The TMNP MPA was established in 2004 and occupies a transition between the Agulhas and Southern Benguela ecoregions, with Cape Point constituting a well-known coastal biogeographic break (Emanuel et al., 1992; Sink et al., 2019), across which gene flow is reduced (Von der Heyden, 2009; Wright et al., 2015). We surveyed shores on both the east and west sides of the Cape Peninsula, thus covering both ecoregions. Diverse rock types are present, but Cape Granite and Table Mountain Sandstone (TMS) predominate (McQuaid and Branch, 1985; Pfaff et al., 2019). Both rock types are represented in no-take ‘restricted’ areas and in harvested ‘controlled’ areas ( Figure 1 ). Site names, protection level, rock type, ecoregion and GPS coordinates are recorded in Supplementary Table A1 .

Sixteen sites were selected for study, representing two protection levels, two ecoregions and two rock types, with two replicate sites for each combination. Site selection was spatially randomized by placing replicate sites in different no-take areas. Wave action was standardized by selecting sites that fell in the range defined as ‘semi-exposed’ to ‘exposed’ by Steffani and Branch (2003), thus avoiding ‘sheltered’ or ‘extremely exposed’ shores.

Surveys of community composition were conducted at all 16 sites. Sampling was undertaken under permits issued by the Department of Forestry Fisheries and Environment and SANParks. For each site, four intertidal zones were identified based on their relative level and biotic indicators, termed the low, mid, high and top shore, at respective heights of approximately +0.2, +0.8, +1.2 and +1.5 m above low water spring tide (mean spring tidal range being 1.87 m). In each zone, 15 replicate 50×50-cm quadrats (divided into 25 cells of 4% cover each) were randomly placed and visually non-destructively sampled. All macro-organisms were identified and quantified as percentage cover for sessile organisms and as counts for mobile fauna (Bustamante and Branch, 1996; Blamey and Branch, 2009). To standardize, counts were later converted to percentage cover (Wieters et al., 2009).

Densities and sizes (shell lengths) of three limpet species commonly harvested in the region, Cymbula granatina, C. oculus and Scutellastra argenvillei (Eekhout et al., 1992; Bustamante et al., 1995; Branch and Odendaal, 2003), and three species of limpet rarely harvested there, S. cochlear, S. granularis and S. longicosta, were compared between harvested and no-take areas. Classification of species as ‘commonly harvested’ or ‘rarely harvested’ was based on our own observations, as reported in the results, and those of Lasiak (1991b) and Nakin and McQuaid (2014). These authors did report Scutellastra longicosta as being harvested on the south-east coast of South Africa, but there was no evidence of this in our study area. Measurements of densities and sizes were made separately in all four intertidal zones within the 15 replicate quadrats, with a minimum of 50 individuals being measured per species per zone per site.

Data for the shell lengths (measured with calipers to 0.1 mm accuracy) and relative abundances of shellfish (from counts of material that had been harvested) were obtained from South African National Parks field rangers’ records, and opportunistically from confiscated material and discarded shucked shells. These data allowed assessment of which species are most often harvested, and whether harvesters are selectively collecting larger individuals relative to the size spectrum available. The latter comparisons were restricted to C. granatina at two sites, and S. argenvillei at one site, as it was only there that sample sizes of harvested limpets were large enough for rigorous statistical comparisons.

To compare the mean sizes of Cymbula granatina, individuals collected by harvesters at Wireless Point and Kommetjie with those present on the shore at these respective sites and, likewise, for Scutellastra argenvillei at Wireless Island, Student’s t-tests were used. If assumptions of normality and equality of variance were not met, Mann-Whitney U tests were applied.

Univariate GLMs were employed to test for harvesting effects on densities and sizes (i.e., shell lengths) of individual limpet species. Not all species were present in both ecoregions or for some ecoregion-rock type combinations, therefore the data were analyzed separately per ecoregion and rock type. Individual GLMs included factors Protection level (as proxy for harvesting effects) and Site nested within Protection level. Significance levels were Bonferroni-adjusted to account for type-I error inflation associated with multiple tests of a variable.

