58 −4.18 −15.16 −7.11 Annual mean temperature 13.75 7.54 −9.18 4.88 −2.84 −15.35 −6.82 Annual mean precipitation 10.06 −11.67 −10.53 11.02 8.80 8.49 2.21 Percentage of clay 0.51 2.35 −1.64 −1.75 1.98 −0.11 −4.28 Precipitation coldest quarter −8.18 −7.84 0.67 −2.97 4.55 16.32 11.93 Precipitation seasonality 16.88 2.59 − 13.73 8.54 −1.73 −10.20 −0.83 Radiation seasonality −15.19 −6.37 12.91 −5.48 −1.27 13.98 4.71 Ridge Top flatness 2.56 0.42 4.80 −2.65 −4.82 −5.35 −1.34 Rock grain size −1.39 −7.45 8.51 −0.46 0.04 0.28 3.63 Percentage of sand 1.63 −0.91 −0.27 1.60 −1.80 −1.14 2.10 Temperature seasonality −12.79 12.39 6.81 −10.16 −3.20 −3.96 −4.36 Discussion Eucalypt phytogeographical regions Our seven phytogeographical regions and 21 subregions contribute significantly to the understanding of contemporary Australian biogeography. The major split of the main two clusters suggests the existence of a north‐south split of the eucalypt phytogeographical regions. The southern area might reflect historical adaptation to habitats affected by higher winter rainfall whereas the northern area might be more influenced by the summer rainfall adapting to high humidity and extreme heat. Within the southern area, the Eremaean South and south‐west share higher similarity than with the south‐east. In contrast, the clusters corresponding to the monsoon and Eremaean north share higher similarity than with tropical/subtropical clusters. The limits of the southern area extend east‐west from Western Australia to central New South Wales. This major pattern could be explained by the effect of regolith carbonate accumulations in the soil (Hill et al ., ; McQueen et al ., ). Such a geochemical difference is a potential explanation for the presence of the major north‐south split of the continent wide phytogeographical regions of the eucalypts. Our proposed eucalypt phytogeographical regions (Fig. ) are in broad agreement with the biomes of Australia (Burbidge, ; Schodde, ; Crisp et al ., ). The main difference between the result of our analysis and those of Burbidge ( ) is that the northern Tropical Cape York region is not differentiated as part of the monsoonal tropics biome, instead it is part of the north‐eastern tropical region (Fig. c). The extent of Burbidge's monsoonal tropics is larger than ours, specifically towards the central eastern region. The Eremaean zone is a single unit in Burbidge's classification while here we divide it into two distinct regions with the Eremaean north clustering with the monsoonal bioregion and not with the Eremaean south. This same relationship was also found in Acacia (González‐Orozco et al ., ). Species endemism is a key concept to explain biogeographical patterns (Parenti & Ebach, ). We found that in six of the seven proposed eucalypt phytogeographical regions there is at least one centre of endemism (Table ). Occurrence of centres of species richness and endemism per phytogeographical regions of A ustralian eucalypts. Numbers denote the number of centres present in each bioregion, a dash (–) denotes no centres Eucalypt phytogeographical regions Species Richness Endemism Monsoon – 1 Eremaean north – 2 Eremaean south – – Tropical – 1 Subtropical 1 3 South–east 1 4 South–west 1 3 Linking eucalypt phytogeographical regions to environment There are many examples of bioregionalizations in Australia, such as the agro‐climatic classification, the environmental domain classification and the primary productivity regionalization of Australia (Hutchinson et al ., ; Mackey et al ., ). These are important frameworks but are strongly biophysical, as opposed to the biologically centred approach used here. The only study comparing species distribution of eucalypts to climate at the continental scale was conducted by Gill et al . ( ). They identified two eucalypt clades adapted to warm climates ( Blakella‐Corymbia and Eudesmia ) and another two composed of mostly of species that are adapted to cool climates. The genus Eucalyptus contains 10 subgenera, but subgena Blakella and Corymbia comprise the subgenera Corymbia . Gill et al . also noted that temperature during the wettest quarter of the year and rainfall patterns overall had a great effect on the species located in the northern tropical regions. Our work identified a predominant north‐east and south‐west geographical pattern, which fits the climatic rationale presented by Burbidge ( ). Burbidge also identified the Tropic of Capricorn as a major biome delineator: higher summer rainfall to the north and winter rainfall to the south. The inconsistencies of environmental drivers between our study and that of Gill et al . ( ) may be due to the fact that we used more species records, a greater number of species and applied a metric that minimizes the imbalance of species richness in representing biotic turnover (see Table ). Gill et al . also used meteorological data from the weather station nearest to each cell, whereas our climate data are derived by spatial interpolation using all surrounding stations. Ten principal differences between the present study and that of Gill et al . ( ) for the analysis of the distribution of species dissimilarity and diversity of A ustralian eucalypts Study parameters Gill et al . ( ) Present study Spatial scale 1.0 × 1.5 degrees (lat × long) 100 × 100 km grid cells Turnover metric Jaccard and Czeckanowski Beta‐Simpson index Number of species 551 798 Taxonomic resolution Species and subgenera (without Angophora ) Eucalyptus , Angophora and Corymbia Statistical technique PCA NMDS Cluster metric UPGMA WPGMA Environmental variables Ten (five temperature and five rainfall variables with no soils) Eleven seasonal and aseasonal variables Terminology Zones Bioregions and subregions Distribution Continental Continental and Malesia Diversity metric used Species richness and absolute endemism Species richness and relative endemism Number of bioregions 11 major zones and 18 subgroups Six main bioregions and 13 subregions Comparing eucalypt phytogeographical regions to major eucalypt communities Our phytogeographical regions closely match with previous broad scale studies on vegetation communities in Australia that were based on expert opinion. Beadle ( ) identified four main Eucalyptus vegetation alliances and eight generic biogeographical regions of eucalypts. The main difference of our study from Beadle's is that he did not define the monsoon region per se and the semi‐arid and arid areas in central Australia are not split into two regions as we propose here. A more recent vegetation classification based on the National Vegetation Information System (NVIS) classification (DEWR, ) defined six major eucalypt vegetation groups, which broadly correspond with our phytogeographical regions. The results of this research support previous findings regarding the location of the main centres of species richness but, interestingly, we identified new centres of endemism. The eucalypt centres of species diversity and endemism tend to be different from other groups of Australian plants and mostly occur south of the Tropic of Capricorn. For example, the biodiversity hotspot of south‐west Australia is an area of high species richness and endemism for both Acacia (González‐Orozco et al ., ) and eucalypts but not for Glycine (González‐Orozco