Effects of human disturbance on composition and structure of Brachystegia woodland in Arabuko-Sokoke Forest, Kenya

Authors

  • Joseph O. Oyugi,

    Corresponding author
    1. Department of Biological Sciences, University of Illinois at Chicago, 845 West Taylor Street, Chicago, IL 60607, U.S.A.
    2. National Museums of Kenya, Department of Ornithology, P. O. Box 40658 Nairobi, Kenya
    3. Department of Biology, Wright College, 4300N. Narragansett, Chicago, IL 60634, U.S.A.
    Search for more papers by this author
  • Joel S. Brown,

    1. Department of Biological Sciences, University of Illinois at Chicago, 845 West Taylor Street, Chicago, IL 60607, U.S.A.
    Search for more papers by this author
  • Christopher J. Whelan

    1. Illinois Natural History Survey, Midewin National Tallgrass Prairie, 30239 South State Route 53, Wilmington, IL 60481, U.S.A.
    Search for more papers by this author

*E-mail: joyugi@ccc.edu

Abstract

We examined tree species diversity, density, dispersion patterns and size class distributions in Brachystegia woodland of Arabuko-Sokoke Forest, Kenya. The metrics varied with human disturbance (disturbed versus relatively undisturbed areas). Julbernardia magnistipulata Harms occurred only in the undisturbed site. Brachystegia spiciformis Benth. had the highest importance value (IV) at both study sites, whereas the IVs for the other tree species were greater within the undisturbed than the disturbed sites. Simpson’s Diversity Index and tree densities were greater in the undisturbed site than in the disturbed site. All seven tree species exhibited random dispersions in the disturbed site, but three species were clumped in the undisturbed areas. Smaller individuals of B. spiciformis were over-represented in the disturbed habitat relative to the undisturbed habitat. In contrast, J. magnistipulata, Lannea schweinfurthii Engl. and Mimusops obtusifolia Wall. exhibited a greater proportion of smaller trees in the undisturbed site. The tree size class distributions may provide an index of regeneration for these four tree species in the disturbed and undisturbed sites respectively. Active management and restoration may be desirable for J. magnistipulata but appears unnecessary for the remaining common tree species. Illegal logging has persisted in the forest despite its conservation status over many decades. If logging activities cease, it would be instructive to document changes in vegetation composition and structure over time.

Résumé

Nous avons examiné la diversité, la densité, les schémas de dispersion et la distribution des classes d’âge de différentes espèces d’arbres dans la forêt àBratystegia d’Arabuko-Sokoke, au Kenya. Les données mesurées varient avec les perturbations humaines (zone perturbée vs. relativement non perturbée). Julbernardia magnistipulata Harms ne se trouvait que dans la zone non perturbée. Brachystegia spiciformis Benth. avait l’Importance value (IV) la plus grande dans les deux sites d’étude tandis que l’IV des autres espèces d’arbres était plus grande dans le site non perturbé que dans le site perturbé. L’Indice de diversité de Simpson et la densité des arbres étaient plus grands dans le site non perturbé que dans le site perturbé. Les sept espèces d’arbres présentaient des dispersions aléatoires dans le site perturbé, mais trois espèces poussaient en groupes dans le site non perturbé. Les individus plus petits de B. spiciformisétaient sur-représentés dans l’habitat perturbé par rapport à l’habitat non perturbé. Par contre, J. magnistipulata, Lannea schweinfurthii Engl. et Mimusops obtusifolia Wall. présentaient une plus grande proportion de petits arbres dans le site non perturbé. La distribution des classes de taille d’arbres peut donner un indice de la régénération de ces quatre espèces dans les deux sites respectivement. La gestion et la restauration actives peuvent être souhaitables pour J. magnistipulata mais ne semblent pas nécessaires pour les autres espèces communes d’arbres. La coupe d’arbres illégale se poursuit dans la forêt malgré le statut de conservation dont elle jouit depuis des décennies. Si ces activités cessaient, il serait instructif de documenter les changements de la composition et de la structure de la végétation avec le temps.