Multivariate analyses to determine harvesting effects on the entire rocky shore community were conducted separately for each intertidal zone representing a unique biotope, i.e. low, mid, high and top zones. To meaningfully compare community composition between the two ecoregions comprising different species but comparable functional groups, species within each quadrat were grouped into functional groups based on criteria specified below ( Supplementary Table A2 ). Grouping may obscure the effects for rare species, but reduces statistical challenges associated with numerous dependent variables (Warton et al., 2012; Warton et al., 2015). Algal grouping was based on morphology, productivity and susceptibility to herbivory (Steneck and Dethier, 1994). Invertebrates were grouped by trophic category and foraging strategy (Wieters et al., 2009). Herbivores were divided into ‘grazers’ that crop algae, ‘trappers’ snaring drift algae or fronds (individuals of the limpets Cymbula granatina > 30 mm and Scutellastra argenvillei > 40 mm), and ‘gardeners’ cultivating patches of algae.

Percentage cover data of functional groups were analyzed with four-way multivariate GLMs using the mvabund package in R (Wang et al., 2012). The factors included in the models were Ecoregion, Rock Type, Protection (all fixed and crossed) and Site (random, nested in Rock Type). The nature of any significant differences was statistically explored using post-hoc pairwise multivariate comparisons included in the PERMANOVA toolbox in PRIMER; multidimensional scaling (MDS) ordination plots were used to visualize and interpret the community patterns. SIMPER analyses explored which functional groups were responsible for observed differences between harvested and no-take areas of the MPA.

Statistical analyses were performed in R (R Core Team, 2019), using the car package (Fox and Weisberg, 2019) and mvabund package (Wang et al., 2012). Pairwise PERMANOVA, MDS and SIMPER analyses were conducted with PRIMER V7 (Clarke and Gorley, 2015).

Harvesters at Kommetjie, Wireless Point and Wireless Island targeted three species of limpets, together with the mussels Choromytilus meridionalis and Mytilus galloprovincialis (pooled as ‘mussels’ as rangers’ records treated them as a group) ( Table 1 ). Cymbula granatina was the most harvested limpet, followed by C. oculus. Scutellastra argenvillei was collected at Wireless Island only, while S. cochlear and S. granularis were never taken.

Harvesters focused on large individuals of C. granatina at both sites examined, with the difference between the mean sizes of harvested individuals and those in the natural populations being significant (Mann Whitney U test, p ≤ 0.001 in both cases; Figures 2A, B ). The smallest harvested individuals were respectively 40-45 mm and 60 mm, whereas natural populations had individuals as small as 5-20 mm. Although a similar pattern emerged for S. argenvillei, with harvested individuals falling in the range 60–100 mm and the natural population spanning 5–100 mm, the difference in mean sizes was in this case not significant ( Figure 2C ; Mann Whitney U test, p = 0.3094).

On sandstone (TMS), all three commonly harvested limpets Cymbula granatina, C. oculus and Scutellastra argenvillei had significantly lower densities in harvested areas than in no-take areas. Harvesting effects were significant despite significant variability between replicate sites in most cases and the reduced power of Bonferroni-adjusted analyses. Densities on granite were low regardless of the level of protection and showed no harvesting effects or - in the case of C. oculus - greater densities in Agulhas harvested areas ( Supplementary Table A3 post-hoc ; Figures 3A–C ).

In contrast, densities of the rarely harvested limpets showed inconsistent or no effects of protection. Scutellastra longicosta densities did not differ significantly between levels of protection on either rock type ( Figure 3D ). Scutellastra cochlear densities were lower in harvested areas on TMS in both ecoregions, while no or reversed protection effects existed on granite ( Figure 3E ). Scutellastra granularis also showed inconsistent harvesting effects, with greater densities in protected than harvested sites on granite in the Southern Benguela, whereas no significant or negative effects of protection were evident at the other three combinations of ecoregion and rock type ( Figure 3F ).

