The assembly processes that structure biodiversity across a range of scales form the theme of macroecology (Chave, 2013). This paper deals with the composition of seaweed assemblages along the ~2,900 km South African coastline, including the identification, description and explanation for the spatially-structured patterns at scales from 100s to 1,000s of kilometers. Inshore conditions along this coastline range from cool through warm temperate to fringe tropical (Bolton and Anderson, 2004; Bolton et al., 2004), and are influenced by two major ocean currents in two oceans that set up a strong thermal gradient along the shore (Smit et al., 2013). We investigate how a gradient in seaweed community composition may have arisen in response to this temperature gradient (e.g., Qian and Ricklefs, 2007), and the extent to which deterministic, niche-based species assembly processes could have contributed toward the assembly of the seaweed flora of the region.

Biodiversity may be viewed in terms of α-, γ- and β-diversity, with the former two referring to “local” and “regional” diversity, respectively. β-diversity as defined by Whittaker (1960, 1972) is a measure of variation in species composition from place to place and is comprised of two processes (Baselga, 2012): the replacement of species independently of the difference in species richness (called “turnover,” β_sim) and a term that considers this difference in species richness called (“nestedness-resultant,” β_sne). The relative contribution of these two processes is not immediately evident from species dissimilarities, and yet such considerations should be implicit in macroecological studies. The ability to decompose β-diversity into turnover and nestedness-resultant components has become possible within the last decade (Baselga, 2010), but the use of this form of β-diversity partitioning has not yet widely permeated the phycological literature (or even in marine studies more broadly; Anderson et al., 2013). Cognisance of spatial scaling should also be deeply rooted in the study of biodiversity more generally, but such questions have only recently begun to be addressed (Barton et al., 2013).

Since studies of β-diversity consider the variation in species assembly processes from place to place, and necessitate an understanding of mechanisms that drive these processes (Davidar et al., 2007), studying β-diversity of adjacent coastal marine bioregions or marine provinces (Spalding et al., 2007) could provide deeper insight into how such processes operate along gradients (Qian and Ricklefs, 2007). Few studies dealing specifically with β-diversity exist for marine biota, especially over spatial areas in excess of 1,000s of kilometers. Recent examples include studies on fish distribution (Zintzen et al., 2011; Anderson et al., 2013), the biogeography of macroalgae along 6,600 km of Australian coastline (Leaper et al., 2011), and a study on deep-sea bivalves (McClain et al., 2011). All four have β-diversity as a primary interest. However, macroecological studies in the terrestrial realm have yielded the most comprehensive insights into the drivers responsible for assembling species into communities (Whittaker, 1960; Davidar et al., 2007; Qian and Ricklefs, 2007; Soininen et al., 2007). What these studies showed is that, perhaps universally, high rates of species turnover are associated with steep environmental gradients, such as those occurring along mountain slopes or across latitudes. In the ocean, the biggest driver latitudinally seems to be temperature (Tittensor et al., 2010; Straub et al., 2016; Stuart-Smith et al., 2017); other drivers such as nutrients, salinity, turbidity, wave action and photoperiod may also structure species composition, but these are likely to depend more on the regional context, and some, such as nutrients (e.g., Waldron and Probyn, 1992), are often strongly correlated with a more distal variable, such as temperature.