Introduction

An extensive, though narrow, strip of coastal forest once extended along the east coast of Africa from southern Somalia in the north to the Eastern Cape in the south (Moll & White, 1978). The composition of this coastal forest changed continuously from north to south, and it possessed a high faunistic and floristic endemism (Moll & White, 1978). Although most of this forest has been converted to grassland, wooded grassland, or agriculture, a number of significant forest remnants remain. The largest of the remnants north of Mozambique is the Arabuko-Sokoke Forest in Kenya (Burgess, Fitzgibbon & Clarke, 1998). Arabuko-Sokoke Forest is an internationally recognized repository of biodiversity. It houses 50 species of globally or nationally rare plant species, at least four endemic butterfly species, three globally threatened mammal species and six globally threatened bird species, as well as many reptile and invertebrate species (Collar & Stuart, 1998; Lange, 2003). Arabuko-Sokoke Forest ranks second in importance for forest conservation of rare and threatened bird species in Africa (Collar & Stuart, 1998; Bennun & Njoroge, 1999).

Despite its status and recognition as a biodiversity hotspot (Collar & Stuart, 1998), little quantitative information exists on the vegetation structure and composition of the Arabuko-Sokoke Forest. Broad vegetations surveys and inventories of the forest have been conducted (Moomaw, 1960; Kelsey & Langton, 1983; Awimbo & Wairungu, 1990; Mutangah & Mwaura, 1992; Davies, 1993; Hawthorne, 1993; Robertson & Luke, 1993; Wairungu, Awimbo & Kigomo, 1993; Blackett, 1994; Muchiri, Kiriinya & Mbithi, 2001), and while informative, they were largely qualitative and descriptive. Some studies have assessed endemism of fauna and flora of East African coastal forests, including Arabuko-Sokoke Forest (Dale, 1939; Britton & Zimmermann, 1979; Burgess et al., 1998, 2003).

There have been detailed investigations of Miombo woodlands (Brachystegia-dominated savanna) elsewhere (Frost, 1996), yet Miombo woodlands appear only superficially similar to the forest types at Arabuko-Sokoke Forest (Keay, 1959 in Frost, 1996). As part of a larger study focusing on the foraging ecology and habitat relationships of sunbirds within the Arabuko-Sokoke Forest (Oyugi et al., unpublished data), we conducted a quantitative investigation of the composition and structure of the disturbed and relatively undisturbed patches of Brachystegia woodland within the Arabuko-Sokoke Forest. The differences in disturbance resulted from past legal and current illegal logging activities. We counted cut stems, foot paths and old logging tracks to identify disturbance levels (disturbed versus relatively undisturbed areas).

We address the following issues and hypotheses:

  • 1) If logging primarily removed rarer tree species then the disturbed forest should be even more heavily dominated by B. spiciformis and bear less resemblance to the intact Miombo woodlands of Zambia and Zimbabwe. The reverse would be expected if logging primarily focused on B. spiciformis. We can also see whether the Brachystegia woodlands of Arabuko-Sokoke can be categorized as Miombo woodland.
  • 2) If the disturbance has been sufficiently large, we may expect to see succession where a distinctly different tree composition may be emerging in the disturbed forest.
  • 3) Armesto, Mitchell & Villagran (1986) hypothesized that frequent, large-scale disturbances should create random spatial patterns, whereas individuals of a tree species should appear clumped in forests where canopy gaps are the major cause of structural change. Hence, the disturbed forest should exhibit more random dispersions among individuals of a species, and the undisturbed should exhibit more clumped dispersions.
  • 4) Logging in the disturbed forest should result in fewer large trees and so we expect mean diameter at breast height to be higher in the undisturbed forest. Furthermore, species experiencing successful regeneration in the disturbed forest should have size distributions more skewed towards saplings and smaller trees than the same species within the undisturbed forest. Unsuccessfully regenerating species should either be unusually rare or lack these smaller size classes of trees.

Methods

Study area

The Arabuko-Sokoke Forest lies between 39°50′E and, 39°40′E longitude and 3°10′S and 3°30′S latitude, along the eastern Kenyan coast (110 km north of Mombasa and 18 km south of Malindi). The topography rises 60–135 m above sea level, and the mean annual rainfall ranges from 600 mm in the northwest part to 1100 mm at the Gedi station in the northeast, with the short rains falling from November to December (Muchiri et al., 2001).

