Patterns in assemblage composition
In total, some 67,000 individual macrofauna, representing 160 species, were examined in 2,100 core samples from the Zostera beds during the study. These ranged from typical freshwater/dilute-brackish species such as Afrochiltonia capensis, Corallana africana and Melanoides tuberculata through to fully marine forms such as Gibbula cicer, Limaria tuberculata, Nebalia capensis and Parechinus angulosus. Ordination by nMDS of Bray–Curtis similarity data from the 23 intertidal sites suggested that four significantly different faunal clusters occurred in the system (ANOSIM R = 0.88; P < 0.0001) (Figs. 1and 2), an essentially similar pattern to that derived earlier using non-standardized (but fourth-root transformed) abundance data (Barnes 2013a). These represented: (i) the sandy mouth region immediately adjacent to the true mouth, (ii) the outer marine embayment, and (iii) the lagoon plus lower-estuary divisions of the main axial channel, and (iv) the fringing backwater-creek system of the smaller, saltmarsh-enclosed creeks and channels that separate the bay's two large islands (each c. 82–84 ha) from the mainland, together with sites in the upper estuary. Separation of the backwaters/upper-estuary sites from those along the main axial channel was the most marked, with a Bray–Curtis similarity between the two blocks of sites of only 20%, and the mouth region was an outlier within the axial channel. The three points of change along the longitudinal axis of the bay, however, were not marked by sharp faunal contrasts (Fig. 3). Indeed, SIMPER indicates that most (> 50%) of the differences are brought about by the relative abundances of just eight common and widespread species, the gastropod molluscs Hydrobia knysnaensis and 'Assiminea' capensis (dominant in region iv), Turritella and Alaba (dominant in i), and Nassarius (dominant in iii), and the polychaetes Prionospio (dominant in iii), and Caulleriella and Simplisetia (dominant in ii). Despite statistically significant regionalization, 32% of the species (representing > 75% of the total individuals) occurred in all four regions in more than token quantities (Table 1 lists the more numerous of these shared taxa, and Table 2 displays those characteristic of each region). Patterns of relative species abundance within the four regions did not differ (ANCOVA equality of means F = 0.17, P = 0.92; equality of slopes F = 0.59, P = 0.62) (Fig. 4), further indicating similarity between the different local assemblages. Number of species per sample did not vary across test areas of up to 1.5 ha at a given site, whether in the bay or in the lagoon (Table 3). The observation that the seagrass macrofauna of the brackish upper estuary did not differ from that in the fully saline, saltmarsh-enclosed backwater channels of the marine embayment is noteworthy and reinforces the earlier comments of Day (1959) and Barnes (1989) that so-called estuarine faunas may be as characteristic of sheltered areas of fully marine soft sediment as they are of regions subject to low salinity.
Major differences, however, did occur in the relative importance of infauna versus epifauna. Except at the lagoonal site 9, where the small biofilm-feeding cushion star Parvulastra exigua occurs in large numbers, the intertidal zone of the whole axial channel apart from the upper estuary is dominated by infaunal species (Fig. 5a), principally by polychaetes. From sites 1 to 15, the infauna comprised 68% (SE 3.7) of animals with no significant trend in their relative importance along the gradient (Sρ = 0.31; P = 0.26) (Fig. 5B). In contrast, the shores of the upper estuary and the marine backwater channels were dominated by epifaunal truncatelloid microgastropods, especially by 'Assiminea' capensis and Hydrobia knysnaensis, epifauna here comprising 64.4% of individuals. Only a few subtidal Z. capensis sites have so far been examined, but such areas are also overwhelmingly dominated by an epifaunal microgastropod, here by the cerithioid Alaba pinnae, although the importance of this species and hence of the subtidal epifauna in general decreases upstream so that epifauna and infauna contribute equally in the upper estuary (Fig. 5a). Thus in the bay region there is a transition at some LWS between a burrowing polychaete infauna and a seagrass-leaf-associated gastropod epifauna, and although upstream sub- and intertidal faunas are relatively similar, downstream in the bay they are markedly different (Fig. 6a). Few data are available to help explain the great downstream subtidal abundance of the epifaunal Alaba (a mean density of 28,000 m−2), although various studies have suggested that few fish consume significant numbers of shelled gastropods, even relatively small ones (McCormick 1998; Reynolds et al. 2018), not least because of their low nutritive value per unit intake (Vinson and Baker 2008). It is known that in South Africa, mugilids will take microgastropods (Whitfield and Blaber 1978; Whitfield 1988), but at Knysna mugilids do not characterize the dense sublittoral eelgrass beds favored by Alaba (Pollard et al. 2017). Several equivalent subtidal areas of seagrass in other continents are also dominated by species of Alaba, although Knysna is the only known such locality outside the tropics (Barnes and Claassens 2020). These other areas are of relatively high salinity which may help to account for the lesser importance of this gastropod in and near the upper estuary. Why the same suite of truncatelloid microgastropods dominates the otherwise contrasting habitats of the intertidal backwaters and upper estuary is not known for certain, but their common shelter (see paragraph above) is likely to be an important component.