Recent decades have seen temperature emerging as a readily available environmental driver that's able to resolve the ocean's heat content across a multitude of temporal and spatial scales. Temperature is also very useful as a predictor of species range limits due to it being one of the key output variables of coupled global ocean-atmosphere circulation models used to project the physical milieu of the future Earth that has experienced anthropogenic climatic change (IPCC, 2013), hence making it the focus of our present study. The controlling effect of seawater temperature on the survival and reproduction of benthic organisms and patterns in the evolution and ecology of biological assemblages at regional scales is well known (van den Hoek, 1982b; Breeman, 1988; Blanchet et al., 2008; Broitman et al., 2008; Byrne et al., 2009; Verbruggen et al., 2009; Wieters et al., 2009; Couce et al., 2012; Potts et al., 2014). It influences the species composition of the biota associated with various thermal zones, which forms patterns on global, regional and local scales (Tittensor et al., 2010; Spalding et al., 2012). At the global scale, ocean currents and well-characterized latitudinal solar heat flux gradients maintain the ocean's thermal regime. At local scales (<10 km) at the land/sea margin, however, additional less well-known physical phenomena contribute to thermal patterns and dynamics that may differ markedly from those at the mesoscale (i.e., spatial scales > 50 km), and the different properties of the temperature regime which set up the biogeographical patterns are less well understood there—this is especially true at regional and local scales near the land where satellite data poorly reflect reality (Smit et al., 2013). Most studies that recognize temperature as a major driver of species distribution take the annual mean temperature (Tittensor et al., 2010); fewer recognize the importance of variability and range (e.g., Couce et al., 2012; Tyberghein et al., 2012). Detailed laboratory culture studies with seaweed species, for example, closely link biogeographical distribution limits with maximum and minimum monthly mean temperatures (van den Hoek, 1982b; Breeman, 1988). The time-integrated “thermal environment” is an amalgam of various statistical properties derived from a multitude of instantaneous temperature recordings, and it is pertinent to question if one or a few of these properties have an overriding imprint on the species assembly.

Understanding the processes influencing species turnover and geographical range is of practical importance to spatial biodiversity planning. This is especially relevant in today's warming world because it will allow us to project future biodiversity responses. Focusing conservation efforts on areas of high β-diversity will ensure the preservation of a diversity of species and their environmental niches. Peaks in β-diversity along ecoclines may indicate boundaries between bioregions, highlighting regions where communities are poised at their environmental limits, thus aligning with long-term monitoring surveys that aim to project ecosystem responses to climatic change.

Using a detailed data set of seaweed presence and absence records coupled with coastal in situ seawater temperature climatologies that are able to resolve the coastal zone, we investigated the thermal properties and species composition of 58 coastal sections spaced around the South African coastline. The coastline is broadly influenced by two major ocean currents, the Benguela Current and the Agulhas Current. One is an eastern boundary upwelling system and defines a cool temperate environment, and the other a western boundary current driving meridional transport of sub-tropical water toward the tip of Africa. We were primarily interested in establishing how these ocean currents, of which the effect at the coast can readily be measured as gradients in thermal properties, may have influenced the species assembly processes operating along an approximately 2,900 km long coastline. To do so, we first link matrices of species dissimilarity to environmental distance matrices and Moran's eigenvector maps (MEM) using distance-based redundancy analysis, and then apply variance partitioning (Peres-Neto and Legendre, 2010) to determine the relative contributions of thermal metrics and other spatially-organized drivers of seaweed community composition (as seen in the β-diversity components, β_sim and β_sne). Second, we examine how the thermal and spatial structuring agents operate at smaller spatial scales within marine provinces of the region by undertaking a more detailed analysis of the various distance matrices, and also ask which of the thermal properties we examined were most influential in setting up the patterns that emerged. Our findings are congruent with existing knowledge (e.g., Stephenson, 1948; Lombard et al., 2004; Spalding et al., 2007) and add a more nuanced understanding of the processes responsible for structuring the seaweed biodiversity in the sub-region.

In this paper, we considered the drivers of seaweed β-diversity at the scale of bioregions (marine provinces) nested within a ~2,900 km stretch of coastline. We examined β-diversity using a numerical ecological approach that permits assessing its spatial pattern independently of the influence of α-diversity (Baselga et al., 2017), which in our study displays major regional differences. There are fewer species of macroalgae on the Benguela Current dominated west coast, as is also shown by studies on invertebrates and fishes (Thandar, 1989; Turpie et al., 2000). There is a similar species richness per coastal section on the south coast and the east coast. This was not the case up until the early 2000s (Bolton and Stegenga, 2002), when a more restricted data set showed more species per section on the south than east coast. This difference disappeared during the mid-2000s, when major collections on the east coast filled out the species lists available for the region (De Clerck et al., 2005). Continuous sampling over three decades has now resulted in a robust α-diversity data set, and it is still different from the animal data: fish and invertebrates become more abundant per section on the east coast (toward the tropics) but the seaweeds do not. This matches previously published global seaweed diversity patterns, which suggested that biogeographical seaweed diversity does not peak in the tropics. Globally, seaweed species diversity at a biogeographical scale is highest in warm temperate regions (Bolton, 1994; Santelices et al., 2009). Examining spatial trends in α-diversity on its own offers an incomplete understanding of the structuring of biodiversity at regional scales, and more insight may be obtained by looking at species richness (i.e., lists of species, such as which we use in this study) across two scales, sensu Whittaker (1972)—i.e., α-diversity (species richness at the local scale, such as per coastal section) and γ-diversity (total species richness in all coastal sections). In this broad sense, the two scales of species richness are not independent, and the loss or gain in species from section to section will be seen in β-diversity. In this work, we employ Baselga's (2010) framework to represent overall β-diversity as two components, one of which excludes the effect of regional differences in α-diversity.