The forest has been described in detail by Moomaw, 1960; Kelsey & Langton, 1983; Thomas, 1988; Awimbo & Wairungu, 1990; Mutangah & Mwaura, 1992; and Robertson & Luke, 1993;. Arabuko-Sokoke Forest is one of the few remaining indigenous forests in Kenya, and one of the largest extant fragments of a coastal forest that once covered much of the East African coast (Burgess et al., 2003). European timber merchants harvested trees from the forest in the early 1900s and removed most of the commercially valuable timber. The most harvested tree species included Sterculia appendiculata K.Schum. ex Engl., Manilkara sansibarensis Engl., Afzelia quanzensis Welw., Brachylaena huillensis O.Hoffm. and B. spiciformis (Robertson & Luke, 1993; Wright, 1999). Robertson & Luke (1993) note how literature on Arabuko-Sokoke Forest frequently lists S. appendiculata as logged for timber even though this species is absent within the forest boundaries. Robertson & Luke (1993) suggest that much written about the Arabuko-Sokoke Forest in fact refers to a much wider area, now under permanent cultivation. These former forests include areas closer to the shore towards Kilifi (now on the east side of the main road that separates the protected forest area from areas of cultivation), and towards the west into the Mangea hills.

Arabuko-Sokoke Forest was proclaimed a crown forest in 1932 and gazetted as a forest reserve in 1943. Under the Forestry Department management, commercial exploitation was supposed to stop. However, five sawmills legally operated in Arabuko-Sokoke Forest until the 1950s, and one of these sawmills operated in the Brachystegia habitat and retained a license for the removal of B. spiciformis until the 1990s (Davies, 1993; Robertson & Luke, 1993; Blackett, 1994). In 1977, a special nature reserve was created within the forest where all extractive activities were declared illegal. The area was to be monitored through survey and permanent sample plots. Sporadic vegetation monitoring has taken place with surveys by the Kenya Forestry Research Institute (Wairungu et al., 1993; Muchiri et al., 2001), Kenya Indigenous Forest Conservation Program (Davies, 1993; Blackett, 1994) and International Council for Bird Preservation (Kelsey and Langton, 1983). In 1991, a memorandum of understanding was signed between the Kenya Forestry Department and Kenya Wildlife Service that gave both parties a management role. Each service ostensibly manages about half of the remaining Brachystegia woodland. Since 1991, illegal-logging activities have subsided but not totally ceased in the forest (M. Mwavita personal communication).

This study took place within the 7636 ha Brachystegia woodland which runs in a central strip through the Arabuko-Sokoke Forest. This is relatively open habitat dominated by B. spiciformis growing on soil of white sands (Kelsey & Langton, 1983). The decades of logging activities and current human activities have created matrices of relatively more and less disturbed patches. Hereafter, we refer to these differences as disturbed and undisturbed. Disturbed areas have conspicuously more foot paths, cut stems and old logging tracks than the relatively undisturbed ones. We counted nineteen versus eight foot paths; five versus two cut stems and two versus zero old logging tracks in the disturbed and undisturbed patches respectively.

In Arabuko-Sokoke, B. spiciformis (Leguminosae) grows up to 25 m (Chudnoff, 1984). Little is known about its regeneration requirements, annual seed production, or seed consumers within the Arabuko-Sokoke Forest. However, its phenology, flowering and fruiting and dispersal, seed germination and regeneration, growth and mortality are known for the Miombo Woodland of Zambia and Zimbambwe (Chidumayo & Frost, 1996).

The Arabuko-Sokoke Forest has two other distinct forest types (Arabuko-Sokoke Strategic Forest Management Plan 2002–2027, 2002); (1) mixed forest with a diversity of relatively dense, tall and undifferentiated trees covering an area of about 7000 ha; (2) Cynomnenta forest and thicket, which covers about 23,500 ha, occurs to the west on red Magarini sands (Kelsey & Langton, 1983), and is dominated by Cynometra webberi Baker f., Manilkara sulcata Dubard, Oldfieldia somalensis (Chiov.) Milne-Redh. and formerly B. huillensis. This latter species, much in demand for the wood-carving trade, has been logged or poached from most of the forest.

