RESULTADOS Y DISCUSION Aspectos estructurales del aparato bucal

biodiversity

Abstract

Habitat classifications define habitat types that act as surrogates for biodiversity patterns.

These classifications may be derived from actual data or expert opinion and may include physical variables and/or biological variables. The aims of this chapter were firstly to determine the validity of a data-derived seascape and an expert-derived habitat classification; secondly to compare the performance of these classifications to each other and against a classification that was based on both biological and abiotic data (biotopes);

and lastly to determine the best classification. The Canonical Analysis of Principal Coordinates (CAP), a constrained ordination, was utilised to determine whether the macro-infaunal assemblage of each sample was consistent with the assigned habitat in each classification. The fit of each habitat classification with the macro-infaunal data was measured as allocation success. Both the seascapes and expert-derived habitats had high allocation success, however, the expert-derived habitat classification (93-94 %) performed marginally better than the seascape classification (89-92 %). These two classifications also performed well compared to the biotope classification (98 %). This study indicated that there is a trade-off between capacity requirements and cost, and the performance of habitat classification systems as surrogates of biodiversity patterns. The best performing habitat classifications are those that include biological data such as species composition and abundance. But these classifications are also the most expensive to develop due to the cost of biological surveys. Therefore it is recommended that, where possible, habitat classifications with biological data should be used as surrogates for biodiversity patterns. Therefore the biotope map, based on macro-infaunal community patterns and their driving processes, the best performing biotope classification, was developed for the South African west coast to represent biodiversity patterns for inclusion in systematic conservation planning. However, it is imperative that surrogates always be tested against biological data to determine their validity.

132

Introduction

Biodiversity surrogates are virtually always utilised in conservation planning (Sarkar and Margules 2002). This is mainly due to the lack of sufficient biological data for entire planning regions, since such data are generally biased towards areas that are easily accessible (Pressey 2004). In the marine environment, biological data are biased towards the shallow marine ecosystems with less or very little data available for the deeper ocean (Griffiths et al. 2010, Heap et al. 2010, Kenchington and Hutchings 2012). The cost (i.e.

financial and time) of gathering the amount of data required to produce maps of continuous biological data is exceptionally high, therefore surrogates (either biotic or abiotic) that represent patterns of biodiversity (i.e. target features) in a cost-effective way are utilised (McArthur et al. 2010). These surrogates are used for, amongst others, establishing and testing the effectiveness of marine protected areas (Stevens and Connolly 2005, Currie et al. 2009).

Biological surrogates refer to higher-level taxa (e.g. families or phyla), cross-taxa (e.g.

patterns in fish diversity representing patterns in epibenthic macrofaunal communities) or subset taxa (e.g. patterns in polychaetes representing macrofaunal diversity pattern;

Karakassis et al. 2006, Wlodarska-Kowalczuk and Kedra 2007, Mellin et al. 2011). In homogenous environments such as marine unconsolidated sediments, the higher-level taxa are the most effective biological surrogates for multivariate macrofauna data (Mellin et al. 2011). Abiotic surrogates refer to physical variables or combinations thereof that correlate with biological data (e.g. Ward et al. 1999, Post et al. 2006, Van Wynsberge et al. 2012, Shokri and Gladstone 2013).

The correlations between abiotic variables, used to construct physical surrogates, and biological data such as species richness, distributions or assemblages, are scale-dependent. Salinity, temperature and oxygen concentration may be strong predictive variables, but they operate at large spatial scales (1-100s km) so may be less useful at a local scale (Meynard and Quinn 2007, Currie et al. 2009, McArthur et al. 2010).

Productivity and sediment characteristics have complex relationships with benthic fauna and have high variability at the fine scale (1 m to 10s km), although they too may have strong relationships with biodiversity (Sanders 1968, Stevens and Connolly 2004, Post et al. 2006, Svensson et al. 2007, Anderson et al. 2013). Depth, a variable that does not have a direct physiological effect on fauna, is a consistent predictor of biological data at various scales (e.g. Grassle and Maciolek 1992, Karakassis and Eleftheriou 1997, Bergen

133

et al. 2001, Post et al. 2006, Jayaraj et al. 2008, Renaud et al. 2009, Huang et al. 2011, Przeslawski et al. 2013).