et al ., ) or bryophytes (Stevenson et al ., ). However, the species rich and endemic areas in the south‐west do not overlap but are mostly adjacent to each other suggesting separate environmental niche preferences for the two lineages. Acacia species richness and endemism is most pronounced in a belt delimiting the south‐west for the arid zone while the eucalypt species richness and endemism dominate the coastal regions. There is an area of overlap of species richness and endemism in Acacia and the eucalypts in the broader Stirling ranges — Fitzgerald River region of Western Australia (see area 1 in Fig. a). Overlapping patterns of high species richness, with fewer areas of endemism, are found in south‐eastern Australia for the eucalypts, Acacia , hornworts liverworts and mosses (González‐Orozco et al ., , ; Stevenson et al ., ). This suggests that south‐eastern Australia is an area of high speciation and broader geographical distributions. The Wet Tropics is an area of high species richness and endemism for all these lineages. Bioregionalizations have been conducted on continental (Rueda et al ., ; Linder et al ., ) and global scales (Kreft & Jetz, ; Holt et al ., ). These works provide an alternative approach as they consider a subset of the diversity of groups such as plants, insects and vertebrates within large geographical areas whereas our study focuses on a single fully sampled lineage at a continental scale. Linder et al . ( ) also used β sim and identified seven sub‐Saharan biogeographical regions that were broadly congruent among the plant and vertebrate groups studied. However, Rueda et al . ( ) did not find congruence among lineage‐specific bioregionalizations in Europe. An allied study on Acacia (González‐Orozco et al ., ) revealed similar but not identical phytogeographical patterns to the eucalypts. It is possible that the differing evolutionary histories of individual lineages, when combined in a single analysis, would produce the less defined patterns that were seen by Linder et al . ( ) in sub‐Saharan Africa. More studies are needed to better understand whether bioregionalizations of plants and animal groups are congruent (Linder et al ., ) or incongruent (Udvardy, ; Rueda et al ., ) and what drives the patterns. Our finding that climate and dissimilarity of plant distributions correlate with species turnover confirms previous global studies (Buckley & Jetz, ). For example, Buckley and Jetz ( ) found that high levels of species turnover occur regardless of environmental turnover rates, but environmental turnover provides a lower bound for species turnover in amphibians than for birds. Because each continent has specific topographic and climatic attributes, it would not be expected to find major similarities on the main environmental control of turnover. The patterns of eucalypt diversity identified here provide a foundation for future biogeographical studies and investigation of eucalypt environmental niches. In the eucalypts, the patterns of species richness and endemism are not tightly linked, especially in south‐eastern Australia. Those taxa in the species rich areas have broad ranges, thereby lowering the corrected weighted endemism scores and suggesting multiple dispersal episodes following events of active allopatric speciation. Limited inferences can be made on data based on species distribution without a phylogenetic context. It is probable that combining over 800 species into a single study will obscure important signals that can be determined through phylogenetics. The utilization of phylogenetic diversity (Faith ), phylogenetic endemism (Rosauer et al ., ) and related metrics may better unravel the historical factors and evolutionary relationships behind the organization of biodiversity on the landscape. A thorough understanding of the basis for these patterns, including phylogeny and environmental niche preferences, can inform conservation planning decisions. Although our classification is not based on phylogenetic relationships, we found a few similarities with some phylogenetic focussed studies. Ladiges et al . ( ) reported that species of Corymbia and Eudesmia, analysed using area cladogram techniques, suggest differences between the monsoon tropics and the eastern areas of Australia. Here, we identified similar patterns across the larger eucalypts group. The phylogenetic structure of the eucalypts in these bioregions and at smaller scales could be driven by alternative ecological patterns, such as phylogenetic clustering or over‐dispersion (Webb et al ., ). Potential caveats We demonstrate that bioregionalizations of a continent using species turnover of a large taxon group is a potential method to understand biogeography, as shown in other case studies (see Linder et al ., ). However, despite partitioning, the continent into meaningful biogeographical areas of species assemblages, the phytogeographical regions are not evolution‐based and therefore phylogenetic approaches would benefit future studies (Ferrier et al ., ; Rosauer et al ., ). Sampling bias is a common problem with herbarium data, and under‐sampling is thus a common cause of uncertainty. We explored the potential implication of sampling bias by conducting a redundancy analysis, calculated as the ratio of species records to number of samples per grid cell (Garcillán et al ., ). We found that only 274 of all 906 grid cells analysed had redundancies less than 30% (see grid cells in green and blue shades; Appendix S1). This indicates that 70% of the continent is comparatively well sampled. The poorly sampled areas, which are often remote, tend to correspond with areas of low turnover values. The grid cell size may also have an effect on the turnover patterns (Barton et al ., ). A pilot test (results not shown) that assessed the effect of grid cell size (50 km × 50 km; 25 km × 25 km) on the mapping of the phytogeographical regions found no major spatial difference in terms of number of phytogeographical regions generated. However, we observed that the resolution has a small effect on the boundaries of the bioregions. Conclusion Rainfall plays a key role for northern Australian eucalypt assemblages, whereas temperature and solar radiation are more important in the south and south‐eastern regions, specifically south of the Tropic of Capricorn. Current climate projections suggest warmer temperatures for eucalypts in Australia, particularly the southern populations (Hughes et al ., ). Temperature is a key driver of eucalypt species turnover, which highlights the importance of special care when managing biodiversity around areas located in southern latitudes. Species turnover patterns of eucalypts in the western part of the continent correlate more with non‐climatic factors such as topographic and soil properties. Although the proposed phytogeographical regions are exclusive to eucalypts, our method can be applied to any geographical scale providing sufficient data with the appropriate taxonomic and spatial detail are available. Acknowledgements We would like to thank to the CSIRO referees for useful comments on the manuscript and Andrew Slee for help with distribution and taxonomy issues. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Diversity and Distributions Wiley

Loading next page...