Tree survey

We established two 1400 × 800 m study plots, in disturbed (3°19′S and 39°55′E) and relatively undisturbed (3°25′S and 39°52′E) areas within the Brachystegia forest. We surveyed the plots during the dry season; January–May 2002.

Within each large plot, we established five 400 m long transects each 250 m apart. Along each transect, we established five 30 × 30 m subplots at 50 m intervals (Pulido & Diaz, 1997). We identified, counted, and measured diameter at breast height (dbh) of all trees within each subplot. For trees with multiple stems, we summed total basal area and assigned that to the individual.

Data analysis

Tree species abundance and diversity.  We encountered eleven canopy tree species. Because of the rarity of four of these species [Strychnos madagascariensis Poir., Memecylon fragrans A.Fern. & R.Fern., Newtonia paucijuga (Harms) Brenan, Rytigynia spp], we restrict formal statistical analyses to the seven most common tree species, except when calculating species diversity indices where all species were included. We set the density threshold for common tree species at ≥1 individual ha−1. Density estimates came from averaging the density of a species across the subplots of a given habitat. A Mann–Whitney U-test with subplots as replicates tested for differences in tree numbers between the disturbed and undisturbed habitats, except J. magnistipulata which occurred only in the undisturbed habitat. The importance value (IV) for each tree species within the Brachystegia woodland was calculated as (Curtis, 1959): IV = relative density + relative frequency + relative dominance (dbh).

We calculated the diversity of tree species for each subplot using the reciprocal of Simpson’s Diversity Index (Magurran, 2004):

image

where ni = the number of individuals in the ith species and N = the total number of individuals of all species. This measure of diversity increases with the numbers of species and the evenness with which individual trees are spread among the different species. We used a t-test to see whether tree species diversity differed between the two habitat types.

Tree species dispersion patterns.  To characterize each tree species dispersion pattern quantitatively, we calculated Lloyd’s index (Vandermeer, 1981):

image

where s2 = variance in number of individuals among subplots, λ = mean number of trees per subplot. If a tree species is randomly distributed, the variance of the distribution should equal the mean; if trees are evenly distributed, the variance is significantly less than the mean. If tree distribution is clumped, the variance is significantly greater than the mean. To test for nonrandom dispersions (clumped or even), we used a chi-square test to compare observed counts of individual trees per subplot with counts generated using a Poisson distribution (Vandermeer, 1981).

Tree size distributions.  We used MANOVA to test for effect of habitat type (disturbed and undisturbed habitats) on diameter at breast height using the seven trees species as the dependent variables. We then used a coefficient of skewness (g1) to summarize the asymmetry of tree species size distributions with respect to habitat (Bendel et al., 1989; Wright et al., 2003). The coefficient of skewness was calculated using the natural logarithm of dbh (Bendel et al., 1989) for each individual of each tree species. The assumption is that the regular interval along the scale of the natural logarithm of the stems more closely approximate the stems’ ages. Coefficient of skewness

image

where n represents the number of individual trees, xi represents the natural logarithm of dbh for individual tree i, inline image represents the mean of xi, and s represents the standard deviation of the xi. Skewness is significant if skewness/standard error of skewness is greater than 2. The coefficient of skewness (g1) is positive for size distributions with abundant smaller sized trees (saplings) and few larger trees (adults), and is negative for distributions with fewer smaller trees and abundant larger trees. Poorter et al. (1996) interprets greater numbers of smaller trees relative to larger ones as indicative of a stable or a growing tree population, fewer numbers of smaller trees relative to larger ones as indicative of an unstable or declining population.

For B. spiciformis, in which significant skewness occurred in both disturbed and undisturbed habitats, we categorized the trees into four size classes based on diameter: 4–8 cm; 8.1–13.3 cm; 13.4–22 cm and >22.1 cm and tested for differences in tree size distributions between the two habitats by chi-square test.

We used SYSTAT version 10.2 (SPSS Inc., 2000) for all statistical analyses. We report values for tree numbers, densities, diversity and diameter at breast height as mean ± SE. Nomenclature follows Beentje (1994).