Of the commonly harvested limpets, Cymbula granatina was significantly larger inside the no-take than the harvested areas, regardless of rock type ( Supplementary Table A4 ; Figure 4A ). Cymbula oculus was counterintuitively larger in the harvested areas, but only on sandstone shores ( Figure 4B ). For Scutellastra argenvillei, post-hoc sizes were larger in the no-take than harvested areas, but on sandstone shores only; protection had no effect on granite, where its densities were low ( Figure 4C ).

Of the rarely or never harvested limpets ( Supplementary Table A4 ), the size of S. longicosta, was not affected by either protection level or rock type ( Figure 4D ). For S. cochlear, sizes were greater in harvested than no-take areas in the Agulhas ecoregion on granite, but protection had no effect elsewhere ( Figure 4E ). For S. granularis, sizes were significantly larger in harvested areas in the Benguela on both TMS and granite and on Agulhas TMS, but not significantly different on Agulhas sandstone ( Figure 4F ).

MDS plots of the four zones on the shore exhibited a clear biogeographic distinction between the Southern Benguela and Agulhas ecoregions ( Figure 5 ). Differences also existed between replicate sites ( Supplementary Table A5 ). The distinctions between protection levels and sites emerged more clearly when the data were disaggregated to separate the ecoregions and rock types and bring to the fore the protection and site effects ( Supplementary Figure A1 ). Over and above the regional differences, clear harvesting effects existed, but they differed among intertidal zones, ecoregions and rock types. In the low intertidal zone ( Supplementary Figure A1A ), dissimilarities between harvested and no-take areas existed for TMS but not granite ledges where site differences dominated; and were more pronounced in the Southern Benguela than the Agulhas ecoregion. Similarly, in the mid intertidal zone ( Supplementary Figure A1B ) harvesting effects existed only for TMS and not granite shores, in both ecoregions ( Supplementary Figure A1B ). In the high intertidal zone, harvesting effects were most prominent on Southern Benguela TMS, while site differences overrode harvesting differences in all other cases ( Supplementary Figure A1C ). On the top intertidal zone, site differences were pronounced, overriding harvesting effects in all cases ( Supplementary Figure A1D ).

SIMPER analyses ( Figures 6 – 9 ), coupled with the MDS ordinations discussed above, indicated that marked dissimilarities existed between no-take and harvested areas, but varied among ecoregions, rock types and intertidal zones. On the low shore ( Figure 6 ), ephemeral algae were consistently more abundant on harvested shores, most obviously in the Southern Benguela ecoregion. Corticated algae followed suit and distinguished harvested sites in all cases except Southern Benguela granite. Algal crusts were more prevalent on harvested shores on granite in both ecoregions, but on sandstone the reverse was true. Herbivores (particularly trappers and gardeners) were more abundant at no-take sites.

Within the mid intertidal zone ( Figure 7 ), corticated algae were more abundant on harvested shores (with the exception of Southern Benguela granite), and harvested sites housed larger amounts of ephemeral algae than no-take sites in the Southern Benguela ecoregion. Algal crusts were, contrary to the low-shore pattern, more abundant on no-take shores, being distinguishing members of no-take shores in three cases. Filter feeders also distinguished no-take sites in three of the four comparisons.

In the high shore ( Figure 8 ), ephemeral algae were prevalent at the majority of harvested sites, distinguishing these shores in three of four cases. Corticated algae, however, reversed the trend evident lower on the shore and were more abundant at, and distinguished, no-take sites in two instances. Of the herbivores, trappers yielded no consistent results with respect to protection level, but greater grazer abundances were found on protected shores in three cases, two of which diagnostic of no-take shores. Filter feeders yielded mixed results, being more abundant on no-take shores in two cases, but less abundant there in the other two cases.

On the top shore ( Figure 9 ) ephemeral algae were more abundant in no-take areas, with one exception. Corticated algae remained more common in harvested sites in three out of four cases. Grazers– dominated by the rarely harvested limpet S. granularis - attained greater densities on harvested shore in all cases bar Southern Benguela granite. In three of four instances, filter feeders distinguished and achieved greater abundance in no-take areas.

