Biodiversity and Conservation 12: 2279–2294, 2003.  2003 Kluwer Academic Publishers. Printed in the Netherlands.

Bird community composition in an actively managed savanna reserve, importance of vegetation structure and vegetation composition A.L. SKOWNO 1,2, * and W.J. BOND 1 1

Department of Botany, University of Cape Town, Private Bag, 7701 Rondebosch, South Africa; Current address: Kirstenbosch Research Centre, National Botanical Institute, Private Bag X7, 7735 Claremont, South Africa; * Author for correspondence (e-mail: skowno@ nbict.nbi.ac.za; fax: 127 -21 799 -6903) 2

Received 15 July 2002; accepted in revised form 9 December 2002

Key words: Bird assemblages, Multivariate analysis, Ordination, Savanna management, Secondary succession, Vegetation structure Abstract. The mosaic of trees, shrubs and open grassland in mesic African savannas is highly dynamic and strongly influenced by mammal herbivory and fire. We investigated the bird fauna in four different savanna habitats to help assess the impacts of vegetation change on this component of faunal diversity. Birds were censused, plant species were identified and vegetation structure was measured in four different vegetation types (Acacia nilotica woodland, Acacia nigrescens woodland, broadleaf thicket and open grassland) in the Hluhluwe-Umfolozi Park in northern KwaZulu Natal, South Africa. Multivariate ordination analyses were used to determine the relative importance of vegetation structure and floristic composition in defining bird assemblages. The bird communities of the grasslands, the acacia woodlands, and the broadleaf woodlands were clearly separated on the first axis of the detrended canonical correspondence analysis (DCCA). Canopy cover and foliage height diversity (FHD) were strongly correlated with the first axis of DCCA, possibly reflecting a secondary successional series from grassland to woodland, known as bush encroachment. Floristic composition (based on presence–absence data only) seemed to be less important for bird community composition than vegetation structure. The results indicate that changes in vegetation structure, caused by bush encroachment, could cause concomitant changes in bird community composition.

Introduction Many conservation areas are effectively habitat islands in a ‘sea’ of agricultural or urban areas in which natural disturbance regimes have been altered or limited. Managers of such areas often need to artificially maintain disturbance regimes in order to control ecosystem processes such as vegetation succession (Richards et al. 1999). Secondary succession, or woody plant encroachment, is a well documented phenomenon in savannas throughout the world (Smith and Goodman 1987; Archer et al. 1988, 1995). In mesic savannas such as those in the Hluhluwe-Umfolozi Park (HUP), KwaZulu Natal, the shift from open grassland to acacia savanna woodland to closed canopy broadleaf thickets can occur within 40 years (Watson and Macdonald 1983; Hoffman and O’Connor 1999; Skowno et al. 1999). Such rapid physiognomic changes are bound to cause significant changes in the associated

2280 faunal assemblages. However, there is limited data on the nature and magnitude of these habitat changes, which could be important for assessing effects on biodiversity of different management types. Before these changes can be investigated directly, some understanding of the importance of vegetation structure and composition in bird community structure needs to be gained. To investigate this we looked at bird assemblages in four savanna vegetation types in HUP, three of which (open grasslands, A. nilotica woodland and broadleaf thicket) form part of the vegetation succession series mentioned above. The fourth habitat, A. nigrescens woodland, is not typically part of the known successional series but covers a significant portion of HUP and could conceivably replace grasslands in the southern portion of the park. It is widely believed that the structure of the vegetation, its complexity and vertical arrangement are primary defining factors in bird communities (MacArthur and MacArthur 1961; Willson 1974). Yet, there are also studies showing that floristic composition plays an important role (Ralph 1985; Herremans 1993; Whelan 2000), especially when fruit-bearing tree species (Willson et al. 1994), or plants that provide distinct foraging environments for insectivorous bird species (Whelan 2000) are involved. Loiselle and Blake (1994) suggest that a range of successional stages is necessary to support the full complement of bird species that can occur in a given area. These successional stages may provide habitats of different structure and / or different plant species composition. Maintenance of a mosaic of vegetation types in a particular region or reserve may require substantial and active management (Loiselle and Blake 1994). In HUP the structure and composition of the vegetation is largely controlled by indigenous herbivores and disturbances such as fire, overlaid on local soils and climatic conditions (Whateley and Porter 1983). The park houses constantly fluctuating populations of all the indigenous large vertebrates, including large herbivores such as the white rhinoceros (Ceratotherium simium), elephant (Loxodonta africana), buffalo (Syncerus caffer), zebra (Equus burchelli) and wildebeest (Connochaetes taurinus). Balfour and Howsion (2002) suggest that herbivore densities are currently 10–30% higher than those indicated as typical for large African savanna mammals. In HUP, as in many other southern African reserves, burning is used to limit the invasion of woody plants into grasslands and savannas (see Skowno et al. (1999) for references). This partially controlled disturbance regime, which produces a spatial mosaic of patches of varying size and vegetation structure in HUP, is labour intensive and expensive to conduct. Consequently there remains the possibility that a more ‘hands off’ approach may be taken in the future, leaving the successional process to proceed. Decisions and actions on vegetation management could affect the various components of biodiversity by influencing habitat structure and composition. However, there is little data available on the biodiversity costs or benefits of alternative vegetation management options (i.e. species composition in different habitats). The aim of this study was to investigate the bird community composition in habitats with different vegetation structure and floristic composition in HUP, and in this way explore the potential effects of vegetation management. We used multivariate community analyses to investigate: (1) how the bird and plant assem-