With one exception, no evidence of any strong species interactions within any given site was forthcoming. The exception was the positive correlation between numbers of the ectoparasitic pyramidellid snail Sayella sp. and those of its probable host Hydrobia at the backwater site 'A' in Fig. 1 (Rs = 0.78; P < 0.00001). Such a parasite/host association is known from the western Atlantic (e.g., Hershler and Davis 1980), but although the pyramidellid concerned is a widely distributed animal, it is otherwise not recorded from Africa (GBIF 2020). That exception apart, however, in a large sample (325 cores) from the Kingfisher Creek seagrass (site 2 in Fig. 1), for example, Barnes (2013b) recorded 75 macrofaunal species at overall and mean densities of 2581 and 34 m−2, respectively. Considering the 34 relatively common species there that each attained a mean density of at least 10 m−2 (and together comprised 96% of the total individuals), all pairwise correlations of species abundance were very weak to non-existent (sensu Moore et al. 2018), positives averaging only Pρ = 0.069 (± 0.060 SD) and negatives Pρ = 0.047 (± 0.035 SD); and allowing for the familywise errors inherent in such a large correlation matrix (via Bonferroni correction), no negative correlations and only three positive ones were significant at a critical α of < 0.05 (between the polychaetes Simplisetia and Caulleriella, Glycera and Cirriformia, and between the polychaete Prionospio and the gastropod Nassarius). Equivalently, although qualitative co-occurrence patterns across the whole of the marine-influenced embayment at Knysna show deterministic structuring (Barnes and Elwood 2011), as indeed might be expected granted the location of the sampled sites in three distinct faunal regions (sandy mouth, marine bay, and backwater system), syntopic species within a single one of those regions did not differ from random co-occurrences (Barnes and Barnes 2014b).
In the absence of strong bioturbators such as Kraussillichirus kraussi (Callianassa kraussi in the older literature) from the majority of the system, faunal relationships between seagrass and bare sediment at Knysna are not the classic one of seagrass supporting the greater number of species and of individuals per unit area (Hemminga and Duarte 2000; Pillay et al. 2011; Hyman et al. 2019, etc.). To date studies have only concerned the outer marine embayment, but there seagrass macrofauna at a given site is more similar to those occurring in adjacent areas of bare sediment than either habitat is to other areas of the same type in the general region [Bray–Curtis faunal similarity between the two contiguous habitat types being a mean 0.58, whereas within-habitat-type similarity averaged 0.26 for the seagrass and 0.25 for the bare sediment (ANOVA F1,14 = 5.05; P < 0.05)] (see Fig. 6b). In general, seagrass beds supported lower, not higher, levels in half the metric comparisons in which there was a significant difference (Barnes and Barnes 2014a). Overall, faunal abundance was lower in seagrass in the ratio of 0.64: 1, while species density was indeed higher, but only by 1.13 to 1, with in large measure the higher numbers in the unvegetated sediments resulting from a quadrupled abundance of infaunal polychaetes, maybe because of the greater volume of available sedimentary habitat in the absence of eelgrass rootmass, although numbers of epifaunal crustaceans were 15 times less there (from a much smaller base). The same overall effect was not the case, however, in bare areas created by the death of seagrass following blanketing by the chlorophyte blooms described by Allanson et al. (2016) and Human et al. (2016). In these circumstances, the former seagrass sites clustered together, as did the same areas when de-vegetated, although macrofaunal abundance was again significantly lower in the former seagrass than it was in the replacement bare sediment (in a ratio of 0.62: 1) and again largely because of an increased number of polychaetes and decreased number of crustaceans in the unvegetated sediment (Barnes 2019a).