SIMPER identified groups that distinguished ecoregions ( Figure 10 ). Most differences were zone-specific, but ephemeral algae consistently distinguished the Southern Benguela, as did crustose algae in all zones bar the low shore, where they distinguished the Agulhas ecoregion, associated with a greater abundance of gardening limpets in that zone and region. Grazers were abundant in only the high and top zones, where they distinguished the Southern Benguela. Articulated coralline algae were abundant only in the low shore, where they distinguished the Agulhas. Corticated algae distinguished the Agulhas in the mid and high shore where their abundance was greatest but distinguished the Southern Benguela in the low and top zones. Filter feeders distinguished the Agulhas in the high and top zones where they were most abundant, but the Southern Benguela in the low and mid zones.

Previous studies have demonstrated that shellfish gatherers have preferences for particular species (Durán et al., 1987; Airoldi et al., 2005; Nakin and McQuaid, 2014). This leads to a small number of species dominating the catches (Hockey and Bosman, 1986; Lasiak, 1991b; Lasiak, 1993b; García-Escárzaga et al., 2017). For example, Durán et al. (1987) showed that in Chile shellfish gatherers concentrate their harvest on Concholepas concholepas and Fissurella crassa. Our study revealed that, unlike the east coast of South Africa where the mussel Perna perna dominates catches by shellfish gatherers (Hockey and Bosman, 1986; Hockey et al., 1988; Lasiak, 1991a; Lasiak, 1992), the limpet Cymbula granatina was the most exploited intertidal shellfish in Table Mountain National Park. This could be attributed to its large size, high densities, accessibility on relatively sheltered shores (Eekhout et al., 1992; Bustamante et al., 1995; Blamey and Branch, 2009), and comparatively low strength of attachment (Branch and Marsh, 1978). All these characteristics lead to easy harvesting of this species on rocky shores, making it vulnerable to overfishing (Lasiak, 1991b). Indeed, there is a long history of exploitation of species such as this in southern Africa, extending back at least 100 000 years in the archaeological record (Jerardino et al., 2008; Parkington et al., 2013; Will et al., 2016; Jerardino, 2021). Cymbula oculus was also harvested at three Southern Benguela sites, but in smaller amounts, reflecting the fact that its densities there are lower than those of C. granatina, reflecting its prevalence in the warmer waters of the Agulhas Ecoregion (Branch, 1971). By contrast, Scutellastra argenvillei, which inhabits semi-exposed to exposed areas (Branch and Steffani, 2004), was recorded as being harvested at only one site in any numbers. The absence of S. longicosta, S. cochlear and S. granularis from any catch records supports the view that they are rarely harvested in the region we investigated, although S. longicosta is harvested on the more easterly shores of South Africa (Hockey et al., 1988; Nakin et al., 2012; Nakin and McQuaid, 2014; Nakin and McQuaid, 2016).

The selection of larger individuals of C. granatina by shellfish gatherers upholds our first hypothesis to this effect and supports other findings that shellfish gatherers target large individuals (Durán et al., 1987; Lasiak, 1991a; Lasiak, 1992; Jimenez et al., 2011; Alexander and Gladstone, 2012). This usually leads to reductions in the mean sizes of the individuals in the natural populations (Eekhout et al., 1992; Jimenez et al., 2011), with profound effects on reproductive output (Branch, 1975; Branch and Odendaal, 2003).

Exclusion of shellfish gatherers from MPAs often leads to greater densities of commonly harvested species inside no-take areas relative to harvested areas (Moreno et al., 1984; Hockey and Bosman, 1986; Moreno et al., 1986; Hockey et al., 1988; Edgar and Barrett, 1999; Alexander and Gladstone, 2012; Bednar and Trulio, 2017). For example, higher densities of the limpet Cellana tramoserica were found in the Bouddi National Park Marine Extension relative to unprotected shores (Alexander and Gladstone, 2012). The same was true for Patella ulyssiponensis in La Palma MPA (López et al., 2012), for Cymbula oculus in Dwesa Nature Reserve in South Africa (Branch and Odendaal, 2003), and for Fissurella picta and Fissurella limbata in the Marine Reserve of Mehuin in Chile (Duarte et al., 1996). These increased densities were all attributed to protection. Our results produced similar patterns for the commonly harvested limpets C. granatina, C. oculus and S. argenvillei, particularly on sandstone where these species are most abundant.