 
/lp/wiley/biogeographical-regions-and-phytogeography-of-the-eucalypts-pRqLt4rnM0

References (74)

Publisher
Wiley
Copyright
Copyright © 2014 John Wiley & Sons Ltd
ISSN
1366-9516
eISSN
1472-4642
DOI
10.1111/ddi.12129
Publisher site
See Article on Publisher Site

Abstract

Introduction The eucalypts, the most abundant group of canopy woody plants in the Australian continent, are comprised of three genera: Angophora (Smith) , Corymbia (K.D. Hill & L.A.S. Johnson) and Eucalyptus (L'Hér.) (Brooker et al ., ). The eucalypts consist of around 800 species (894 taxa) with a high level of taxonomic (Nicolle, , ; Brooker, ; Dorothy et al ., ) and phylogenetic diversity (Brooker & Kleining, ; Ladiges et al ., , ; Ladiges, ). The most recent study of distribution patterns (Gill et al ., ) mapped species richness, endemism and general bioregions for Eucalyptus and Corymbia excluding Angophora . Since then an additional 250–300 species have been named and Corymbia has been elevated to a genus (Hill & Johnson, ; Brooker, ; Brooker et al ., ). There has also been a considerable increase in the availability of digitized herbarium records with geographical coordinates and allied environmental datasets, which can be used to explore species distribution patterns. All these elements combined provide an important opportunity to produce a comprehensive and updated analysis of their phytogeography. Bioregionalizations are an useful approach to the understanding of patterns of biodiversity (Wallace, ). For example, they have been successfully applied to delineate zoogeographical regions of the world (Kreft & Jetz, ; Holt et al ., ). Bioregionalizations provide a way to identify units of ecological, evolutionary and historical relevance, enhancing our understanding of the spatial distribution of a biota (Morrone & Crisci, ). Also, by grouping the bioregions into clusters of meaningful geographical units, relationships among regions can be revealed. This is essential to reconstruct the natural history of a continent. Consequently, these aspects are important for future management of biodiversity and ultimately conservation planning. The Interim Biogeographic Regionalisation of Australia (IBRA; Thackway & Cresswell, ) is a comprehensive bioregionalization that integrates many data types, including expert opinion and political boundaries, making it a challenge to interpret biologically. In contrast, climate centred approaches, based on environmental variables, gross primary productivity and remote sensing data, but no biological factors, were explored by Mackey et al . ( ) and Hutchinson et al . ( ). A consistent taxonomy of bioregionalization frameworks would greatly help the research field (Ebach, ). We use Simpson's beta dissimilarity metric to generate a bioregionalization of the eucalypt phytogeographical regions in Australia and Malesia, and then assess the relationships of the phytogeographical regions to the environment. Specifically our goals were to: (1) map spatial patterns of species richness, species endemism and species turnover, (2) propose a biogeographical regionalization of eucalypts based on species turnover and (3) assess the relationship of the phytogeographical regions to environment. These data were compared with other spatial analyses of the Australian flora and together allow the testing of previous biogeographical classifications. Methods Taxonomic and spatial dataset Angophora, Corymbia and Eucalyptus species records were extracted from the Australian Virtual Herbarium database (AVH; CHAH, ). The point records are given Appendix S5. Australian collections were curated to the accepted taxonomy of the Australian Plant Census (Lepschi & Monro, ). We excluded hybrids and used varieties and subspecies data at the species level to reduce taxonomic uncertainty. Species distributions outside of Australia and Malesia were not included if they were not available in the AVH. Spatial errors were identified, and taxonomic corrections were applied using the ArcMap and Google Refine software, respectively. Records without a geographical location were deleted, as were records located in the ocean or outside continental Australia and the Malesian region. The geographical range of each species was corrected manually according to the species distributions in the Euclid database (Brooker et al ., ) to exclude cultivated or naturalized collections. The geographical coordinates of the remaining 219,799 records were projected into a Lambert's conic conformal coordinate system to avoid the latitudinal biases of geographical coordinate systems, and then aggregated to 100 km × 100 km grid cells (906 in total) using the Biodiverse software, version 0.18 (Laffan et al ., ). This grid cell size was chosen to be consistent with Gill et al . ( ) and with González‐Orozco et al . ( ). Environmental dataset A correlation matrix, available through the spatial portal of the Atlas of Living Australia (ALA; htpp:// www.ala.org.au/ ), was used to select 11 environmental variables, which represented different environmental traits and contained minimal correlation (Table ). We also selected some environmental variables based on previous hypotheses of their effectiveness in predicting the distribution of species in Australia (Austin et al ., ; González‐Orozco et al ., ). The climatic variables are described in Houlder ( ) and Hutchinson et al . ( ), who developed the layers using ANUCLIM 6.1. The soil layers were obtained from the National Land & Water Resources Audit (ALWARA, ; Johnston et al ., ). The spatial resolution of the layers was 1 km ( c . 0.01°). The environmental layers were reprojected into the same Lambert's conic conformal coordinate system as the species data and also aggregated to 100 km × 100 km grid cells using Biodiverse . For each environmental variable, we calculated the mean, standard deviation, coefficient of variation, minimum and maximum values of the set of 1 km grid cells within each 100 km grid cell, again using Biodiverse . Environmental variables used in our analyses Environmental variable Description Annual precipitation Monthly precipitation estimates (mm) Annual mean temperature The mean of the week's maximum and minimum temperature (°C) Annual mean radiation The mean of all the weekly radiation estimates (Mj m −2 day −1 ) Precipitation of coldest quarter Total precipitation over the coldest period of the year (mm) Radiation seasonality Standard deviation of the weekly radiation estimates expressed as a percentage of the annual mean (Mj m −2 day −1 ) Precipitation seasonality Standard deviation of the weekly