Habitat classifications combine physical and/or biological variables that operate at different spatial scales into surrogates, which could be ‘habitat types’. These habitat types have been shown to be very effective surrogates representing sandy beach biodiversity patterns where beach morphodynamic types distinguish macrofaunal communities (Defeo and McLachlan 2005, 2011). Recent studies have also investigated the utility of habitat classifications as surrogates for biodiversity in various marine ecosystems (Mumby et al.

2008, Przeslawski et al. 2011, Van Wynsberge et al. 2012, Shokri and Gladstone 2013, Törnroos et al. 2013). Habitat types represented biodiversity well for some ecosystems, for example the unconsolidated sediments (Przeslawski et al. 2011) and coastal sediment/seagrass habitats (Törnroos et al. 2013), but poorly in others, for example estuaries (Shokri and Gladstone 2013).

Habitats, which by definition are spatial units, can be identified and mapped based on physical variables, biological measures or expert-opinion but most often habitat classifications and maps represent a combination of these methods. Data-derived habitats may be based on physical processes alone (as in seascapes) or in combination with biological data (as in biotopes; Costello 2009). Seascapes, a ‘top-down’ approach to defining habitats, are identified based on consistent oceanographic and physical features of the environment that are relatively easy to observe and operate over large scales typically using remote-sensed GIS layers (Roff and Taylor 2000). Biotope classifications, a

‘bottom-up’ approach to defining habitats, incorporate both biological communities and their physical habitats (Connor et al. 2004, Olenin and Ducrotoy 2006). These biotopes may be identified either by multivariate community analyses related to physical variables and scaled to the area of interest (Post et al. 2006), or by determining the most common species in a particular physical habitat (Mumby et al. 2008). Both biotope and seascape classifications and maps are data-intensive. However, generic physical data for seascape attributes (such as sea surface temperature or chlorophyll a concentration) are less expensive to obtain (e.g. remote sensing data are frequently available on internet databases), whereas the collection of biological data is more expensive, more time-consuming and requires taxonomic expertise. Habitat maps derived from expert opinion are mainly used where proficiency exists but data are sparse (Costello 2009). Expert-derived habitats may be as informative as data-Expert-derived habitats and can be classified at a fraction of the cost (Costello 2009). Although habitat classifications may be derived by

134

these various methods, they have not been compared for a single dataset to determine the best surrogate classifications.

The aim of this chapter was three-fold. Firstly, it aimed to test the accuracy with which each of two habitat classifications – namely a data-derived seascape and the expert-derived South African National Marine and Coastal (SANMC) habitat classification – reflect macro-infauna biodiversity on the South African west coast. Secondly, this chapter aimed to compare the efficiency of these habitat classifications to each other and to the biotope classification that incorporates macro-infaunal data and its physical drivers (Chapter 4).

Lastly, the habitat classification that best fit the macro-infaunal diversity was determined in order to produce a biodiversity surrogate map for inclusion in subsequent chapters.

Methods

Habitat classifications are scale dependent (Costello 2009), so the habitats under investigation were based on similar scales i.e. mega-habitat scale which operate over 1-100 km (following Greene et al. 1999) and therefore comparable. Further, the data layers generated relied heavily on the data from the 48 sites (Fig. 5.1) which were sampled by beach, suction or grab sampling as explained in Chapter 4 of this thesis. The resultant species abundance and biomass data were used in further analyses.

135

Figure 5.1: Stations sampled along the west coast of South Africa. The insert indicates the region of South Africa magnified in the larger map. Solid circles indicate numbered stations where 1-6 replicates were obtained, while open circles indicate sites where sampling efforts were unsuccessful due to sea conditions or hard grounds.

136

Figure 5.2: Maps representing (a) Seascapes, (b) Expert-derived habitats and (c) Biotopes. Classified (all coloured symbols) and outlier (δ) sites are indicated on the maps. SWC=South-western Cape and SB=southern Benguela.

137