2281 blages vary in terms of species composition within and between vegetation types, and (2) which components of vegetation structure and floristics relate best to the observed patterns of bird assemblages organisation.

Methods Site description The study was conducted in HUP, KwaZulu-Natal, South Africa (288009–288109 S, 328009–328109 E). The altitude of HUP varies from 90 m above sea level (a.s.l.) in the river valleys, to 580 m a.s.l. on the hilltops in the northeast and extreme south. HUP is generally hilly with some relatively flat areas on the flood plains of the major rivers. There is a strong rainfall gradient in HUP, ranging from 990 mm (mean annual rainfall) in the higher altitude areas in the northwest, to ,635 mm in the river valleys in the southwest (Park records 1932–1990). The wet season runs from October to March. Study design A total of 24 habitat patches, ranging in area from 5 to 538 ha (mean size 26.7 ha), were chosen from four different vegetation types in HUP. Six sites were selected in each of the following vegetation types: (1) broadleaf thicket, (2) Acacia nigrescens (Oliv.) woodland, (3) Acacia nilotica (L.) woodland, and (4) open grassland. We were unable to locate suitable habitat patches of the same size and were forced to select three large (.50 ha) and three small patches (,50 ha) of each habitat. As a consequence we included patch size as a covariate in all analyses. The vegetation types were defined by the dominant tree species, or lack thereof (Whateley and Porter 1983). The broadleaf thickets were dominated by Euclea divinorum (Hiern) and Euclea racemosa (Murray). E. divinorum commonly forms multi-stemmed shrubs up to 2 m in height, and can grow into multistemmed or single stemmed trees of up to 8 m in height (Pooley 1993). E. racemosa typically forms single stemmed trees up to 12 m in height and is common in the more mesic woodlands in HUP. The two acacia woodlands were dominated by Acacia nilotica and Acacia nigrescens, respectively. The grasslands in this study had no trees or only scattered trees (individuals separated by over 40 m). It should be noted that, although a large proportion of HUP is made up of the four vegetation types described above, there are areas where grassland, acacia and broadleaf woodland are found together. Suitable patches of vegetation were identified using aerial photographs, orthophotographs, and a digital vegetation map of the Park based on LANDSAT-TM images (Meyer 1999), viewed using the GIS Arcview (version 3.3, Environmental Systems Research Institute 1999, Redlands, California). Transects with a length of 400 m were randomly placed in each patch. In order to minimise edge effects, transects were placed more than 200 m from the edge of the patch (for sites B6 and

2282 N4 transects were placed 100 m from the edge of the patch due to small patch size). Sites were placed at least 1 km apart. Bird censuses and habitat measurements were completed between 11 December 1998 and 25 January 1999 (wet season). Bird census procedures Birds were censused using transects approximately 400 m long and 50 m wide (25 m on either side of the path chosen). Twenty minutes were spent walking each transect census. The slow walking speed and frequent pauses also allowed detection of inconspicuous woodland species by sighting and song. The length of the transect ensured useful censusing of the grassland patches. This method is a compromise between the long transects usually used in grassland bird censuses and the stationary point counts used in woodlands and forests (Bibby et al. 1992; Pomeroy 1992). The principal benefit of this technique is that it allows comparison between different habitats directly. However, the method does not completely remove the bias of underestimation of bird abundance and species richness in low visibility habitats (woodland) relative to open habitats (grassland). Because the study focussed on bird community structure and not diversity per se, it is unlikely that the analyses were significantly influenced by this underestimation. Each transect was censused four times, on rainless mornings, in the period of maximum activity and vocalisation between sunrise and two and a half hours after sunrise. On average four different transects were censused per day. The number of species and individuals encountered on each census was noted and the cumulative number of individuals for each species was used in analyses. Although 4 days is not sufficient to compile a comprehensive species list for a particular habitat patch, it allowed for the detection of a meaningful subset of species for community comparisons. Aerial feeding birds (swifts, swallows and martins) and raptors were excluded from the census because of the difficulties in assigning sightings to a particular patch. All bird censuses were conducted by A.L. Skowno. Vegetation measurement The structure of the vegetation was quantified using the method devised by MacArthur and MacArthur (1961), and modified by Ralph (1985). At eight points (every 50 m) along each transect we assessed the vegetation to the right and left of the transect line. At each point we estimated the distance at which a 30 3 30 cm board was 50% obscured by foliage at specific heights. These heights were 0, 0.5, 1, 1.5, 2, 3, 5, 8 and 12 m. The average distance for each height was calculated for each transect. This was translated to foliage density using Ralph’s (1985) formula: k50.69315 /D