precipitation estimates expressed as a percentage of the annual mean (mm) Temperature seasonality Standard deviation of the weekly mean temperatures estimates expressed as a percentage of the annual mean (°C) Ridge top flatness Metric of the topographic flatness derived from a 9 arc‐second resolution raster digital elevation model (dimensionless; Gallant & Dowling, ) Rock grain size Lithological property of the bedrocks related to the mean grain size (0–10 units) Percentage of sand Content of sand (%) in the top 30 cm of soil layer estimated from soil maps at a resolution of 1 km (%) Percentage of clay Content of clay (%) in the top 30 cm of soil layer estimated from soil maps at a resolution of 1 km (%) Species richness and endemism Species richness (SR), weighted endemism (WE) and corrected weighted endemism (CWE) were calculated in the Biodiverse software version 0.18 (Crisp et al ., ; Laffan et al ., , ) for each 100 km × 100 km grid cell. CWE is a relative measure of endemism and is essentially a function of range restriction. It can be interpreted as the degree to which ranges of species found in the grid cell are, on average, restricted to that grid cell (Laffan & Crisp, ). Once the species richness and endemism scores were calculated for all grid cells, we defined centres of species richness and endemism by selecting those grid cells with the highest 1% of scores (Orme et al ., ; González‐Orozco et al ., ). Species turnover A matrix of Simpson's beta (β sim ) species turnover was generated for all pairwise grid cell combinations (Tuomisto, ). Simpson's beta was used because it reduces the effect of any species richness imbalance between locations. β sim i , j = 1 − a a + min ( b , c ) . where a refers to the number of species common to cells i and j, b is the number found in cell i but not cell j, and c is the number found in cell j but not cell i. A low β sim value indicates that many taxa are shared between two grid cells (low dissimilarity) and a high β sim means a small number of shared taxa (high dissimilarity). Cluster analysis The β sim pairwise distance matrix was used in an agglomerative cluster analysis to generate a WPGMA (weighted pair‐group method using arithmetic averages) hierarchical cluster in Biodiverse . WPGMA weights the contributions of clusters by the number of terminal nodes (data set cells) they contain, ensuring each cell contributes equally to each merger of which it is a part. Its performance was determined to be as successful as UPGMA in Kreft & Jetz ( ). We implemented a tie breaker approach such that, when more than one pair of clusters had the minimum turnover score and thus could be merged, the algorithm selected the pair that maximized the corrected weighted endemism score in the cluster (Crisp et al ., ; Laffan & Crisp, ). This approach guarantees a stable solution that is replicated each time the analysis is run, as well as increasing the degree of endemism and thus the spatial compactness of the resultant bioregions. We identified the phytogeographical regions from the clusters based on two criteria: (1) a phytogeographical region is preferably represented by a group of contiguous, or near‐contiguous, grid cells, (2) each cluster that represents a phytogeographical region needs to be clearly separated from its children or parent. Relative Environmental Turnover Relative environmental turnover (RET) was applied to understand the relationship of environmental variables to the phytogeographical regions estimated using species turnover, as first tested with the Australian genus Acacia (González‐Orozco et al ., ). Previous studies used the term environmental turnover to explore rates of change of dissimilarity in vertebrates and their relationship to environment depending on the geographical distance (Buckley & Jetz, ). RET differs from other studies because it contains two types of analyses: ordinations of the β sim with the environmental variables and the grid cell‐based Getis‐Ord Gi* hotspot statistic. For the ordination analyses, a non‐metric multidimensional scaling (NMDS) ordination was generated using the β sim values. We used NMDS because it is an unconstrained metric that assumes neither normally distributed variables nor linear relationships between variables. This approach allows us to obtain a dimensionless ordination diagram with axes that can be interpreted in terms of an environmental gradient. More traditional methods such as principal component analyses (PCA) were not applied because they assume the data follow a linear normal distribution. Mean values for each of the grid cells were extracted from the 100 km environmental layers. We used the R statistical software (R Development Core Team, ) to produce the β sim values in a matrix format. We then matched the records between the β sim dissimilarity matrix and the mean per cell values of the environmental datasets (see Appendix S2 for heat maps of the environmental variables). The function ‘metaMDS’ of the vegan package (Oksanen et al ., ) was then used to generate the ordination. The β sim values were overlaid onto the ordination and fitted with the environmental variable matrix using the vector fitting of the envfit function from the vegan package. The environmental variables that best explained the patterns of turnover were then displayed as vectors only for those cases with high predictability ( P < 0.001), assessed using 999 permutations. The stress values of the envfit results were used to estimate the efficiency of the NMDS ordinations. The environmental analyses excluded the Malesian region because most environmental variables were not available at an appropriate resolution for that region. The values of grid cells along the first and second axes of the NMDS ordination of β sim distances were extracted and mapped in ArcMap. The geographical rate of change of each axis was calculated using the slope tool in ArcGIS. Those locations with steep slope values correspond with rapid changes (breaks) in the geographical distributions of turnover. The Getis‐Ord Gi* hotspot statistic (Laffan, ; Külheim et al ., ) was calculated using Biodiverse to assess whether the environmental values within each phytogeographical region were significantly different from those for the Australian continent as a whole, where each region was represented by its set of 100 km × 100 km cells. The Gi* statistic is expressed as a z‐score indicating the degree to which the values of a subset of grid cells, in this case the cells comprising a cluster, are greater or less than the mean of the dataset. Those clusters with Gi* values >2 