The conclusion that protection is responsible for increased densities of harvested species is strengthened by the fact that the rarely harvested limpets S. granularis, S. longicosta and S. cochlear had equivalent densities at both protection levels in most instances (six out of ten comparisons). This difference is what would be expected if it is harvesting, rather than any other factors, that accounts for the decline in harvested species in sites that are not protected. Of the species that are not harvested, in the case of S. longicosta, no differences in density existed inside and outside no-take areas. For the other two species that are not harvested, there were, however, departures from this pattern, but they still amounted to an absence of any protection effect, or at least inconsistencies in its effects. For S. cochlear, densities inside and outside no-take MPAs were the same in two cases, greater in no-take areas in one case, and greater in harvested areas in one case. Scutellastra granularis attained high densities inside no-take areas in only one instance. It is possible that that its abundance is influenced more by the arrival of the alien mussel M. galloprovincialis (Griffiths et al., 1992; Branch et al., 2008; Branch et al., 2010) than any other factor, as this mussel greatly boosts this limpet’s recruitment (Van Erkom Schurink and Griffiths, 1990). In the Agulhas Ecoregion, S. granularis density is also often restricted by barnacles such as Tetraclita serrata and Chthamalus dentatus (Branch, 1976; Nakin and McQuaid, 2014), which are substantially less abundant in the Southern Benguela than the Agulhas (Boland, 1997). There were significant and complex interactions among ecoregion, rock types and protection level for S. granularis and S. cochlear as a result of these variable outcomes, which was not unexpected in the absence of any likely direct effect of protection. The absence or inconsistent outcomes of any protection effect observed for rarely harvested species also concurs with previous studies. For example, another rarely harvested species, Cellana capensis, had higher densities in harvested than no-take areas on the southeast coast of South Africa (Lasiak, 1993a).

Protection of even harvested species by way of no-take areas is, however, not always effective. An absence of any protection effect was reported by Nakin and McQuaid (2014) for limpets in Dwesa Nature Reserve, attributable to illegal harvesting in the reserve and differences in the intensity of harvesting outside the reserve. Similarly, Coppa et al. (2012; 2015) associated the decline of Patella ferruginea in Italy’s Mal di Ventre MPA with ineffective enforcement of regulations. Although C. oculus had high densities inside sandstone no-take areas, its overall abundance appears to have dramatically decreased in the TMNP MPA over the past 11 years, from 12.35 m^-2 to 8 m^-2 (Maneveldt et al., 2009). The lack of protection effect on the granite rocky shore might be due to higher mortality associated with dislodgement of limpets from granite rocks, as it has been suggested that they cannot attach as firmly to this rock type (McQuaid et al., 1985). Thus, low densities of C. oculus on granite rocky shores may not reflect a failure of protection inside no-take areas of the MPA, but rather an influence of habitat type on the density of the species.

We had anticipated that there would be differences in the abundances of limpets between the two ecoregions spanned by our study. This manifested itself in two ways. First, some species were either confined to, or substantially more abundant, in one of the two ecoregions. That was the case for C. granatina and S. argenvillei (predominantly Southern Benguela), and C. oculus and S. longicosta (predominantly Agulhas). For those species that occurred in both ecoregions, we had expected that there might be greater abundances in the Southern Benguela than the Agulhas Ecoregion because of the greater productivity on the west coast and the fact that greater biomasses of grazers have been recorded there (Bustamante et al., 1995). That was, however, not the case. There were no obvious differences in abundances of either S. granularis or S. cochlear between ecoregions.

The densities of S. longicosta that we recorded were magnitudes less than those recorded in Dwesa Nature Reserve (Nakin and McQuaid, 2014). This might reflect that the species is reaching the western limit of its distribution range in the TMNP, resulting in low densities (Hidas et al., 2010).