(1) 2

3

where k is foliage density (m / m ), D is the distance to the imaginary board and 0.69315 is the natural logarithm of 2 (log e 2). A vertical vegetation profile was then constructed by plotting height above the ground (on the y axis) against the logarithm of vegetation density (on the x axis), and joining the points on the scatter.

2283 FHD was calculated using the Shannon–Wiener information theory formula (Equation 2): s

H952 S pi log 10 pi i51

(2)

where s is the number of categories and pi is the proportion of the observations in the ith category. In order to calculate FHD, the foliage profile is divided into three horizontal layers, 0–1, 1–2, and .2 m, and the proportions of the whole that each constitutes is the pi used in the Shannon–Wiener diversity formula. The resulting value is known as the FHD (MacArthur and MacArthur 1961). In addition to foliage density, canopy cover (CC), canopy height (Cn ht), grass height (Gr ht), grass cover (GC), grass density (Gr den), tree density (Tr den), shrub density (Sh den) and plant species composition were measured along each transect. Percentage CC was estimated by noting the presence or absence of CC every 10 m along the transect. For the purposes of this study woody plants over 3 m in height were considered to be trees. Plants between 2 and 3 m in height were classed as trees if they had one stem and as shrubs if they were multi-stemmed. Woody plants less than 2 m in height were classed as shrubs. Cn ht and Gr ht were estimated in a similar manner, using a 3 m measurement pole. Percentage GC was estimated in a 1 m 2 quadrat every 50 m. The distance to the nearest tree and shrub in each of the cardinal directions was also measured every 50 m. The reciprocal of the average distance to a tree was taken as a measure of density. The same was done for shrubs. The floristic survey, conducted in eight 50 m 2 quadrats along each transect, focused on grass, shrub and tree species, and excluded forbs (non-grass herbaceous plant species), which are a poorly described component of the flora, and in HUP contribute less than 1% to peak standing grass biomass (Zululand Grass Project, W.J. Bond, unpublished information). The environmental variables are presented in Appendix 1. Ordination Detrended correspondence analysis (DCA) was used to extract the dominant patterns of variation in community composition for the bird and plant species data. DCA is an indirect gradient analysis which ordinates only the species data and does not include environmental factors (ter Braak and Verdonschot 1995). Patch area was a potentially confounding variable and was therefore included as a covariate in all ordinations. In order to relate the bird community composition and vegetation structure directly, we used detrended canonical correspondence analysis (DCCA). DCCA is a direct gradient analysis technique in which a set of species data is related directly to a set of environmental variables. This method detects the patterns of variation in species data that can be explained by environmental variables. The environmental data used in the analysis of the bird communities were measures of vegetation structure. The data in all the analyses were not transformed and were detrended by segments. To assess the importance of each structural variable, an initial DCCA

2284 Table 1. The total, range and median number of bird species, and individuals (abundance) encountered in each of the four habitat types (n 5 6) over four consecutive mornings in mid-summer. Species

Broadleaf A. nigrescens A. nilotica Grassland

Abundance

Total

Range

Median

Total

Range

Median

45 46 59 24

13 13 12 8

23 23 20 11

624 453 564 385

30 48 33 89

104 84 95 65

including all the variables was run. The intraset correlations between the first two DCCA axes and the structural variables are effectively a measure of importance for each variable. Intraset correlations are the interset correlations divided by the species environment correlation of the axis and tend to be a more stable measure than interset correlations (ter Braak 1987). The computer programme CANOCO for Windows (version 4.0, Microcomputer Power, Ithaca, New York) was used for all ordinations. Distance analysis Another method of gradient analysis for community data is distance analysis. The objective of the analysis was to quantify the differences between both the bird and the plant assemblages in the four habitats. Unlike ordination, in which all sites are considered purely on their species compositions, the sites must be ‘grouped’ in distance analyses. The percentage dissimilarity between, and within the sites in each habitat type are the products of the analysis. The computer programme Releve´ Manager (ter Braak 1988, Agricultural Mathematics Group DLO, Wageningen, The Netherlands) was used for the distance analysis.