Results

Tree species abundance and diversity

The IV, an index that combines the relative density, frequency of occurrence and dominance of each species in the community, is shown in Fig. 1 for the seven principal tree species in the Brachystegia woodland. B. spificiformis had a higher IV in the disturbed habitat than in the undisturbed. Other trees had relatively higher IV’s in the undisturbed habitat than in the disturbed one. Tree species diversity was greater in the undisturbed habitats (D = 3.90 ± 0.31) than in the disturbed (D = 2.47 ± 0.39) habitats (t48 = 2.89, = 0.006).

Figure 1.

 Proportional abundance as represented by importance values (IV’s) of tree species within the disturbed (a) and undisturbed (b) Brachystegia woodland of Arabuko-Sokoke Forest. IV is a composite measure including relative dominance (basal area), relative frequency and relative density. The greater the IV, the more abundant the tree species. Afz = Afzelia quanzensis; Brch = Brachystegia spiciformis, Hyme = Hymenaea verrucosa; Julb = Julbernardia magnistipulata; Lann = Lannea schweinfurthii; Mank = Manilkara sansibarensis; Mimu = Mimusops obtusifolia

Among tree species, B. spificiformis, Hymenaea verrucosa Gaertn. and A. quanzensis showed no significant differences in the number of individual trees between the disturbed and undisturbed habitats (Mann–Whitney U-test, df = 1, > 0.05). In contrast, the number of individuals was higher in the undisturbed habitat than in the disturbed habitat for M. sansibarensis (36.44 ± 6.83 trees ha−1 versus 3.55 ± 1.23 trees ha−1; U1 = 69, < 0.001), M. obtusifolia (46.22 ± 4.78 trees ha−1 versus 1.33 ±0.73 trees ha−1; U1 = 6, < 0.001), and L. schweinfurthii (16.88 ± 2.87 trees ha−1 versus 6.22 ± 1.28 trees ha−1; U1 = 173.5, < 0.001). Overall, tree numbers were higher in undisturbed habitat than disturbed habitat (276.44 ± 15.52 trees ha−1 versus 99.56 ± 7.01 trees ha−1; U1 = 2.0, < 0.001).

Tree species dispersion patterns

Patchiness indices varied considerably among tree species and in some cases between the habitats (Table 1). All of the six tree species that occurred in the disturbed site were distributed randomly. B. spificiformis, J. magnistipulata and M. sansibarensis were significantly clumped in the undisturbed site and A. quanzensis had a statistically nonsignificantly clumped distribution in the undisturbed site. M. obtusifolia, H. verrucosa, L. schweinfurthii were randomly dispersed in the undisturbed site.

Table 1.   Patchiness in spatial dispersion of seven tree species in the disturbed and undisturbed habitats as shown by Llyod’s (L) indices. If tree distribution is clumped, the patchiness index is >1. If trees are evenly distributed, the index is <1. We tested the significance of the nonrandom dispersion (clumped or even) using chi-square test
Tree speciesDisturbedUndisturbed
Mean ± SE (trees per subplots)Lχ2dfP-valueMean ± SE (trees per subplots)Lχ2dfP-value
Brachystegia 6.44 ± 0.611.0735.11240.077.88 ± 0.941.2469.9724<0.001
JulbernardiaAbsentAbsent6.28 ± 1.612.38245.1324<0.001
Manilkara0.32 ± 0.110.9023.25240.513.28 ± 1.611.5769.2224<0.001
Mimusops0.08 ± 0.060.3122.00241.584.08 ± 0.431.0326.73240.31
Hymenaea1.04 ± 0.241.0024.00240.460.64 ± 0.140.6318.38240.76
Lannea0.60 ± 0.130.4916.68240.861.56 ± 0.271.1127.89240.26
Afzelia0.24 ± 0.111.5827.33240.290.32 ± 0.142.5335.75240.06

Tree size distributions

All tree species had similar mean dbh in both disturbed and undisturbed habitats, except L. schweinfurthii (F1, 50 = 8.71, = 0.01). Its dbh was significantly greater in the disturbed habitat than in the undisturbed habitat (8.81 ± 1.04 cm and 6.66 ± 0.23 cm, disturbed and undisturbed respectively). Distributions of dbh for canopy tree species were generally unimodal (Fig. 2). B. spiciformis displayed significantly negative and positive skewness in undisturbed and disturbed habitats respectively. H. verrucosa showed significantly negative skewness in the disturbed habitat. L. schweinfurthii had significantly positive skewness in the undisturbed habitat. Skewness was significantly positive for M. obtusifolia in the undisturbed habitat. J. magnistipulata showed positive skewness in the undisturbed habitat where it occurred.