With the exception of the some Australian studies (Smale et al., 2010, 2011; Waters et al., 2010; Leaper et al., 2011; Wernberg et al., 2013), one in Europe (Tuya et al., 2012) and another two in the northwest Indian Ocean/Red Sea region (Schils and Wilson, 2006; Issa et al., 2016), studies of this scale and nature have so far been infrequently seen for marine macroalgae. Of the above-mentioned studies, only two specifically considered β-diversity (Leaper et al., 2011; Issa et al., 2016). Our data focus on the taxonomic representivity of a region by aggregating species occurrence records within 50 km long sections of coastline, and are blind to the effects of small-scale habitat heterogeneity (e.g., as seen in Smale et al., 2010). For this reason we cannot infer influences of stochastic processes and other aspects of environmental heterogeneity at scales of <50 km; as such we excluded drivers that may be only of local relevance, such as salinity, turbidity (light), and wave action. Rather, our data emphasize broad biogeographic patterns (Lawton, 1999), with induced spatial dependence (Peres-Neto and Legendre, 2010) emerging as the main species assembly process east of the Cape Peninsula. This points to deterministic niche influences that correlate with environmental drivers that function over a broad spatial scale. These environmental and biotic patterns reflect the nature of the two dominant ocean currents of the region, and we see that β-diversity at the scale of the country is moderately high and generally influenced by processes that cause species turnover (β_sim). Our selected environmental drivers comprise mostly thermal properties of seawater with a strong spatial dependence across a broad scale (82.4%) and a smaller amount of unknown non-thermal broad-scale spatial influences (13.7%). Only in the species-poor west coast region is there evidence of neutral assembly processes (e.g., dispersal limitation, and stochastic processes as one might expect within kelp forests), as shown by the higher rates of nestedness-resultant β-diversity (β_sne) emerging there. We recognize that some environmental drivers such as nutrients can and do co-vary with temperature, but they were not used in our analysis because they are not as readily measured as temperature, they are not always directly produced by coupled global ocean-atmosphere models and thus have less value as predictive (long-term) variables, and because of the issues arising due to collinearity.

The most comprehensive assessment of global marine biogeography (Spalding et al., 2007) places much of South Africa's coast within two realms: Temperate Southern Africa (from southern Angola to around 56) and the western-most edge of the Western Indo-Pacific Realm (across the Western Indian Ocean to Sumatra). The transition into the Western Indian Ocean realm is just visible in section 57 where our data show a steady rise to tropical water temperatures with an annual minimum that exceeds 20°C, between 28.5°S and 29°S. Our study region is too restricted to capture the rates of β-diversity change at the transition of realms. Here we would expect higher rates of turnover than we report for the smaller-scale “Provinces” and transition zones that make up the Temperate Southern African Realm (see below). We would also expect this at higher taxonomic levels, consistent with the definition for a realm (Spalding et al., 2007): this is true for the temperate Southern African seaweed floras, which have high species endemism but very low generic endemism (except for the Fucales, Ochrophyta). We hypothesize that this is because the cool temperate region may be geologically recent (4.6–3.2 Ma, Marlow et al., 2000).