Results General A total of 92 bird species was encountered in the four vegetation types censused in summer. The most abundant species were the Rattling Cisticola (Cisticola chiniana) and Blackeyed Bulbul (Pycnonotus barbatus). The most abundant bird in broadleaf woodlands was the Greenbacked Bleating Warbler (Camaroptera brachyura), while the Croaking Cisticola (Cisticola natalensis) was the most abundant species in the grasslands. The Rattling Cisticola was the most abundant bird in the A. nigrescens woodland, and A. nilotica woodland (Appendix 2). The broadleaf thickets supported the highest number of individual birds, while the A. nilotica woodlands had the highest total number of species encountered (Table 1). The acacia woodlands have similar numbers of both grass and woody plant

2285 Table 2. The total number of grass species and woody plant species encountered in each vegetation type. Vegetation type (n 5 6)

Grass species

Woody species

Total plant species

Broadleaf woodland A. nigrescens woodland A. nilotica woodland Grassland

14 25 22 31

35 24 24 12

49 49 46 43

species. Not surprisingly, the grasslands have the highest number of grass species, while the broadleaf thickets have the highest number of woody species (Table 2). Ordination DCA – bird communities A gradient of increasing vegetation structural complexity is evident from right to left in the diagram. The grassland sites form a distinct cluster on the right-hand side of the diagram, and the broadleaf thickets form a cluster on the left-hand side (Figure 1). The two acacia woodlands occupy the ordination space between the open and the broadleaf clusters. The A. nilotica sites and A. nigrescens sites separate out on the second ordination axis, except for one A. nigrescens site (6g) which is placed amongst the A. nilotica sites. The relatively high eigenvalues of the axes (0.776 for axis 1 and 0.267 for axis 2) indicate that the observed patterns are well supported (Figure 1). DCA – plant communities based on presence–absence data Like the bird DCA, there are three main groupings in ordination space of the plant DCA (Figure 2). However, in this instance, on the first axis the grassland and A. nilotica sites formed a single cluster between the A. nigrescens sites on the left and the broadleaf sites on the right. There was little separation of sites along the second axis. Floristically the grasslands and A. nilotica woodland sites form one group,

Figure 1. DCA ordination of the first two axes of the bird census data. The sites are labelled as follows: 1b–6b, broadleaf thicket sites; 1g–6g, A. nigrescens woodland sites; 1n–6n, A. nilotica woodland sites; 1o–6o, grassland sites. Patch area was included as a covariate.

2286

Figure 2. DCA ordination of the first two axes of the plant presence–absence data from 24 sites sampled. 1b–6b, broadleaf thicket sites; 1g–6g, A. nigrescens woodland sites; 1n–6n, A. nilotica woodland sites; 1o–6o, grassland sites. Patch area was included as a covariate.

while the broadleaf and A. nigrescens woodlands are distinct. The first and second ordination axes (eigenvalues 0.513 and 0.24, respectively) explained 22% of the variation in the plant species data. DCCA – bird and vegetation structure The environmental variables, including nine structural variables and two plant species variables, explained over 60% of the variance in the bird data. Monte Carlo permutation tests of the F-ratios of the first axis eigenvalue (F 5 3.27) and trace statistic (F 5 1.84) (the sum of all eigenvalues) were both significant (P , 0.005), indicating that the complete set of environmental variables adequately explains the variation in the species data. The determinants of the DCCA axes are most likely those variables with the highest (intraset) correlations, namely CC and FHD (axis 1, R 5 20.947, R 5 20.907) and Gr ht (axis 2, R 5 0.811), see Table 3. The structural variables shrub height and Gr den, and the two plant species variables, derived from the plant DCA, showed weak intraset correlations (i.e. R , 0.6) with both the first and second ordination axes and were thus excluded from the ordination diagram (Table 3). Tr den and Cn ht were excluded due to collinearity with CC (R 5 0.91) and FHD (R 5 0.81), respectively. This left a set of five vegetation structural variables (Gr ht, CC, FHD, shrub density and ground cover), which were included in the species–vegetation structure biplot of the 24 sites (Figure 3). CC, FHD and shrub density (Figure 3) all reflect a successional series of increasing woody plant biomass, from the grassland sites on the right through the A. nilotica woodlands in the centre to the broadleaf thickets on the left. The A. nigrescens communities, clustered below the A. nilotica woodland sites, do not form part of this series, and occupy a position of intermediate CC and of low Gr ht. Dissimilarity analysis The dissimilarity analysis of the bird communities (Table 4) supports the clusters