or <−2 represent sets of cells that have environmental values significantly different from expected ( P < 0.05). Results Species richness and endemism The most species rich and endemic regions of eucalypts were located south of the Tropic of Capricorn. We identified three main centres of species richness (Fig. a). Six of the ten grid cells with the highest species richness scores (86–116 species per grid cell) were located in the south‐east coastal province of the Southwest Floristic Region of Western Australia (location 1 in Fig. a). The remaining four of the ten centres (richness scores from 86 to 92) were located in south‐eastern Australia. The New South Wales and Queensland border (location 3 in Fig. a) has not been previously identified as a centre of eucalypt species richness, although the general region was recognized as a centre of species richness for the Australian flora (Crisp et al ., ). Maps of species richness (a) and endemism (b) for eucalypts in A ustralia and M alesia. Numbers in the maps are referred to in the text. Fourteen centres of high eucalypt endemism were identified (Fig b), with the highest scores (between 0.1 and 0.5) being located in Western Australia and East Timor. Five of these endemism centres are in eastern Australia (6, 9, 10, 11 and 12 in Fig. b) and were not identified by Crisp et al . ( ) as being among the 12 endemic areas of the Australian flora. Phytogeographical regions Seven eucalypt phytogeographical regions are proposed, of which six are in continental Australia and one comprises the islands to its north (Fig. a). There are two small clusters in the Malesian islands: one in East Timor and the second further east including part of Indonesia, East Timor and Papua New Guinea, all of which have endemic eucalypt species. The remainder of the Malesian region clusters with larger Australian regions. One cluster differentiates the south‐eastern Australia region from the rest of the continent (Fig b). (a) Map of the seven eucalypt phytogeographical regions, (b) corresponding cluster dendogram and (c) the biomes of A ustralia according to Crisp et al . ( ). Map regions are coloured based on relationship in the corresponding dendogram. For example the blue regions for a higher level cluster as do the red colours. Figure a contains two green clusters in M alesia. These two clusters subtend the rest of the dendogram (2b) and are shown in black and grey for clarity. The dendogram branch lengths (Table ) for each of the geographical clusters (Fig. ) show Australia is subdivided into six major phytogeographical regions (A–F). The Eremaean north region is subdivided into four subregions, where the Pilbara (a) is most similar to western coastal (b) whereas central (c) is most similar to the Eastern (d) (See [(a,b) (c,d)] in Fig. A). The monsoon phytogeographical region is composed of mainland Australia and Malesian subregions, where the Top End/Malesia subregion (a) is most dissimilar to the rest of the subregions (See [(a) (b,c,d)] in Fig. B). The inland monsoon subregions are divided east‐west in three subregions: east (b), central (c) and west (d). The tropical/subtropical region is an independent cluster related to the northern regions of the monsoon and the Eremaean north (See [(a) (b,c)] in Fig. C). This region has a strong north to south pattern with Cape York (a) to the north and a southern coastal cluster, which can be further separated into Queensland Central (b) and New South Wales south (c) subregions. Values of branch length for clusters of floristic regions and subregions of A ustralian eucalypts Eucalypt phytogeographical regions Geographic subdivisions Branch length to parent (brackets refer to subregions labels in Fig. ) Malesia Top End/Malesia 0.04 Monsoonal Belt 0.08 South‐east (Major split) 0.23 Alpine 0.21 (a) Tasmanian/southern Victoria 0.15 Tasmanian 0.27 (b) South‐eastern mainland 0.03 (c) Eremaean south/south‐west (Major split) 0.01/0.02 Eremaean south: 0.03 Southern MDB 0.14 (a) Nullarbor Plain 0.18 (b) Northern 0.14 (c) Southern 0.14 (d) South‐west: 0.18 Eastern/Central 0.40 (a) Western (Southern) 0.21 (b) Western (Northern) 0.53 (c) Tropical/Subtropical (Major split) 0.03/0.08 Tropical: 0.20 (a) Cape York 0.17 (b) QLD North 0.24 Subtropical: 0.34 (c) NSW South Monsoon/Eremaean north (Major split) 0.06 Monsoon: 0.11 Malesia 0.16 (a) East 0.16 (b) Central 0.06 (c) West 0.12 (d) Eremaean north: 0.04 West 0.06 Western‐ Pilbara 0.13 (a) Western‐ Coastal 0.15 (b) Central eastern 0.10 Central 0.07 (c) Eastern 0.04 (d) Map of the eucalypt phytogeographical regions and their respective subregions. Regions are denoted in upper case letters: A = Eremaean north; B = monsoon; C = Tropical/subtropical; D = south‐west; E = Eremaean south; F = south‐east. Subregions are denoted in the parenthetical notation using lower case (see Table for corresponding names) and are coloured based on relationship in the corresponding dendogram. Eucalypt phytogeographical regions and subregions defined by analysis of S impson's beta (β sim ) species turnover (letters in brackets refer to regions and subregions shown in Fig ) Eucalypt phytogeographical regions Geographic subdivisions Subregions Monsoon (B) Top End/Malesia Top End/Malesia (a) Monsoonal Belt East (b) Central (c) West (d) Eremaean north (A) West Pilbara (a) Coastal (b) Central/eastern Central (c) East (d) Eremaean south (E) Eastern/Central Southern MDB (a) Nullarbor Plain (b) Western South (d) North (c) Tropical/Subtropical (C) Tropical Cape York (a) Subtropical QLD North (b) NSW South (c) South‐east (F) Alpine Alpine (a) Tasmanian/southern Victoria Tasmania (b) South‐eastern Victoria (c) South‐west (D) South‐eastern South‐eastern (a) Western North (c) South (b) The Eremaean south and the south‐west phytogeographical regions form a single cluster separate from the monsoon, Eremaean north and tropical/subtropical regions. While the branch combining the two areas is short, the branches supporting Eremaean south and the south‐west are longer. We recognize these two eucalypt phytogeographical regions based on long nested branches. The south‐west region is divided into south and west subregions (see [(a) (b,c)] in Fig. D). The Eremaean south region is subdivided into two subregions that stretch from east to west (See [(a,b) (c,d)] in Fig. E). The most eastern/central subregion further subdivides into a southern Murray‐Darling Basin subregion (a) and a subregion along the Nullarbor coastal plain (b) whereas western subregions (c–d) of inland and coastal areas connect with the south‐west region. The south‐east region is subdivided