Nested within the realms are marine provinces, areas that according to Spalding et al. (2007) are large, with distinct biotas, some level of endemism (mainly at species level), and distinctive abiotic environments. In this context, the Benguela and Agulhas Currents prescribe the broadest scale hydrographic features whose imprint can be seen, at that scale, on the seaweed flora: they maintain the Benguela Marine Province (BMP) and Agulhas Marine Province (AMP, as per Spalding et al., 2007) of South Africa, separated by an area of transition. The southward flowing Agulhas Current has an overriding effect on the east coast, extending as far as the eastern portion of the Western Cape Province (58–22). Along the east coast it is responsible for setting up a region that transitions from sub-tropical in the north to warm temperate near Cwebe (i.e., the region 58–44), which is termed the East Coast Transition Zone (ECTZ). From 44–22 the Agulhas Current continues to maintain an influence on the coast, but its direct effect is subdued because of the widening of the continental shelf southwards of Cwebe, and the warm-temperate biogeographical region, the AMP, is consequentially formed. The northward flowing Benguela Current from which upwelling is maintained by prevailing south-easterly trade winds influences the remainder of the Western Cape Province (west of 22), particularly from the western side of the Cape Peninsula (16/17) northwards to about 16°S. The influence of the Benguela Current here defines a cool-temperate regime, the BMP, with the range of monthly mean temperatures at most sections intermediate between cold-temperate and warm-temperate, according to the definitions of Lüning et al. (1990).

The oceanographic features of the Benguela and Agulhas Currents set up gradients in seawater temperature that run predominantly from east to west (i.e., August mean temperatures decreasing in this direction), but also from north to south (i.e., with the southern coastline becoming more variable in its thermal regime). Of particular ecological importance is the gradient in August mean temperature. It clearly structures the thermal distance decay curve, which shows a steadily increasing dissimilarity in thermal distance along the shore from 1. The gradient may be smooth or punctuated at irregular intervals by peaks in dissimilarity. Specifically, the mean temperature for Austral winter (August) increases more-or-less smoothly from the west. Within the region, species richness along the coastline reflects the general global trend—for most taxa but ironically not for seaweeds (Bolton, 1994; Santelices et al., 2009)—of diminishing diversity with decreasing temperature, which at that scale is seen as a latitudinal gradient: the cool temperate area of the west coast has a much lower α-diversity than the warm temperate and sub-tropical south and east coasts, with the latter two regions having similar values. This same pattern exists for fish and invertebrates in this region, albeit in data which include coastal waters further from the shore (Griffiths et al., 2010). In the previous presentation of the seaweed dataset (Bolton and Stegenga, 2002), the ECTZ had far fewer species per section than the AMP, but this change in the dataset results from significant research in the period 2000–2005 on the seaweeds of the ECTZ (De Clerck et al., 2005). In contrast, bringing the annual mean temperature or thermal range during Austral summer into the calculation of the thermal distance introduces several large punctuations in dissimilarities between some sections—most notably near Cape Point into False Bay (17), and near 30 and 48. The region of the coastline where low biodiversity community composition changes to highly biodiverse communities is reflected in the β-diversity, which peaks around the Cape Peninsula (16/17). Correlations between β-diversity (resulting from a change in α-diversity) and areas of transition have also been noted in several studies on terrestrial biota (Melo et al., 2009; Tonial et al., 2012), but such investigations are less established for marine biota (Schils and Wilson, 2006; Anderson et al., 2013).

β-diversity is generally lower within individual bioregions than at the scale of the country, because the coastlines are shorter and large differences in diversity cannot approach that which can occur over the entire coastline. Even within bioregions, β-diversity is still generally dominated by species turnover rather than by nestedness-resultant processes. The exception is the BMP where nestedness-resultant (β_sne) assembly processes contribute to around 58% of the β-diversity. The generally strong influence of turnover-based processes elsewhere highlights the role of environmental drivers, which here are spatially structured as one would expect of the steep environmental gradients, and hence of niche influences (i.e., environmental filtering) in determining species assembly (Fitzpatrick et al., 2013).