2287 Table 3. Intraset correlations between the vegetation structural variables and the first two axes of the DCCA. All structural variables are included. Axis 1 Eigenvalue Vegetation structure variable FHD Gr den Gr ht GC Tr den Sh den CC Cn ht Sh ht Plant axis 1 Plant axis 2

Axis 2

0.763

0.260

20.8822 0.453 0.3721 0.4932 20.9078 20.7929 20.9518 20.6408 20.4437 0.5698 0.0793

20.324 0.6299 0.8512 a 0.642 a 0.1117 0.2277 a 0.0967 a 20.4898 0.3796 20.4657 0.1289

a

Patch area was included as a covariable. a Included in biplot of sites and structural variables (Figure 3).

Figure 3. DCCA ordination of the first two axes of the bird census data. Arrows indicate the direction and relative magnitude of selected vegetation structural gradients. The site vs. structural variable biplot shows avifaunal similarities among sites and their relationship to the selected structural variables. FHD – total foliage height diversity; Sh den – shrub density; CC – canopy cover; Gr ht – grass height; GC – ground cover. Patch area was included as a covariate.

formed in the DCA (Figure 1). The bird assemblages of the grasslands are different from all the woodlands. The broadleaf woodlands also have distinctive bird assemblages, which are over 70% dissimilar to all the other habitats. The percentage dissimilarity within the A. nilotica woodland sites is high, indicating that the bird assemblages within woodlands are more variable than those of the other habitats. The plant assemblages of the A. nilotica woodlands and grasslands are relatively similar, with 52% dissimilarity between the habitats (Table 5). As expected, the grassland and broadleaf thicket sites are most dissimilar in terms of plant community composition (81% dissimilarity). The level of dissimilarity within each of the habitats varies between 38% for A. nilotica woodlands and 53% for grasslands. An

2288 Table 4. Percentage dissimilarity of vegetation types based on bird species presence–absence data, and the % dissimilarity of the sites within each vegetation type (marked with *). Dissimilarity %

Broadleaf A. nigrescens A. nilotica Grassland

Broadleaf

A. nigrescens

A. nilotica

Grassland

41.1*

77.5 43.3*

71.9 60.6 62.1*

88.6 83.0 73.7 44.3*

Table 5. Percentage dissimilarity of vegetation types based on plant species presence–absence data, and the % dissimilarity of the sites within each vegetation type (marked with *). Dissimilarity %

Broadleaf A. nigrescens A. nilotica Grassland

Broadleaf

A. nigrescens

A. nilotica

Grassland

51.9*

78.2 46.7*

62.8 67.5 37.5*

80.9 80.0 52.2 53.1*

interesting difference between the plants and birds is that A. nilotica woodlands have relatively variable bird communities but relatively similar plant communities.

Discussion The principal findings of this study are that grasslands, both acacia woodlands and broadleaf woodlands, have markedly different bird assemblages. The magnitude of the differences, measured in terms of dissimilarity, indicates that vegetation changes associated with secondary succession could cause significant changes in bird assemblages in HUP. Bird community composition in the habitats sampled was associated with vegetation structure, and vegetation density and vertical complexity were the vegetation structure variables most strongly associated with the first axis of the DCCA. The variables CC and FHD are closely correlated with the ordination axis 1, along which the sites are linearly distributed. This arrangement of the sites agrees with the secondary successional sequence in HUP in which open grasslands are invaded by microphyllous shrub and trees species, which in turn provide shade and suitable habitat for broadleaf thicket species establishment (this successional sequence is limited or reversed by disturbance such as fire; see Skowno et al. (1999) for references). Along this vegetation series, there is rapid turnover in bird species associated with the occurrence of a tree layer (i.e. between the grassland avifauna and the A. nilotica woodland avifauna). Willson (1974), who first observed this phenomenon, noted that the turnover was driven by the addition of new feeding guilds of birds, rather than the expansion of guilds already present in open habitats. In this case the turnover in bird species, associated with the occurrence of a tree