into an alpine (a) subregion and a Tasmanian/southern Victorian subregion (See [(a) (b,c)] in Fig. F). The southern portion subdivides into Tasmanian (b) and south‐eastern mainland (c) subregions. Relative environmental turnover (RET) The geographical surface of the first NMDS axis indicates a major geographical break oriented north‐west to east across the continent (Fig. a, see also the red line in Fig. c). This break is aligned with the summer–winter rainfall line defined by Burbidge ( ). The pattern produced by the second axis (Fig. b) matches the division of arid central Australia, which has less eucalypt diversity, and the mesic zone on the coastal regions of eastern Australia. Areas noted as having high slope values are regions with abrupt changes of species turnover and are inferred turnover breaks. For example, the major break we found is detected by NMDS1 (red line in Fig. c), and follows the subregion boundaries (Fig. ). No major pattern is found for NMDS2, but ten small inferred turnover break areas are scattered across the continent (Fig. d). Of these, area 1 is a known barrier for the Wet Tropics, area 2 corresponds to the MacPherson/MacLeay overlap zone between NSW and Queensland east coast, area 4 corresponds with the Murray and Gippsland Basin barriers, and area 10 reflects the known barrier of the Nullarbor Plain of southern Australia. Species turnover of A ustralian eucalypts measured with the non‐metric multidimensional scaling ( NMDS ) method for (a) axis 1 ( NMDS 1) and (b) axis 2 ( NMDS 2). The two right side panels plot the slopes of NMDS 1 (c) and NMDS 2 (d) and the inferred biogeographic breaks (red lines and polygons). A low stress value (0.1016) suggests a good fit of the environmental data to the β sim clusters in the ordination model (Fig. ). With the exception of clay content, all variables were significantly ( P < 0.001) correlated with β sim . For example, the values towards the northern areas of monsoonal Australia correspond with the points located in the top left corner of the plot (indicated by the North arrow). The vectors for annual precipitation and precipitation during winter vectors correlate with species turnover towards the east and north‐west of Australia. Vectors for temperature seasonality, annual mean radiation, topography, sand content and lithology correlate with species turnover towards the west and south‐west of Australia. Vectors for annual mean temperature and annual mean radiation vectors correlate with species turnover towards the western parts of the monsoon region as well as the central Australian deserts located in the eastern part of the Eremaean north region. Vectors for precipitation during the coldest quarter and radiation seasonality were highly correlated with the southern regions. Non‐metric multidimensional scaling (NMDS) of eleven environmental variables fitted to turnover (β sim ) of A ustralian eucalypts. Vector names representing seasonal related variables are bolded while aseasonal and non‐climatic variables are not. The N orth arrow in the top left corner represents the geographic orientation of the NMDS values. The results of the gridded approach, where the effects of 11 environmental variables were tested in each of the eucalypt bioregions, are shown in Table . A value >2 or <−2 indicates that the values of an environmental variable in that floristic region are, collectively, significantly different from the study region as a whole. Overall, climate variables such as precipitation and temperature appear more important for explaining the turnover patterns of the eucalypt phytogeographical regions than are non‐climatic variables such as soil and topography. However, there is some degree of variability from region to region. For example, precipitation seasonality is the main driver for the monsoon region, while precipitation during coldest quarter of the year is the main driver for the south‐eastern region. Temperature seasonality is the main driver for the Eremaean north region, meaning that monthly variation in temperature correlates well with species turnover in the drier and hotter areas north of the Tropic of Capricorn. Interestingly, precipitation seasonality represented as winter rainfall (south of the Tropic of Capricorn) becomes the main driver for the Eremaean south region, meaning that rainfall patterns between June and August play a determinant role. As expected, the main driver of the tropical and subtropical north‐eastern and eastern regions is annual mean precipitation. The species turnover in the temperate regions of south‐east and south‐west Australia is highly correlated with the amount of precipitation during winter months. The results of the gridded approach tested on each of the eucalypt subregions are given in Table S3 of Appendix S3. Gi* statistic scores for the cells comprising the six major phytogeographical regions of A ustralian eucalypts. Bold text signifies statistically significant (α = 0.05). N = number of grid cells per region Monsoon ( N = 127) Eremaean north ( N = 341) Eremaean south ( N = 204) Tropical ( N = 38) Subtropical ( N = 38) South‐east ( N = 65) South‐west ( N = 21) Annual mean radiation 6.46 9.83 −2.47 −0.58 −4.18 −15.16 −7.11 Annual mean temperature 13.75 7.54 −9.18 4.88 −2.84 −15.35 −6.82 Annual mean precipitation 10.06 −11.67 −10.53 11.02 8.80 8.49 2.21 Percentage of clay 0.51 2.35 −1.64 −1.75 1.98 −0.11 −4.28 Precipitation coldest quarter −8.18 −7.84 0.67 −2.97 4.55 16.32 11.93 Precipitation seasonality 16.88 2.59 − 13.73 8.54 −1.73 −10.20 −0.83 Radiation seasonality −15.19 −6.37 12.91 −5.48 −1.27 13.98 4.71 Ridge Top flatness 2.56 0.42 4.80 −2.65 −4.82 −5.35 −1.34 Rock grain size −1.39 −7.45 8.51 −0.46 0.04 0.28 3.63 Percentage of sand 1.63 −0.91 −0.27 1.60 −1.80 −1.14 2.10 Temperature seasonality −12.79 12.39 6.81 −10.16 −3.20 −3.96 −4.36 Discussion Eucalypt phytogeographical regions Our seven phytogeographical regions and 21 subregions contribute significantly to the understanding of contemporary Australian biogeography. The major split of the main two clusters suggests the existence of a north‐south split of the eucalypt phytogeographical regions. The southern area might reflect historical adaptation to habitats affected by higher winter rainfall whereas the northern area might be more influenced by the summer rainfall adapting to high humidity and extreme heat. Within the southern area, the Eremaean South and south‐west share higher similarity than with the south‐east. In contrast, the clusters corresponding to the monsoon and Eremaean north share higher similarity than with