This study investigates the effects of forest fragmentation and matrix type on the diversity and distribution of epigaeic invertebrates in selected Afromontane forests of the KwaZulu-Natal midlands. The consequences of area reduction and matrix transformation, due to commercial forestry, for epigaeic invertebrate diversity, and the role of small forest fragments in their conservation, are particularly examined. Data are used to provide guidelines for the management and conservation of Afromontane forest and adjacent land use. Epigaeic invertebrates were sampled using pitfall-trapping quadrats set along gradsect lines in eleven Afromontane mistbelt mixed Podocarpus forests, five of which were surrounded by commercial pine, and five surrounded by the natural grassland matrix. The two sets of five forests were approximately matched in pairs by area (0.5ha to 215ha) and compared for differences in the epigaeic invertebrate communities to determine potential effects of fragmentation and landscape transformation on these communities. A large tract of forest (Leopards Bush Private Nature Reserve, 705ha, hereafter Leopards Bush) situated within the continuous Karkloof forest, served as the control. Trapping intensity increased with forest area. Invertebrates were identified to morphospecies, and where possible, further identification was carried out by specialists. A total of 61 282 epigaeic invertebrates, representing 168 morphospecies, were collected. There was a significant positive species-area effect evident in the Balgowan complex (grassland matrix), but not in the Gilboa complex (pine matrix). There was thus a confounding influence of the pine matrix in the Gilboa, since the matrix represented the single most important difference between the two forest complexes. In afforested landscapes it is thus crucial to conserve the largest intact forest fragments to preserve overall epigaeic invertebrate species richness. Density compensation was evident in the Balgowan complex but not in the Gilboa complex. Recolonisation dynamics may play a small role in the regional persistence of a species within forest in a pine matrix, and persistence would be ensured by the preservation of a small number of large forest fragments containing large, extinction-resistant populations. The establishment of ecologically functional grassland corridors (i.e. wide enough to maintain "natural" ecological processes) between mistbelt mixed Podocarpus forest fragments would facilitate dispersal of epigaeic invertebrates between forest fragments. The results of this study indicated that edge effects were experienced deeper in the forests than the expected 32m suggested by Kotze and Samways (1999) and thus any changes to the epigaeic invertebrate communities induced by edge effects had already taken place inthese forests. Cluster analyses revealed that twenty-two percent (n = 37) of invertebrate species recorded were common to all forests and these shared species were generalist feeders. The effect of matrix type on diversity of epigaeic invertebrate communities was most notable for large forests (i.e. over 30 ha). Large fragments with core areas unaffected by edge-induced disturbance would support more forest dependent species than small fragments due to a lower susceptibility to invasions. Although not significantly nested, epigaeic invertebrate communities in both forest complexes tended toward nestedness. Isolation of forest fragments appeared to play a lesser role than patch area in determining the invertebrate community composition. Assuming that communities are extinction-dominated, community convergence in small fragments has probably already occurred, with invasions from the matrix confounding patterns of deterministic extinction of forest-dependent epigaeic invertebrate species. A habitat disturbance gradient was evident from the relatively undisturbed control (Karkloof quadrats) to the more disturbed pine dominated Gilboa quadrats, with intermediate disturbance values for the Balgowan (natural grassland matrix) fragments. The undisturbed Karkloof forest was characterised by a deep, abundant leaf litter layer, dense sub-canopy, and an abundance of seedlings indicating high rates of natural regeneration. Leaf litter depth was the most important variable in explaining the variation of epigaeic invertebrate species. A "shopping basket" of eight selected ecological indicators are mostly unrelated species. In general, this study supports the wealth of evidence advocating the use of epigaeic invertebrates, especially Carabid and Staphylinid beetles, as ecological indicators. In this study, eight species were identified by canonical correspondence analysis as ecological indicators that were sensitive to forest disturbance. This suite of species in the mistbelt mixed Podocarpus forests of the KwaZulu-Natal midlands will provide an accurate indication of forest condition in summer when abundance data is used. / Thesis (M.Sc.)-University of Natal, Pietermaritzburg, 2001.