Knysna's marine embayment forms a natural harbor, has been in the past a busy port (Grindley 1985), and today supports several marinas, and hence it is one of the centers of ship-borne alien immigrant species in South Africa (Griffiths et al. 2009). Alien species of Boccardia, Polydora, Dipolydora, Pseudopolydora, Diopatra, Capitella, Desdemona, Ericthonius, Jassa, Monocorophium, Paracerceis, Elysia, Favorinus and Indothais all form part of its seagrass fauna, as do amphipods such as Cymadusa filosa, Melita zeylanica and Americorophium triaeonyx that are regarded by Robinson et al. (2005) and Mead et al. (2011) as being cryptogenic—to which could presumably be added Victoriopisa chilkensis. Relatively recently, these aliens have been joined by more northerly species spreading southward probably as a result of global warming. Smaragdia souverbiana, for example, is now a member of the subtidal seagrass fauna (Barnes and Claassens 2020). In the Knysna intertidal, Melanoides tuberculata has arrived and joined Cerithidea decollata (Hodgson and Dickens 2012) and Austruca occidentalis (formerly Uca annulipes) (Peer et al. 2015), the latter two in the adjacent saltmarsh or at the seagrass/saltmarsh interface.
Patterns in assemblage metrics along the axial gradient
As would be expected, the number of species at the 17 sites that were spaced along the system's longitudinal axis decreased with distance upstream (Sρ = −0.82; P < 0.0001; Fig. 7a), but the form of the decrease in species density suggests the occurrence of a step change within the general area of the lower estuary, with the downstream sites showing a considerable degree of uniformity of species density (Fig. 7b; Table 4). The points in Fig. 7 are based on the whole available 12-year dataset, and hence, the location of faunal and regional boundaries will have been blurred by temporal shifts, but an individual survey of macrofaunal animals along the axial channel in 2012 showed an almost identical (and sharper) feature (Barnes and Ellwood 2012) in the same general location. Comparison of data across different spatial scales shows that decline upstream in number of species when assessed per site (Fig. 7) is greater than when assessed per region (Table 4): Clearly, the bay and lagoon + lower-estuary regions are large and, as also would be expected, the total fauna contained in each is considerably in excess of that at any individual site.
Assemblage abundance per unit area (Sρ = -0.35; P = 0.16; Fig. 8a) and patchiness in assemblage abundance (Sρ = −0.20; P = 0.45; Fig. 8b), however, showed no significant change with distance upstream; indeed, degree of patchiness along the axial gradient was significantly unchanging (Barnes 2019b). Neither was there any significant relationship between number of species per site and overall assemblage abundance there (Sρ = 0.43; P = 0.08). However, significant relationships have been found between how patchy an individual species is and its occupancy and, to a lesser extent, its abundance: The more abundant and widespread the species, the less its patchiness, both in subtidal and in intertidal seagrass (Barnes 2019c, 2020), and both in interspecific comparisons (Barnes 2020) and intraspecifically (Barnes, in prep.) (Fig. 9). This suggests that the well-known macroecological abundance-occupancy pattern (e.g., He and Gaston 2003) can be extended into a patchiness-abundance-occupancy one, at least in this habitat type. As can be seen in Fig. 9, the slopes of the power laws relating logit occupancy to log patchiness in individual species are much more variable than those interspecifically in the different faunal regions; thus, while the interspecific occupancy-patchiness slopes representing different regions do not differ (ANCOVA F = 1.3, P = 0.3), the equivalent intraspecific slopes are heterogeneous (ANCOVA F = 4.9, P < 0.0001) with a further six of the dominants (including the epifaunal Alaba and Cymadusa, and infaunal Caulleriella and Salmacoma) not showing significant occupancy-patchiness relationships at all. This also indicates that disparate species together form assemblages with similar properties in the various regions. There were no discernable trends in either metric upstream, although the upper estuary did display the largest value of both β and R2.