It seems plausible that factors linked with the establishment of oceanographic regimes (historical factors sensu Baselga et al., 2012) that influenced dispersal and speciation along the coast resulted in the current-day gradients in β-diversity within bioregions. This hypothesis was put forward to explain the low endemism and low species richness of Ochrophyta in the upwelling dominated west coast of southern Africa (Bolton, 1986). But is the underlying mechanism the selection of cold tolerant species from amongst a more diverse pool in the wider region (Bolton, 1986), or is it due to a dispersal barrier that prevented elements of the warmer south coast seaweed flora from occupying the west coast? Evidence from seaweed phylogeography postulates different evolutionary origins for seaweeds of the BMP and AMP (Hommersand, 1986; Hommersand and Fredericq, 2003). The mechanism could be the intensification of upwelling during the mid-Pliocene 4.6–3.2 Ma (Marlow et al., 2000). Turnover is almost entirely absent in favor of nestedness-resultant processes in the BMP, but the AMP is defined entirely by a moderate rate of turnover with respect to both geographical and thermal distance. The dominance of the nestedness-resultant component of β-diversity within the BMP suggests that the complex, non-linear influence of the annual mean temperature selectively influences species richness from section to section in a pattern that is not spatially coherent, but this climatic variable (i.e., fraction [E], which has some spatial structure) explains only 39.1% of the β_sne patterning in the region. Unaccounted-for non-climatic variables and perhaps non-environmental variables may explain some of the remaining variation. This points to habitat and/or environmental heterogeneity and neutral processes such as dispersal limitation, stochastic influences or habitat heterogeneity as the main structuring agents of communities within the region. Indeed, habitat heterogeneity generated through stochastic influences such as patchiness due to storms denuding portions of kelp beds (which is the dominant habitat type along the west coast) is well-known in kelp forests (Smale et al., 2011). Furthermore, the prevalence of two kinds of assembly processes to either side of the Cape Peninsula (17), coinciding with the point where a region of low α-diversity (BMP) transitions into a region of higher α-diversity (the Benguela-Agulhas Transition Zone, B-ATZ), suggests that the answer to our question is probably that some historical event was responsible for that disjunction, and that this barrier is still maintained today due to limited mixing between the Agulhas and Benguela Currents (although some of the sections along the west coast between 16/17 and 10 certainly have a reasonable probability of sharing species with sections at the western-most side of the AMP, e.g., 23, 24; see more on the B-ATZ, below).

Rapid rates of species turnover with respect to geographic and thermal distance along the east coast, and the steep decrease in species richness around the 16/17–22 transition region, suggest that relatively more endemic species should occur within the BMP and AMP. Seawater becomes warmer along a steep gradient along the east coast northward toward 54 where-after there is a further steepening of the gradient toward Mozambique (north of 58). This steepening of the already strong temperature gradient in the northern portion of the ECTZ supports the conclusion of Bolton and Anderson (2004) that it represents the transition from a tropical Indian Ocean seaweed flora in Mozambique to a temperate flora in the south. A similar pattern exists for the rocky intertidal biota in the region (Sink et al., 2005). The temperature of the coldest month of the year sufficiently accounts for the environmental gradient in all bioregions, except for in the BMP. This strong coupling between the thermal gradient and β_sim suggests that a niche difference mechanism is the primary species compositional assembly process (Nekola et al., 1999). This implies species being sorted based on their physiological tolerances along a gradient, and the particularly high rate of species turnover along the east coast reflects the species' narrow thermal ranges. Species here have narrow local distributions, and there is only a 2% seaweed endemism in the region (Bolton and Anderson, 2004).

The overlap region between 22 and the southern tip of the Cape Peninsula (16/17), which we call the B-ATZ, is an area where aspects of both currents may be periodically seen (Largier et al., 1992), and it is not surprising that biogeographically this region includes biotic elements from the marine provinces on both sides of it. This is clearly visible in our analyses, which show the highest rates of species turnover (β_sim) with respect to geographical distance, the mean temperature for August and the range in temperature in February. This is a form of regional biogeographical structuring called the temporal/spatial constraint model (Nekola et al., 1999). Similar conclusions with regards to the B-ATZ being an transitional area have been reached by Stephenson (1948), Bolton (1986), Bolton and Anderson (2004), and Mead et al. (2013). For marine macroalgae at least, this area cannot be recognized as a marine province as it does not meet the criteria for it to be classified as such (Spalding et al., 2007), i.e., it lacks cohesion, levels of endemism, and there is an absence of distinct abiotic features (with exception of such conditions periodically arising in False Bay due to prevailing mesoscale oceanographic conditions).