Figure 2.

 Tree size class distributions of seven tree species in undisturbed and disturbed areas. Coefficient of skewness (g1) represents the degree to which the distributions differ from a normal distribution. Significant coefficients of skewness (*) indicate significant deviation from a symmetric normal distribution

B. spiciformis was the only tree species with significant skewness in both habitats. Interestingly, it had the same mean dbh in both the disturbed and undisturbed habitats (t358 = 0.62, = 0.54; 14.89 ± 0.79 cm and 15.37 ±0.35 cm, disturbed and undisturbed respectively). Despite no difference in mean dbh, the size class distributions of Brachystegia differed significantly between habitats (χ23 = 36.09, < 0.001). Relative to the undisturbed habitat, the disturbed habitat had an over-representation of very small and very large individuals.

Discussion

Tree species composition

A single tree species, B. spiciformis, dominated our site (Fig. 1). The miombo woodlands of south central Africa closely resemble our study. Miombo woodlands, like the Brachystegia woodland of the Arabuko-Sokoke Forest, exhibit a low diversity of canopy tree species (Frost, 1996). The undisturbed portion of the Arabuko-Sokoke Forest superficially resembles the miombo forests of Zimbabwe (Guy, 1989; Grundy, Campbell & Frost, 1994; Grundy, 1995) with B. spiciformis and Julbernardia globiflora (Benth.) Troupin as co-dominants. Similar forests include those with B. spiciformis, Brachystegia boehmii Taub. and J. globiflora in Zambia (Chidumayo, 2002, 2004), and those dominated by B. spiciformis, B. boehmii and Pterocarpus angolensis DC. in Tanzania (Dondeyne et al., 2004). These miombo forests, like Arabuko-Sokoke, have deep, often sandy soils of low fertility.

It remains an open question whether Arabuko-Sokoke should be classified as a form of Miombo. If one prefers broad categories, this represents a northern Miombo forest type. Otherwise, it does have some important differences, namely a taller canopy height (20–25 m instead of the 8–15 m stature of Tanzania (see Fors, 2002), a higher density of trees and summed dbh per unit area, and a greater dominance of B. spiciformis. Either because of precipitation, latitude or other edaphic factors, B. spiciformis grows taller in Arabuko-Sokoke than in forests further south. Selective logging and past disturbances may have amplified the dominance of B. spiciformis over all of the other common and more heavily logged species.

Högberg & Nylund (1981) found ectomycorrhizae occurrence in the roots of these dominant tree species in the miombo woodland. These fungi may facilitate a tree species’ exploitation of the infertile soils more effectively than those without ectomycorrhizae. What factors favour the more extreme dominance of B. spiciformis in the Arabuko-Sokoke Forest is an interesting question but remains largely unanswered; perhaps this uniqueness is a biogeographical anomaly (P. Frost, personal communication).

We found a number of differences in the composition and structure of the undisturbed and disturbed sites within Arabuko-Sokoke. One species, J. magnistipulata, reached a high density in the undisturbed site but was absent in the disturbed one. Although B. spiciformis had the greatest IVs regardless of disturbance level, IVs for the other tree species tended to be greater within the undisturbed site than within the disturbed site. The Simpson’s Diversity Index also reflects this where it is higher in the undisturbed site than in the disturbed site. Tree density was greater in the undisturbed site than in the disturbed site. Four of seven tree species exhibited clumped dispersion in the undisturbed site, but all tree species were randomly dispersed in the disturbed site.