2289 layer, is exaggerated because we selected sites consisting of relatively dense woodland. There are probably also open woodlands in which certain grassland and woodland bird species may coexist. Biological edge effects have been widely shown to influence bird diversity in small habitat patches (Ambuel and Temple 1983; McCoy and Mushinsky 1994; Schieck et al. 1995). In this case birds (usually generalist species) from surrounding habitats tend to utilise small patches of ‘inappropriate’ habitat. In particular small grassland patches tended to have higher numbers of woodland generalist birds such as Rattling Cisticolas (Cisticola chiniana) and Blackeyed Bulbuls (Pycnonotus barbatus) than the large grassland areas. Small A. nilotica patches were similarly occupied by more broadleaf thicket bird species than large A. nilotica patches. However, because there were relatively few generalist species and individuals occupying the smaller patches, these edge effects did not significantly affect the ordination analyses. Despite being structurally very different, the grassland and A. nilotica woodland sites are floristically similar. Because the plant community analysis was based on presence–absence data, obvious differences in species dominance between grasslands and woodlands were not taken into account. Many of the trees and shrubs characteristic of the A. nilotica woodlands in HUP are already present, as shrubs, in the grasslands but occur in low numbers. It is therefore unlikely that the changes in bird community composition between grassland and acacia woodland are driven by changes in plant species composition. Broadleaf thickets, on the other hand, are floristically distinct from the other vegetation types, although they are variable in their composition. Secondary succession from A. nilotica woodland to a broadleaf thicket would therefore involve a large turnover in plant species, which may contribute to the turnover in bird species. A. nigrescens woodlands, however, have a relatively distinct flora, which has some elements in common with grasslands and A. nilotica woodlands, but very little with the broadleaf thickets. The A. nigrescens woodlands, in general, occupy relatively dry low-altitude areas in the south of the Park, while broadleaf thickets are most common in higher altitude mesic areas, suggesting that the first axis of the plant DCA (Figure 2) may represent a moisture gradient, decreasing from left to right.

Conclusions The results of this study suggest that bird assemblages in HUP are determined, to a degree, by vegetation structure. Vegetation management in HUP could, therefore, have a direct effect on bird assemblages. If, for example, managers take a ‘hands off’ approach and allow bush encroachment to continue, areas of thicket and woodland would increase and open grassland areas would decrease. Bird species that require large open grassland areas would be lost and those requiring thicket and woodland would benefit. Although we did not investigate bird community composition in the full range of habitats in HUP, our results suggest that grassland bird

2290 communities would be particularly sensitive to changes in vegetation structure. More specifically, allowing shrubs which are encroaching in grassland areas to grow into trees (by excluding fire, for example) is likely to cause significant changes in bird community composition. The relative importance of structure and floristics in determining bird community composition are more difficult to distinguish in dense woodlands and thickets, because there are significant differences in vegetation structure, floristics, and bird community composition. From a management point of view, bush encroachment is a relatively easy process to monitor. Changes in vegetation structure are observable both on the ground and from aerial photographs and satellite images. With the development of sophisticated GIS analysis, spatially explicit data from these sources can be combined with specific data regarding faunal responses to vegetation structural change. This provides managers and scientists in savanna conservation areas with the tools to assess the potential impacts on biodiversity that their management strategies may have.

Acknowledgements We would like to thank Dave Balfour and his staff in the Hluhluwe-Umfolozi Park for expert field assistance, and KwaZulu-Natal Nature Conservation Services for providing unlimited access to HUP. The manuscript was improved by the comments of Dr C. Whelan and an anonymous reviewer. This work was funded by the National Research Foundation (NRF) of South Africa.

2291 Appendix 1. Vegetation structural variables measured at each site; full variable names are given below. Site

FHD1

Gr den

Gr ht (m)

GC (%)

Tr den

Sh den

CC (%)

Cn ht (m)

Sh ht (m)

1b 2b 3b 4b 5b 6b 1g 2g 3g 4g 5g 6g 1n 2n 3n 4n 5n 6n 1o 2o 3o 4o 5o 6o

1.40 1.31 1.28 1.38 1.34 1.33 1.23 1.23 1.20 1.22 1.24 1.08 0.84 1.26 0.47 0.97 0.75 0.45 0.24 0.14 0.08 0.43 0.18 0.20

0.31 0.99 0.64 0.54 0.72 0.56 0.28 0.64 0.25 0.37 0.30 0.97 1.07 0.80 2.09 0.96 1.19 2.57 2.22 1.19 1.27 0.57 1.13 2.56

0.39 0.39 0.25 0.80 0.46 0.31 0.12 0.33 0.05 0.14 0.08 0.48 0.52 0.54 0.59 0.97 0.70 1.16 0.86 0.11 0.45 0.42 0.79 0.12