tropical/subtropical clusters. The limits of the southern area extend east‐west from Western Australia to central New South Wales. This major pattern could be explained by the effect of regolith carbonate accumulations in the soil (Hill et al ., ; McQueen et al ., ). Such a geochemical difference is a potential explanation for the presence of the major north‐south split of the continent wide phytogeographical regions of the eucalypts. Our proposed eucalypt phytogeographical regions (Fig. ) are in broad agreement with the biomes of Australia (Burbidge, ; Schodde, ; Crisp et al ., ). The main difference between the result of our analysis and those of Burbidge ( ) is that the northern Tropical Cape York region is not differentiated as part of the monsoonal tropics biome, instead it is part of the north‐eastern tropical region (Fig. c). The extent of Burbidge's monsoonal tropics is larger than ours, specifically towards the central eastern region. The Eremaean zone is a single unit in Burbidge's classification while here we divide it into two distinct regions with the Eremaean north clustering with the monsoonal bioregion and not with the Eremaean south. This same relationship was also found in Acacia (González‐Orozco et al ., ). Species endemism is a key concept to explain biogeographical patterns (Parenti & Ebach, ). We found that in six of the seven proposed eucalypt phytogeographical regions there is at least one centre of endemism (Table ). Occurrence of centres of species richness and endemism per phytogeographical regions of A ustralian eucalypts. Numbers denote the number of centres present in each bioregion, a dash (–) denotes no centres Eucalypt phytogeographical regions Species Richness Endemism Monsoon – 1 Eremaean north – 2 Eremaean south – – Tropical – 1 Subtropical 1 3 South–east 1 4 South–west 1 3 Linking eucalypt phytogeographical regions to environment There are many examples of bioregionalizations in Australia, such as the agro‐climatic classification, the environmental domain classification and the primary productivity regionalization of Australia (Hutchinson et al ., ; Mackey et al ., ). These are important frameworks but are strongly biophysical, as opposed to the biologically centred approach used here. The only study comparing species distribution of eucalypts to climate at the continental scale was conducted by Gill et al . ( ). They identified two eucalypt clades adapted to warm climates ( Blakella‐Corymbia and Eudesmia ) and another two composed of mostly of species that are adapted to cool climates. The genus Eucalyptus contains 10 subgenera, but subgena Blakella and Corymbia comprise the subgenera Corymbia . Gill et al . also noted that temperature during the wettest quarter of the year and rainfall patterns overall had a great effect on the species located in the northern tropical regions. Our work identified a predominant north‐east and south‐west geographical pattern, which fits the climatic rationale presented by Burbidge ( ). Burbidge also identified the Tropic of Capricorn as a major biome delineator: higher summer rainfall to the north and winter rainfall to the south. The inconsistencies of environmental drivers between our study and that of Gill et al . ( ) may be due to the fact that we used more species records, a greater number of species and applied a metric that minimizes the imbalance of species richness in representing biotic turnover (see Table ). Gill et al . also used meteorological data from the weather station nearest to each cell, whereas our climate data are derived by spatial interpolation using all surrounding stations. Ten principal differences between the present study and that of Gill et al . ( ) for the analysis of the distribution of species dissimilarity and diversity of A ustralian eucalypts Study parameters Gill et al . ( ) Present study Spatial scale 1.0 × 1.5 degrees (lat × long) 100 × 100 km grid cells Turnover metric Jaccard and Czeckanowski Beta‐Simpson index Number of species 551 798 Taxonomic resolution Species and subgenera (without Angophora ) Eucalyptus , Angophora and Corymbia Statistical technique PCA NMDS Cluster metric UPGMA WPGMA Environmental variables Ten (five temperature and five rainfall variables with no soils) Eleven seasonal and aseasonal variables Terminology Zones Bioregions and subregions Distribution Continental Continental and Malesia Diversity metric used Species richness and absolute endemism Species richness and relative endemism Number of bioregions 11 major zones and 18 subgroups Six main bioregions and 13 subregions Comparing eucalypt phytogeographical regions to major eucalypt communities Our phytogeographical regions closely match with previous broad scale studies on vegetation communities in Australia that were based on expert opinion. Beadle ( ) identified four main Eucalyptus vegetation alliances and eight generic biogeographical regions of eucalypts. The main difference of our study from Beadle's is that he did not define the monsoon region per se and the semi‐arid and arid areas in central Australia are not split into two regions as we propose here. A more recent vegetation classification based on the National Vegetation Information System (NVIS) classification (DEWR, ) defined six major eucalypt vegetation groups, which broadly correspond with our phytogeographical regions. The results of this research support previous findings regarding the location of the main centres of species richness but, interestingly, we identified new centres of endemism. The eucalypt centres of species diversity and endemism tend to be different from other groups of Australian plants and mostly occur south of the Tropic of Capricorn. For example, the biodiversity hotspot of south‐west Australia is an area of high species richness and endemism for both Acacia (González‐Orozco et al ., ) and eucalypts but not for Glycine (González‐Orozco et al ., ) or bryophytes (Stevenson et al ., ). However, the species rich and endemic areas in the south‐west do not overlap but are mostly adjacent to each other suggesting separate environmental niche preferences for the two lineages. Acacia species richness and endemism is most pronounced in a belt delimiting the south‐west for the arid zone while the eucalypt species richness and endemism dominate the coastal regions. There is an area of overlap of species richness and endemism in Acacia and the eucalypts in the broader Stirling ranges — Fitzgerald River region of Western Australia (see area 1 in Fig. a). Overlapping patterns of high species richness, with fewer areas of endemism, are found in south‐eastern