Dynamics of the forest vegetation of the Umtiza Nature Reserve, East London

The forest community at the Umtiza Nature Reserve near East London was surveyed using 24 plots (0.04 ha) in which all woody stems >0 .5 m tall were enumerated. Based on a classification using numbers of stems of canopy species, it was assumed that basically only one forest community was sampled. Further multivariate analyses suggest that this forest is fine-grained. Sample plots were similarly placed in ordination space irrespective of whether woody species occurrence was used as importance value or if species occurrence per size class was used separately [seedlings (0.5-1.0 m), saplings (1-5 m) or canopy individuals ( > 5 m)). An analysis of size-class distributions of the most common canopy species indicated that the majority of species exhibited inverse J-shaped size-class distributions. This is the expected pattern for a fine-grained forest. In these measures of dynamics, this forest is not fundamentally different to the more temperate Afromontane forests.


INTRODUCTION
In brief, forest dynamics is the complex product of inter actions between disturbance regime (e.g. type o f disturb ance. turnover rate), life histories of constituent species along a shade-tolerant to shade-intolerant continuum and particulars of the regeneration arena (e.g. regeneration bot tlenecks due to biotic or abiotic events). As such the study of dynamics needs several inputs. In this preliminary study we concentrated on aspects of the grain and life history components of dynamics. By life histories we mean whether a species is relatively shade-tolerant 'clim ax' or shade-intolerant 'pioneer' and this we have inferred from size-class distributions.

By grain we mean 'the mosaic of structural phases' (Whitmore 1989); or the spatial patterns o f seedlings/ saplings and canopy individuals o f the different species.
Coarse-grained forests tend to have at least some (recently disturbed) areas which are dominated by shade-intolerant species. When plotted in ordination space, coarse-grained forests should show clear separation o f plots according to whether pioneers or climax species are present, especially when their sizes are considered. In fine-grained forests most species can regenerate close to adults and therefore there are no great differences in species composition or size-class distributions amongst plots. In other words the successional process occurs on a small spatial scale in fme-grained forests.
Recent work has shown that in many different forests at any one time gaps occupy about 1-2% of the area (Barden 1989;Connell 1989). In forests not subject to catastrophes these values suggest that stems in the small size classes of shade-intolerant species should be rare in comparison to the large bank of advance regeneration of shade-tolerant species. Thus the size-class distributions of shade-intolerant species should be flatter than those of shade-tolerant species. A forest comprised of many species with relatively flat size-class distributions should thus be a relatively coarse-grained forest.
Very little has been published on the dynamics of South African forests. At the moment there is therefore no model to predict the grain of a South African forest either for a given environment (e.g. climate and soils) or for forest structure (e.g. whether there is a preponderance of large or small-sized individuals). Midgley et al. (1990) indicated that the southern Cape forests [part of White's (1983) Afromontane Forest type] were extremely fine-grained. At the scale of 0.04 ha plots, most species were able to recruit continuously and most canopy species were shade-tolerant. This was interpreted as being due to the unproductive environment (cool climate, poor soils) and conservative disturbance regime (few large gaps) which has favoured shade-tolerant species and restrained both woody and herbaceous shade-intolerant species.
In contrast to the southern Cape forests, our working hypothesis for tropical and subtropical gap-phase forests, such as are found at our study site, is that they can be expected to have a greater component of shade-intolerant species, of both herbaceous and woody types. This is due to their situation in a more productive environment (higher temperatures, summer rainfall) which increases the opportunities for regeneration of fast-growing pioneer species. Such forests should therefore be more coarse grained due to intermittent recruitment of shade-intolerant canopy dominants leading to spatially segregated size classes. Thus an ordination or classification of sample plots using the occurrence of a species in the seedling, sapling or canopy size classes, should reveal spatially distinct groups in a coarse-grained forest. In a fine-grained forest, where species are able to regenerate close to their own adults, data of species occurrence b; size class should have little influence on classifications or ordinations.

Our survey was done in a coastal forest of the eastern Cape. According to Acock's (1988) classification these forests appear to fall between the Coastal Tropical Forest Type (Veld Type No. 1) and Valley Bushveld (Veld Type No. 23). White (1983) mapped these forests as the Tongaland-Pondoland Forest type and most recently they have been mapped as Dune Thicket (Lubke, Tinley & Cowling 1988). Information on forest dynamics in the eastern Cape is conspicuous by its absence in the over view of vegetation in the eastern Cape by Lubke, Tinley & Cowling (1988) and in Everard (1987).
Our objectives were to use information from plots to: 1, briefly describe the vegetation; 2, compare ordinations of plots using information on presence/absence of potential canopy species in the seedling, sapling or canopy layers, to infer grain; and 3, analyse the size-class distributions of the important species to infer their life histories.