Conservation of streams involves an understanding of their physical , chemical and biological entities. SASS5 is a biomonitoring method developed to monitor the habitat quality of a water body. It is based on differential scores attributed to various macroinvertebrate families with varying degrees of sensitivity to anthropogenic impact. This method , however, does not assess impacts on particular species. Odonata are good candidates for study at the species level as they are well researched and males are easily identified . As adults, they are known to be sensitive indicators of both riparian and river conditions. Yet Odonata cannot be an umbrella taxon for all other taxa . Therefore, the main aim of this study is to determine the complementarity of the two metrics (Odonata assemblages and SASS5), establishing whether Odonata assemblages offer additional information on, or insight into, riverine habitat quality as portrayed by SASS5. To accomplish this, certain objectives were addressed . 1) The variation of SASS5 scores and 2) Odonata assemblages between river systems, structural habitat types (open or closed canopies) and compositional habitat types (indigenous or alien vegetation). 3) Whether SASS5 scores vary to the same extent, and, 4) on the same spatial scale (river system and point localities) as Odonata abundance and species richness . The relationship between these two metrics was determined along three rivers in the Pietermaritzburg basin. Sampling units (SUs) with extremes in vegetation structure (sunlight and shaded SUs) and vegetation composition (alien or indigenous) were selected. Using this range of environmental conditions placed environmental extremes on the macroinvertebrate populations at point localities and having three different river systems added the dimension of variation over a broader scale, thus stretching the two metrics to investigate whether both responded similarly or in different ways. Results indicated that both metrics provide a similar portrait of overall river conditions. At the smaller spatial scale, the Odonata assemblage, unlike SASS, was highly sensitive to the riparian vegetation. Odonata species were less sensitive to vegetation composition but differentially sensitive to vegetation structure. However, landscape context is also important, with point localities being affected by the neighboring dominant habitat type. Larval Odonata alone did not provide this information. Overall, aquatic macroinvertebrates and adult Odonata provide a highly complementary pair of metrics that together provide large spatial scale (river system) and small spatial scale (point localities) information on the level of impact of stressors such as riparian invasive alien trees. / Thesis (M.Sc.)-University of KwaZulu-Natal, Pietermaritzburg, 2005.

Considered key landscape elements, ecotones play an important role in landscape ecology. In heavily fragmented, or heterogeneous landscapes, ecotones become a major, even dominant, feature. Yet, there are relatively few studies investigating communities of invertebrates associated with ecotones, especially across natural boundaries. Furthermore, most analyses of habitat loss do not consider the characteristics of the areas surrounding remaining habitat, the matrix. This thesis attempts to partially fill the gap. Afromontane forest-grassland ecotones are characteristically sharp (usually a few metres), are mainly fire-maintained, and have been in existence for, perhaps, millions of years. Therefore they provide a good opportunity to study ecotone, forest patch and grassland matrix characteristics together, and the associated species assemblages. In short, I investigated the diversity and distribution patterns of epigaeic invertebrates across ecotones between the natural and isolated patches of Afromontane forests and the surrounding natural grassland matrix, in the province of KwaZulu-Natal, South Africa. In particular, the following primary hypothesis was evaluated; do edge effects, in terms of elevated abundance and species richness, and in terms of abrupt, significant changes in environmental conditions, occur across near-natural ecotones. These edge effects quite often occur across anthropogenically-created habitat junctions, but it is not clear whether they do across natural ones. To test this hypothesis, the following secondary hypotheses were evaluated. Firstly, often a single invertebrate taxon is used for assessing changing landscape patterns. However, recent work has suggested that patterns and responses vary widely between taxa, and that management programmes based on the knowledge of a single taxon would not necessarily predict or safeguard that of others. Therefore, in chapter I, the hypothesis whether a single taxon could be used in biodiversity studies, or alternatively, whether it is better to select an array of taxa, was tested. Several invertebrate taxa were selected to investigate this. These included terrestrial amphipods, spiders, carabids, staphylinids and ants. Indeed, results showed that species diversities and assemblage-compositions of epigaeic spiders, carabids, staphylinids and ants were significantly different in different-sized Afromontane forest patches. Only carabids and staphylinids correlated positively with each other in terms of numbers of species. The other taxa showed only weak positive, or negative, correlations in their species richness. Results supported the multi-taxa approach in conservation studies, even among groups sharing a common habitat stratum. Secondly, organismal diversity usually increases at disturbed habitat edges. This phenomenon is commonly referred to as the biological edge effect. This