70 78 51 80 82 59 47 84 61 69 50 81 73 70 94 83 88 91 96 92 84 82 87 99

0.11 0.16 0.08 0.17 0.14 0.14 0.06 0.09 0.05 0.06 0.05 0.08 0.11 0.10 0.08 0.09 0.07 0.06 0.01

0.27 0.28 0.28 0.27 0.29 0.30 0.08 0.15 0.07 0.07 0.10 0.10 0.16 0.31 0.11 0.14 0.18 0.14 0.17 0.07 0.03 0.07 0.01 0.01

69 71 71 75 81 79 38 57 33 59 27 47 47 50 57 47 60 47

5.9 5.4 8.3 4.2 6.0 5.7 6.8 6.7 7.2 6.8 6.6 5.6 4.2 4.4 4.3 5.2 4.4 4.9 3.8

1.3 1.4 1.2 1.4 1.2 1.2 1.2 1.5 1.2 1.0 1.0 1.2 1.4 1.4 1.4 1.3 1.1 1.3 1.1 1.1 0.6 1.0 1.3 1.4

FHD1 – total foliage height diversity; Gr den – grass density; Gr ht – grass height; GC – ground cover; Tr den – tree density; Sh den – shrub density; CC – canopy cover; Cn ht – canopy height; Sh ht – shrub height.

Appendix 2 Summary of bird species found in each habitat type, total number of individuals of each species (six sites per habitat type, four censuses per site); B 5 broadleaf thicket, G 5 Acacia nigrescens woodland, N 5 Acacia nilotica woodland, and O 5 open grassland. Bird species

Site B

Andropadus importunus Anthus cinnamomeus Apalis flavida Apalis thoracica Batis capensis Batis molitor Bostrychia hagedash Buphagus erythrorhynchus Camaroptera brachyura

G

80 13 3 2 3

21 11

104

N

O

5 2 3

2

45 2 2 10

2292 Appendix 2. (continued) Bird species

Site B

G

N

Campephaga flava Campethera abingoni Centropus burchellii Chrysococcyx caprius Chrysococcyx klaas Cinnyricinclus leucogaster Cisticola chiniana Cisticola juncidis Cisticola natalensis Colius striatus Coracias naevia Cossypha humeralis Cossypha natalensis Cuculus clamosus Dendropicos fuscescens Dicrurus adsimilis Dryoscopus cubla Emberiza flaviventris Eremomela icteropygialis Erythropygia leucophrys Esrilda astrild Euplectes ardens Euplectes axillaris Eupodotis melanogaster Francolinus natalensis Francolinus sephaena Halcyon chelicuti Hippolais icterina Lagonosticta rubricata Lamprotornis corruscus Lamprotornis nitens Laniarius ferrugineus Lanius collaris Lanius collurio Lybius torquatus Macronyx croceus Malaconotus blanchoti Melaenoris pallidus Melaenoris pammelaina Mirafra africana Mirafra rufocinnamomae Mirafra sabota Muscicapa caerulescens Myioparus plumbeus Nectarinia bifasciata Nectarinia senegalensis Nectarinia talatala Nectarinia veroxii

7 8

5 2

2 1 1 1 1 2 139

1 3 24

4 9 26 5 9 3

28 81

1 1 2 2 1 17 10 1

1 4

3 2

O

15 43 100

2

2 12 1 5 2 5 6

13 33 65 9

9 2 6 1

5 5 4

1

18 26 14 2 2

3 10 7

1 24 6 2 1 2

1 8 17 1 4 3

5 1 6 4 9 15

1

1

17

2 4 1

2293 Appendix 2. (continued) Bird species

Site B

Nicator gularis Nilaus afer Oriolus larvatus Parus niger Passer diffusus Petronia superciliaris Phyllastrephus terrestris Phylloscopus trochilus Ploceus ocularis Pogoniulus pusillus Prinia subflava Prionops plumatus Pycnonotus barbatus Quelea quelea Rhinopomastus cyanomelas Serinus mozambicus Streptopelia capicola Streptopelia semitorquata Streptopelia senegalensis Sylvietta rufescens Tauraco porphyreolophus Tchagra australis Tchagra senegala Telophorus quadricolor Telophorus sulfureopectus Terpsiphone viridis Trachyphonus vaillantii Tricholaema leucomelas Turtur chalcospilos Turtur tympanistria Upupa africana Uraeginthus angolensis Urocolius indicus Vidua macroura Zosterops pallidus Total species