Australia for the eucalypts, Acacia , hornworts liverworts and mosses (González‐Orozco et al ., , ; Stevenson et al ., ). This suggests that south‐eastern Australia is an area of high speciation and broader geographical distributions. The Wet Tropics is an area of high species richness and endemism for all these lineages. Bioregionalizations have been conducted on continental (Rueda et al ., ; Linder et al ., ) and global scales (Kreft & Jetz, ; Holt et al ., ). These works provide an alternative approach as they consider a subset of the diversity of groups such as plants, insects and vertebrates within large geographical areas whereas our study focuses on a single fully sampled lineage at a continental scale. Linder et al . ( ) also used β sim and identified seven sub‐Saharan biogeographical regions that were broadly congruent among the plant and vertebrate groups studied. However, Rueda et al . ( ) did not find congruence among lineage‐specific bioregionalizations in Europe. An allied study on Acacia (González‐Orozco et al ., ) revealed similar but not identical phytogeographical patterns to the eucalypts. It is possible that the differing evolutionary histories of individual lineages, when combined in a single analysis, would produce the less defined patterns that were seen by Linder et al . ( ) in sub‐Saharan Africa. More studies are needed to better understand whether bioregionalizations of plants and animal groups are congruent (Linder et al ., ) or incongruent (Udvardy, ; Rueda et al ., ) and what drives the patterns. Our finding that climate and dissimilarity of plant distributions correlate with species turnover confirms previous global studies (Buckley & Jetz, ). For example, Buckley and Jetz ( ) found that high levels of species turnover occur regardless of environmental turnover rates, but environmental turnover provides a lower bound for species turnover in amphibians than for birds. Because each continent has specific topographic and climatic attributes, it would not be expected to find major similarities on the main environmental control of turnover. The patterns of eucalypt diversity identified here provide a foundation for future biogeographical studies and investigation of eucalypt environmental niches. In the eucalypts, the patterns of species richness and endemism are not tightly linked, especially in south‐eastern Australia. Those taxa in the species rich areas have broad ranges, thereby lowering the corrected weighted endemism scores and suggesting multiple dispersal episodes following events of active allopatric speciation. Limited inferences can be made on data based on species distribution without a phylogenetic context. It is probable that combining over 800 species into a single study will obscure important signals that can be determined through phylogenetics. The utilization of phylogenetic diversity (Faith ), phylogenetic endemism (Rosauer et al ., ) and related metrics may better unravel the historical factors and evolutionary relationships behind the organization of biodiversity on the landscape. A thorough understanding of the basis for these patterns, including phylogeny and environmental niche preferences, can inform conservation planning decisions. Although our classification is not based on phylogenetic relationships, we found a few similarities with some phylogenetic focussed studies. Ladiges et al . ( ) reported that species of Corymbia and Eudesmia, analysed using area cladogram techniques, suggest differences between the monsoon tropics and the eastern areas of Australia. Here, we identified similar patterns across the larger eucalypts group. The phylogenetic structure of the eucalypts in these bioregions and at smaller scales could be driven by alternative ecological patterns, such as phylogenetic clustering or over‐dispersion (Webb et al ., ). Potential caveats We demonstrate that bioregionalizations of a continent using species turnover of a large taxon group is a potential method to understand biogeography, as shown in other case studies (see Linder et al ., ). However, despite partitioning, the continent into meaningful biogeographical areas of species assemblages, the phytogeographical regions are not evolution‐based and therefore phylogenetic approaches would benefit future studies (Ferrier et al ., ; Rosauer et al ., ). Sampling bias is a common problem with herbarium data, and under‐sampling is thus a common cause of uncertainty. We explored the potential implication of sampling bias by conducting a redundancy analysis, calculated as the ratio of species records to number of samples per grid cell (Garcillán et al ., ). We found that only 274 of all 906 grid cells analysed had redundancies less than 30% (see grid cells in green and blue shades; Appendix S1). This indicates that 70% of the continent is comparatively well sampled. The poorly sampled areas, which are often remote, tend to correspond with areas of low turnover values. The grid cell size may also have an effect on the turnover patterns (Barton et al ., ). A pilot test (results not shown) that assessed the effect of grid cell size (50 km × 50 km; 25 km × 25 km) on the mapping of the phytogeographical regions found no major spatial difference in terms of number of phytogeographical regions generated. However, we observed that the resolution has a small effect on the boundaries of the bioregions. Conclusion Rainfall plays a key role for northern Australian eucalypt assemblages, whereas temperature and solar radiation are more important in the south and south‐eastern regions, specifically south of the Tropic of Capricorn. Current climate projections suggest warmer temperatures for eucalypts in Australia, particularly the southern populations (Hughes et al ., ). Temperature is a key driver of eucalypt species turnover, which highlights the importance of special care when managing biodiversity around areas located in southern latitudes. Species turnover patterns of eucalypts in the western part of the continent correlate more with non‐climatic factors such as topographic and soil properties. Although the proposed phytogeographical regions are exclusive to eucalypts, our method can be applied to any geographical scale providing sufficient data with the appropriate taxonomic and spatial detail are available. Acknowledgements We would like to thank to the CSIRO referees for useful comments on the manuscript and Andrew Slee for help with distribution and taxonomy issues.

Journal

Diversity and DistributionsWiley

Published: Jan 1, 2014

Keywords: ; ; ; ; ; ; ; ;

There are no references for this article.