Shellfish gatherers generally target large individuals, as noted above for Cymbula granatina and Scutellastra argenvillei. This can truncate the size distribution of populations in harvested areas (Sagarin et al., 2007; Jimenez et al., 2011; Alexander and Gladstone, 2012; Coppa et al., 2012; Fenberg et al., 2012; Coppa et al., 2015; Bednar and Trulio, 2017), as has been reported for Cymbula oculus (Branch and Odendaal, 2003), Cellana capensis (Lasiak, 1993a), Lottia gigantea (Sagarin et al., 2007; Lucas and Smith, 2016), and many other species, including fish (Marra et al., 2017; Heyns-Veale et al., 2019).

In our results, harvesting had variable effects on the sizes of key species. The most frequently harvested species, C. granatina, consistently had larger sizes inside protected vs. harvested areas, on both sandstone and granite rocks. The second frequently harvested species, Scutellastra argenvillei, yielded the same outcome on sandstone, but not on granite, possibly because densities there were much lower, and no harvesting of this species was recorded at sites with granite rocks. However, C. oculus, which is also harvested, displayed a weak effect of protection on size, being larger outside no-take areas than inside on sandstone, and not significantly different on granite. Effects of protection may thus be species-specific, habitat-specific and region-specific (Nakin and McQuaid, 2014). The existence of larger individuals in protected areas enhances reproductive output because of their greater fecundity (Branch, 1975; Branch and Odendaal, 2003; Orozco et al., 2013; Zarrouk et al., 2016). For species like C. oculus that are protandric hermaphrodites and change sex from male to female at an intermediate size, a reduction of mean size will also affect population reproductive output as it will skew sex ratios in favor of males (Branch and Odendaal, 2003).

In contrast to the three harvested species of limpets, all three rarely harvested species, S. longicosta, S. granularis and S. cochlear, conformed to expectation and showed either no effects or no consistent effects of protection on their mean sizes. The size of S. longicosta was not related to the level of harvesting, and the same was true in three of the four comparisons possible for S. cochlear. For S. granularis, size was larger at harvested sites – significantly so in three of the four cases.

Similar findings were reported for S. granularis at Dwesa Nature Reserve, with larger sizes outside than inside the reserve (Nakin and McQuaid, 2014). The occurrence of larger individuals in harvested areas may reflect reduced competition due to the removal of harvested competitors, although that has not yet been demonstrated empirically. Lasiak and White (1993) have, however, shown that high densities of the limpet Cellana capensis reduce microalgal availability, with implied competitive effects for other species.

The sizes and densities of key limpet species were forecast to be greater in the Southern Benguela than the Agulhas Ecoregion because of greater productivity there as a result of upwelling (Bustamante et al., 1995). This was, however, not the case.

There were considerable differences in community composition between the harvested and no-take areas, with relatively greater abundances of filter feeders, grazers and trappers inside no-take areas. Reductions of these groups in harvested areas have frequently been reported (Moreno et al., 1984; Hockey and Bosman, 1986; Lasiak and Field, 1995; Sharpe and Keough, 1998; Lasiak, 1998; Lasiak, 1999; Ceccherelli et al., 2005; Sink et al., 2010; Ceccherelli et al., 2011; Barbiero et al., 2011). Their depletion at harvested sites was likely responsible for the increased abundances of algae due to diminished grazing and/or competition for space. Similar findings have been reported for Dwesa-Cwebe, Hluleka and Mkambati Nature Reserves, and in Tsitsikamma MPA, on the east and south coasts of South Africa respectively, where community composition inside these protected areas is dominated by commonly harvested species such as Perna perna, Haliotis spadicea, Scutellastra barbara, Cymbula oculus and Cymbula miniata (Hockey and Bosman, 1986; Lasiak and Field, 1995; Dye, 1998; Lasiak, 1998; Lasiak, 1999; Hanekom, 2011). Comparable outcomes have been noted worldwide: the commonly harvested urchin Heliocidaris erythrogramma is a characteristic and distinguishing taxon for no-take areas in Tasmanian marine reserves (Barrett et al., 2009). Three other harvested taxa, Tridacna spp., Trochus niloticus and Turbo spp., are the major groups distinguishing between protected and no-take areas in New Caledonia (Jimenez et al., 2015; Jimenez et al., 2016). The fact that harvested filter feeders, grazers and trappers were the major functional groups distinguishing between harvested and no-take areas in our data strengthens our conclusion that protection influenced the community composition, as reflected in other studies (Lasiak and Field, 1995; Lasiak, 1998; Lasiak, 1999). Protection in no-take areas often results in a mosaic of communities, with patches of filter feeders, grazers and trappers, while adjacent harvested areas tend to have extensive uniform mats of algae, as described for Dwesa-Cwebe and Hluleka Nature Reserves and Tsitsikamma MPA (Hockey and Bosman, 1986; Lasiak and Field, 1995; Dye, 1998; Lasiak, 1998; Hanekom, 2011).