14 1 3 4

G

N

10

7 2 7

9 12 23

5

O

3

10 6 2 4 21 97 3 3 1 7 2 24 2

7 1

4 4 25

27 8 93

4 23 30 2 1 12

9 29 8 2

3

2 2 3 7 7 1 3

1 20 10 25 1

11 1 4 1 1 7

1

4 6 6

38 22

21

11 33

14

References Archer S., Scifres C.J., Bassham C. and Maggio R. 1988. Autogenic succession in a subtropical savanna: rates, dynamics and processes in the conversion of grassland to thorn woodland. Ecological Monographs 58: 111–127. Archer S., Coughenour M., Dall’Aglio C., Fernandez G.W., Hay J., Hoffman W. et al. 1995. Savanna biodiversity and ecosystem properties. In: Solbrig O.T., Medina E. and Silva J.F. (eds), Biodiversity and Savanna Ecosystem Processes: A Global Perspective. Ecological Studies, Vol. 121. SpringerVerlag, New York, pp. 207–215.

2294 Ambuel B. and Temple S.A. 1983. Area dependent changes in the bird communities and vegetation of southern Wisconsin forests. Ecology 64: 1057–1068. Balfour D.A. and Howsion O.E. 2002. Spatial and temporal variation in a mesic savanna fire regime: responses to variation in annual rainfall. African Journal of Range and Forage Science 19: 45–53. Bibby C.J., Burgess N. and Hill D.A. 1992. Bird Census Techniques. Academic Press, London. Herremans M. 1993. Effects of woodland modification by African elephant Loxodonta africana on bird diversity in northern Botswana. Ecography 18: 440–454. Hoffman T.M. and O’Connor T.G. 1999. Vegetation change over 40 years in the Weenen / Muden area, KwaZulu-Natal: evidence from photo-panoramas. African Journal of Range and Forage Science 16: 71–88. Loiselle B.A. and Blake G. 1994. Annual variation in birds and plants of a tropical second-growth woodland. The Condor 96: 368–380. MacArthur R.H. and MacArthur J.W. 1961. On bird species diversity. Ecology 42: 594–598. McCoy E.D. and Mushinsky H.R. 1994. Effects of fragmentation on the richness of vertebrates in the Florida scrub habitat. Ecology 75: 446–457. Meyer B. 1999. Supervised LANDSAT-TM mapping of the spatially heterogeneous savanna of Hluhluwe-Umfolozi Park, KwaZulu-Natal (unpublished). Pomeroy D. 1992. Counting Birds. African Wildlife Foundation, Nairobi, Kenya. Pooley E. 1993. The Complete Guide to Trees of Natal, Zululand and Transkei. Natal Flora Publications Trust, Durban, South Africa. Ralph C.J. 1985. Habitat association patterns of forest and steppe birds of northern Patagonia, Argentina. The Condor 87: 471–483. Richards S.A., Possingham H.P. and Tizard J. 1999. Optimal fire management for maintaining community diversity. Ecological Applications 9: 880–892. Schieck J., Lertzman K., Nyberg B. and Page R. 1995. Effects of patch size on birds in old growth montane forests. Conservation Biology 9: 1072–1084. Skowno A.L., Midgley J.J., Bond W.J. and Balfour D.A. 1999. Secondary succession in Acacia nilotica (L.) savanna in the Hluhluwe Game Reserve, South Africa. Plant Ecology 145: 1–9. Smith T.M. and Goodman P.S. 1987. Successional dynamics in an Acacia nilotica–Euclea divinorum savannah in southern Africa. Journal of Ecology 75: 603–610. ter Braak C.J.F. 1987. The analysis of vegetation–environment relationships by canonical correspondence analysis. Vegetatio 69: 69–77. ter Braak C.J.F. and Verdonschot P.F.M. 1995. Canonical correspondence analysis and related multivariate methods in aquatic ecology. Aquatic Sciences 57: 255–289. Watson H.K. and Macdonald I.A.W. 1983. Vegetation changes in the Hluhluwe-Umfolozi Game Reserve Complex from 1937 to 1975. Bothalia 14: 265–269. Whateley A. and Porter R.N. 1983. The woody vegetation communities of the Hluhluwe-CorridorUmfolozi Game Reserve Complex. Bothalia 14: 745–758. Whelan C. 2000. Foliage structure influences foraging of insectivorous forest birds: an experimental study. Ecology 82: 219–231. Willson M.F. 1974. Avian community organization and habitat structure. Ecology 55: 1017–1029. Willson M.F., De Santo T.L., Sabag C. and Armesto J.J. 1994. Avian communities of fragmented south-temperate rainforests in Chile. Conservation Biology 8: 508–520.