pattern, however, is not universal and a number of authors have shown evidence contradicting this hypothesis. In chapter II amphipods, ground beetles and ants were collected to test the biological edge effect hypothesis. In addition, a number of abiotic factors were measured across these forest-grassland boundaries in an attempt to relate the biotic with the abiotic. Little evidence was found to support the classical edge-effect hypothesis (elevated species richness at the ecotone). In fact, carabid abundance and species richness was high in forests, decreasing gradually through the ecotone to a low in grasslands. In contrast, ant species richness increased significantly from a low in forests, increasing gradually through the ecotone, to a high in grasslands. Certain species did, however, show a significant increase in abundance at the ecotone, such as Talistroides africana, a terrestrial amphipod, and Tetramorium avium, a seed-predatory ant. Afromontane forest-grassland ecotones are natural and are not the result of anthropogenic clear-cut fragmentation. They also lacked any great changes in micro-environmental conditions. I hypothesise that edge effects are of less importance at more naturally maintained habitat boundaries even if these boundaries are sharp. Thirdly, climatic variation has a major impact on invertebrate communities. The Afromontane landscape experiences hot and wet summers, and cool and dry winters. I hypothesised that invertebrate distribution patterns across an ecotone change from one season to the next (chapter III). For example, it is expected that certain winter-active species might disperse from one location along the gradient to another, perhaps to escape predators, or find winter-refugia. This would, in turn, change the pattern of distribution of the selected taxa across these ecotones. Surprisingly, the general pattern of distribution across these ecotones changed little. However, there were significant differences between summer, spring, winter and autumn catch, and species identities changed from one season to the next. For example, carabid abundance and species richness was higher in the forest, compared to in the grassland, while ants were species richer in the grassland, compared to in the forest. These patterns were consistent from one season to the next. Again, as was found in chapter II, T. africana was significantly more abundant at the ecotone, compared to either forest or grassland interiors, in all seasons throughout the year. To summarise, amphipods favoured the ecotone environment, carabids the forests and ants the grasslands, throughout the year. Finally, scientists have recently become aware of the importance of the matrix surrounding habitat patches, in the survival and occurrence of organisms in the habitat patch. I tested whether the quality of the matrix, as a function of human disturbance, has an influence on invertebrate occurrence and distribution patterns across Afromontane forest-grassland boundaries. Redgrass (Themeda triandra Forssk.) dominated Afromontane grasslands are, and have been experiencing varying degrees of anthropogenic disturbance. Consequently, ecotones vary from being very abrupt (heavy disturbance in the matrix) to gradual (little disturbance in the matrix), although still sharp compared to most ecotones elsewhere. Level of grassland disturbance influenced amphipod, carabid and ant assemblage-structure across Afromontane ecotones (chapter IV). Results support the hypothesis that the dynamics of remnant areas are influenced by factors arising in the surrounding landscape. In particular, carabid assemblage-composition changed highly significantly from undisturbed to disturbed sites (this taxon was mainly captured from forests). Furthermore, only a single carabid individual was captured from 8 to 128 m into the grassland and only 14 T. africana individuals were captured from 2 m inside the forest to 128 m into the grassland of the most disturbed site. Clearly, matrix quality influenced not only the patterns of occurrence of organisms in the grassland matrix, but also at the ecotones and in the forest patches. In conclusion, it is imperative to investigate a number of taxonomic groups in conservation ecology to give more reliable results, and thus conservation recommendations. Of course, not all taxa can be considered, and the selection of appropriate taxa still poses a problem, but a set of taxa that are considerably different biologically is a good start. Native Afromontane forest-grassland mosaics are in urgent need of conservation, as much of this habitat is subject to heavy anthropogenic disturbance such as human settlement, forestry, cattle grazing, agriculture, and frequent, out of season, fires. Unfortunately only 2% of this biome is protected in South Africa. Furthermore, matrix quality is important because it determines the survival rate of propagules moving between remnant patches of habitat, and therefore the success rate of such movements. Private land-owners own most of this grassland area in the Afromontane region, but no guidelines are available to them on how to protect this habitat. Throughout this thesis I emphasised the importance of protecting both native forest and the surrounding native grassland. A first initiative is simply to protect a zone of grassland around the forest patches, both in terms of less frequent burning regimes, and less, or no cattle grazing here. By doing so, a rich grassland ant fauna will be conserved, the grassland matrix quality will improve, and a rich carabid fauna, favouring predominantly Afromontane forest remnants, will be conserved. / Thesis (Ph.D.)-University of Natal, Pietermaritzburg, 2000.