It was recently suggested that studies which associate marine macroalgal distributional patterns with broad-scale temperature gradients neglect to consider alternative or additional explanations, such as that offered by connectivity due to ocean currents (Wernberg et al., 2013). It is possible that seaweeds are similarly influenced by ocean currents around South Africa, and it is temping to suggest that the direction of turnover along the east coast is from north to south, and along the south coast from east to west, as this would coincide with the direction of the Agulhas Current. From a purely thermal point of view, the imprints of those ocean currents are certainly present at the coast in the zone that our seaweed data represent, but the physical processes that operate in that coastal zone are considerably more complex and decoupled from mesoscale influences (Schlegel et al., 2017). Personal observations by one of us (AS) recount dislodged individuals of the kelp, Ecklonia maxima, having washed ashore some 400 km east of the eastern-most distributional limit for the species in South Africa—this is against the direction of the current, which in this region is situated considerably far from the shore along the edge of the Agulhas Bank. Our data on the rate of species turnover as a function of the thermal difference between coastal sections (β_sim vs. d_E) within bioregions show an almost precise relationship between thermal difference and change in species composition (even though the relationship between d_E and geographical distance differs between the east and south coasts), and provide very strong support for temperature, rather than connectivity due to currents, as an overriding driver of seaweed community composition over large spatial scales. Within the context of Beijerinck's “Law” that “everything is everywhere but the environment selects” (De Wit and Bouvier, 2006), the only role that ocean currents can have in setting up biogeographical patterns is to cause everything not to be everywhere. At these scales, with or without the influence of ocean currents, temperature still selects. In other words, at broad spatial scales within the region encompassed by the ECTZ and the AMP, the seaweed flora is structured predominantly by niche-assembly processes driven by a thermal gradient.

However, we should point out that ca. 13.7% of the seaweed flora's turnover is also explained by some other broad scale spatial influences (MEM2, MEM3 and MEM5), which represent exogenous (unmeasured) environmental variables and autocorrelation. The importance of these relative to each other is unknown. Future work may find that some of these environmental influences may indeed be that of ocean currents, which operate at those broad scales, but our current interpretation does not support this notion. Rather, we suggest that the unknown influence is more likely to be neutral influences (Hubbell, 2001) manifesting at a reasonably broad scale. These may include stochastic events (e.g., Smale et al., 2011) or biotic processes caused by differences in species demography, dispersal or the influence of grazers, etc. Such processes are usually finer scaled, but, having said that, the entire west coast region that makes up the BMP is dominated by nestedness-resultant β-diversity, which is also symptomatic of neutral assembly processes.

Our approach throughout this paper hinges upon the biogeographical provinces and overlap regions defined using both seaweed and thermal data. Together, thermal and species gradients/patterns provide the most parsimonious explanation for the processes assembling the seaweed flora within the study region. The explanation is that thermal gradients set up β-diversity based on species turnover as a result of environmental filtering within the southern and eastern coastal sections, while along the west coast a very different kind of assembly process, nestedness-resultant β-diversity, emerges. We suggest that this disjunction results from a historical event that is currently still limiting connectivity between the Indian and Atlantic Oceans around the Cape Peninsula.

We also show how thermal metrics other than the annual mean temperatures can aid in the delineation of macroecological patterns. Some species are limited by cold and some by warm temperatures, and the constraining factor may differ at a species' southern and northern limits (e.g., van den Hoek, 1982a; Breeman, 1988). If these are important determinants, as we indeed show there are, other measures of climate should be used in our repertoire of explanatory variables. For example, the mean temperature for the coldest month, standard deviation, variation in annual temperature extremes, or temperature ranges (e.g., as used by Qian and Ricklefs, 2007; Leaper et al., 2011) may readily be extracted from time series of daily temperatures. These approaches, as we apply them here, may greatly improve macroecological and biogeographic studies and aid our ability to analyse broad-scale communities patterns and the processes that assemble them.

AS conceptualized the scope of the research reported in this paper, undertook all the numerical and statistical analyses, made the first round of interpretation, and did the bulk of the writing. JB and RA collected all the samples over the last 30 years, compiled the database of species distribution records for the region included in this analysis, contributed in equal part to the conceptualization of the research and the interpretation of the findings, and provided significant editorial input into penultimate and final drafts of the document.