Whilst not the primary focus of this study, community composition also differed in a consistent and significant manner between ecoregions, which led to frequent interaction effects between protection and ecoregion. Our results agree with the biogeographic findings of Bustamante et al. (1995) and Bustamante and Branch (1996) that community composition is markedly different between the Agulhas and Southern Benguela ecoregions, which is the most clear-cut biogeographic break along the coast of southern Africa (Emanuel et al., 1992). This is associated with greater nutrient availability and lower temperatures in the Southern Benguela (McQuaid et al., 1985; Emanuel et al., 1992; Bustamante et al., 1995). In our results, the Southern Benguela displayed higher abundances of ephemeral algae, grazers and trappers, probably as a result of upwelling boosting nutrient availability and primary productivity (Bustamante et al., 1995; Bustamante and Branch, 1996).

The presence of larger quantities of algal crusts inside some of the no-take areas probably reflects trophic cascades and competitive interactions among functional groups, since this group has been shown to positively related to the abundance of grazers, which deplete macroalgae (Blamey and Branch, 2009) that would otherwise overgrow crusts. Harvested areas frequently had the highest percentage cover of corticated and ephemeral algae, which signals an altered community composition associated with lower densities of grazers (Bustamante et al., 1997; Lasiak, 1998; Barbiero et al., 2011).

The average levels of similarity (70.9%) in the Agulhas Ecoregion were equivalent to those considered to be converged communities (Sink et al., 2010; Jimenez et al., 2012). In addition, both protection levels were characterized by similar functional groups and dominated by algae, which are usually typical of areas experiencing high harvesting pressure (Hockey and Bosman, 1986; Lasiak and Field, 1995; Lasiak, 1998; Barrett et al., 2009; Sink et al., 2010). Thus, the dominance of some groups of algae in no-take areas in the Agulhas Ecoregion contradicts early findings in other MPAs along the south and east coasts of South Africa, such as Dwesa, Hluleka and Mkambati Nature Reserves (Hockey and Bosman, 1986; Lasiak and Field, 1995; Lasiak, 1998) and Tsitsikamma MPA (Hanekom, 2011), where commonly harvested consumers dominate and characterize no-take areas, with algae being less abundant (Lasiak and Field, 1995; Lasiak, 1998; Hanekom, 2011). This may, however, reflect the influence of upwelling in the Table Mountain National Park MPA, which supplies nutrients and thus facilitates algal growth and productivity, while the contrary findings were from more oligotrophic regions of the country, where grazing may have a more severe effect on algal abundances.

In the Southern Benguela Ecoregion, differences between harvested and no-take areas were more obvious (D = 39.85–72.12%), reflecting greater divergence (Sink et al., 2010; Jimenez et al., 2012) than in the Agulhas Ecoregion. There, communities reflected conventional expectations to a much greater degree, especially on sandstone shores, where corticated and ephemeral algae were prevalent in harvested situations, associated with the depletion of herbivores.

In terms of individual species, one striking absentee from the list of characteristic species of the Agulhas Ecoregion was the indigenous brown mussel Perna perna, which was abundant in False Bay at the southeast limits of the Agulhas Ecoregion in the 1980s but had diminished by 2011 (Griffiths and Mead, 2011; Reimers et al., 2014), and was never recorded in our study. Changes in its abundance and presence in this region are likely associated with cooling of the waters and a southeasterly contraction of this species’ range (Pfaff et al., 2019). Replacing it, the alien Mediterranean mussel Mytilus galloprovincialis has become a dominant element of all the shores studied.