The savanna is an important biome, which is under threat from land transformation, and it is therefore a focus for conservation planning. Yet, the invertebrate fauna of this biome is poorly documented and hence there is a need to provide baseline data for this component of biodiversity. This project aimed to provide relevant information that can be used by conservation planners and ecologists, by recommending a sampling strategy for the collection of specific taxa for savanna invertebrate surveys. The effectiveness and efficiency of a sampling strategy using passive and active sampling methods was assessed to provide recommendations for a multi-taxa approach to sampling invertebrates in a savanna ecosystem. In the collection of data, volunteers assisted and they were evaluated in comparison with experienced researchers to assess the effectiveness, efficiency and benefits of using volunteers to carry out multi-taxa invertebrate surveys. In addition, cross-taxon congruency and congruency across taxonomic levels were assessed between nine invertebrate taxa, to select potential surrogates to reduce biodiversity survey costs for conservation planning. Fieldwork was carried out in the Mkhuze Game Reserve (27.67°S:32.27°E, 400km2 ), Phinda Private Game Reserve (27.78°S:32.35°E, 140km2 ) and False Bay Park (27.94°S:32.38°E, 25km2 ) in north-eastern Kwazulu-Natal, South Africa. Forty-three different sites were sampled between November 2002 and March 2005 (summer months). Twenty of these sites were re-sampled across years and in different months during the summer season, giving 77 sampling events. Fifty-four volunteers recruited by the Earthwatch Institute assisted in the collection of data. Lepidoptera, Hymenoptera (Apoidea), Diptera (Asilidae, Bombyliidae), Neuroptera, Odonata, Hemiptera (Cicadellidae), Coleoptera (Cetoniinae, Scarabaeinae), Orthoptera, Blattodea, Isoptera, Araneae (Araneidae, Thomisidae, Oxyopidae), Scorpionida, Myriapoda (Diplopoda, Chilopoda), Mollusca and Annelida were sampled using four active searching methods (transects, tree beating, leaf litter and sweep sampling) and two passive methods (pan traps and baited traps). In its entirety, this project sampled 50 558 individuals from 797 invertebrate species and an extensive database consisting of 33 257 records now exists. A standardised sampling protocol is described for the effective sampling of multiple invertebrate taxa in a savanna biome and recommendations are made for improving the efficacy and completeness of invertebrate surveys based on the application of species accumulation models. Restrictive active searching methods (quadrats) were found to be more effective for sampling epigaeic invertebrates and should be used in conjunction with leaf litter samples. Flying and plant-dwelling invertebrates should be sampled using a range of sampling methods which include baited, malaise and pan traps, active searching along transects and vacuum sampling. I suggest over 75% of the Lovel/, s.1. - MSc. Thesis i ii total estimated fauna to be a satisfactory and realistic level of inventory completeness for making valid comparisons between regions and across sites. Volunteers sampled lower rates of species accumulation, species richness and unique species when using timed, active search methods. Nevertheless, volunteers and researchers were shown to perform equally well when using un-timed, active searching methods. Previous experience or knowledge of scientific method was beneficial when researchers assessed the perceived usefulness of volunteers to researchers for carrying out fieldwork. The project experience raised the volunteers' environmental awareness, knowledge about biodiversity, invertebrates and conservation research, and enabled volunteers to participate in or design locally relevant conservation based projects on their return home. Cross-taxon congruencies were observed. However, relationships were weak and potential surrogates could not be selected. The use of higher taxonomic levels to represent species shows good potential as a surrogate but only in species-poor genera or families. The use of species density to determine congruency and select surrogates is likely to produce different results to those produced by community similarity. Furthermore, when selecting surrogates from congruency assessments an optimal p-value greater than 0.75 should be required. Below this value, the relationship is likely to be weak and if used as a surrogate misinterpretation may occur. / Thesis (M.Sc.)-University of KwaZulu-Natal, Pietermaritzburg, 2006.