STUDY AREA
The study area is situated in the Umtiza Nature Reserve (33°02'S; Z7°47'E) which is located on a northeast-facing slope of the Buffalo River Valley, about 10 kilometres from East London. Thicket/forest vegetation covers about 550 ha of the reserve. One of the aims of this reserve is to afford high conservation status to two forest tree species: Umtiza listeriana Sim, belonging to a monotypic genus of the Fabaceae, and Buxus macowanii Oliv., the box-wood much exploited in the eastern Cape in the past.
Maximum altitude in the reserve is 180 m and many streams dissect the area giving it a variety of aspects. The geology of the area has been mapped as Beaufort Group (Rust 1988) and the soils are mapped as weakly developed soils on rock with black to brown clays and clay loams (Hartmann 1988).

Sampling
The broad vegetation communities of the Umtiza Nature Reserve were mapped from aerial photographs and essen-  tially two types were found to occur; an Acacia kar roo/grassland and a thicket/forest type. A grid was placed over a 1:10 000 map of the reserve and 24 sites were select ed at random in the forest communities (Figure 1). At each site a square plot of 20 X 20 m was laid out and all woo dy plants > 0 .5 m tall were enumerated. These plots were demarcated permanently for further studies on distur bance, mortality and growth. At each sample plot each stem was allocated to one of three size classes: seedlings (0.

Analysis
Initially we classified the forest vegetation using the multivariate package TWINSPAN (Hill 1979a), with numbers of stems as the importance values and the default settings o f the package. In total about 60 species occurred in the canopy ( > 5 m in height; Table 1) in our 24 plots. By using species occurrence in seedling, sapling or canopy levels as separate species, we could increase the apparent number of 'species' to about 150 and the apparent number of 'plots' to 72.

In the second part o f this study we used multivariate techniques to represent grain or succession in space.
Multivariate techniques have long been used for discerning successional trends, e.g. Fox (1990)  We also calculated the position of centroids for ordi nations of plots using presence/absence of canopy species in seedlings, saplings and canopy size classes separately. A meaningful successional trajectory would be indicated by significant directional trend in ordination space as defined by information from each size class.
Our expectations were that in a coarse-grained forest, significantly different eigenvalues or breadth o f axes would result when data from each size class were used separately, rather than when information about the canopy layer only, was used. We expected centroids to be widely separated in ordination space when species by size-class informa tion was incorporated.
For the second part o f the study, the DECORANA (DCA-option)/TWINSPAN packages (Hill 1979a, b) were again used. Presence/absence (due to the great discrepan cies in numbers of stems amongst size classes) was used as the importance value, and again we used the default settings of the package.

Environmental data
Soil colours recorded included very dark brown, dark red-brown, dark brown and brown-black, and soil texture was sandy clay loam. The soils also had a relatively high pH and calcium content (Table 2). Slopes were mostly gen tle and aspect was predominantly between 90° and 270°.

Vegetation description
Very little phytosociological data have been published on the eastern Cape forests and thickets. Furthermore, vegetation patterns are complex (see Acocks 1988). For this reason we have appended an annotated species list ( Table 1). The maximum number of stems observed per 0.04 ha sample for the canopy size class was 74, for seedlings it was 454 and for saplings 25. The maximum number of seedlings per sample plot was high for the following species: Ptaeroxylon obliquum (421), Buxus macommii (189) and Ilex mitis (256). Species richness per 0.04 ha of woody plants ranged from 13-36.

Vegetation dynamics
The groupings of samples produced at the first division using 60 species (i.e. only canopy species) versus 150 species (i.e. canopy species per size class separately), whilst having different orderings, were identical (Table 3) and both had similar and small eigenvalues (0.26 compared to 0.29). This indicates that species by size-class data do not produce a different classification or a different sepa ration of samples.
The ordination using 72 'plots' and the 60 canopy species shows considerable overlap of the three size-class groups (Figure 2A, B, C). This is clearly depicted by the similarity of position of each centroid and their relatively close proximity to the origin. There is no suggestion of orderly directional trend o f movement of centroids. This could indicate that there is no consistent difference in succes sional status amongst plots. The spread along the first ordination axis is narrow (about 3 s.d.) and the eigen vector was also small (0.279).

Size-class distribution
Most of the common canopy species have typical inverse J-shaped size-class distributions (Table 4). To this group could be added many of the less common species we encountered but which had similar size-class distributions. This includes species such as Maytenus heterophylla, M. peduncularis, Cussonia spicata, Euclea natalensis, Brachylaena elliptica and Olea capensis subsp. capensis. DISCUSSION We interpret the multivariate analysis to indicate that this subtropical forest is fine-grained. This forest is there- Height (m) dbh (cm) Species 0.6 -0 .9 9 1 -5 <10 <15 < 2 0 < 2 5 < 3 0 < 3 5 > 3 5