We conclude that the disturbed site represents a recovering forest rather than one that has been disturbed to an earlier successional stage, or which is on a trajectory towards a different forest type. However, by way of management, it may be necessary or desirable to transplant or facilitate the recovery of J. magnistipulata within the disturbed forest. All other species appear to be recovering nicely.

Tree species dispersion

The patterns of dispersion for each of the common tree species supports the hypothesis that forests subjected to frequent, large-scale disturbances exhibit random spatial patterns, whereas clumping occurs in forests where canopy gaps are the major cause of structural change (Armesto et al. (1986). The long-term removal of trees by humans within the disturbed site could thus promote the random dispersion found there. In addition, tree removal of select species should decrease their abundance, and this in turn would lead to declines in their IVs.

The three most abundant tree species in the undisturbed forest showed significantly clumped dispersions. No species exhibited a significantly nonrandom dispersion pattern in the disturbed forest.

Simpson’s Diversity Index incorporates information on both the number and proportional abundance of species (Magurran, 2004). When species richness is similar among habitats, as was the case in our study, this model measures the evenness in population sizes among the species. In this respect, the area with high evenness in tree species abundance will have a higher diversity compared with those with unequal abundances. For the Brachystegia woodland, there was evidence of unequal proportional abundances of the tree species between the disturbed and undisturbed habitats. This resulted in higher evenness in the undisturbed habitat than in the disturbed habitat.

Tree size distributions

The seven tree species studied showed different size distributions in disturbed and undisturbed habitat types raising two questions: (1) which tree species are regenerating and (2) to what extent does disturbance influence recruitment? For slowly growing species, recruitment is interpreted as vibrant when the tree species has a larger number of smaller than larger trees, but recruitment is considered poor when there are larger than smaller trees (Condit et al., 1998; Wright et al., 2003). On the basis of this indicator, we can suggest which tree species may be regenerating successfully in the disturbed and/or undisturbed habitat, and suggest future trends in the forest.

Among all of the tree species, B. spiciformis exhibited contrasting size distributions between the undisturbed habitat and disturbed habitat. In the undisturbed habitat, adults were more abundant than juveniles, but in the disturbed habitat, juveniles were more abundant than adults. This suggests a mature canopy of trees in the intact forest with suppressed seedling establishment and/or sapling survivorship of B. spiciformis. Could B. spiciformis be a light gap dependant species that regenerates primarily after disturbances? Support for this perspective comes from the disturbed forest where the past removal of canopy trees may be leading to many small and possibly recruiting B. spiciformis. Mutangah & Mwaura (1992), in their investigation of intact forest at the transition zone between Brachystegia woodland and mixed forest found high recruitment of B. spiciformis, but most seedlings died off as juveniles with only 30% survivorship to samplings. This supports the idea that B. spiciformis is a light demanding species with better recruitment in disturbed forest. The few extremely large B. spiciformis that we found in the disturbed habitat may have grown unusually large with the removal of canopy competitors.

Four trees species; A. quanzensis, L. schweinfurthii, M. sansibarensis, M. obtusifolia exhibited lower abundances in the disturbed forest. Two of these tree species, L. schweinfurthii and M. obtusifolia had many smaller diameter trees than larger ones in undisturbed habitats. This probably indicates successful recruitment and regeneration of these tree species in the relatively intact forest. While there was no recent stump evidence to suggest recent logging of these trees species in the disturbed habitat, A. quanzensis, and M. sansibarensis could have been logged in the past (Robertson & Luke, 1993). Alternatively, one cannot discount the possibility that some of these tree species had limited recruitment success after removal of B. spiciformis from the disturbed forest. This perspective could apply to H. verrucosa where only larger diameter trees occur in the disturbed habitat, indicating very low current recruitment. One species, the co-dominant J. magnistipulata, has been removed completely from the disturbed forest and it does not appear to be recovering there. Mutangah & Mwaura (1992) found healthy regeneration of J. magnistipulata in the intact forested areas indicating adaptation to undisturbed forest. We favor the hypothesis that the other tree species (with perhaps the exception of H. verrucosa) were more heavily logged than B. spiciformis in the past and that this has reduced tree diversity and fortunately favored the recruitment of all of the persecuted trees save for J. magnistipulata.

In summary, the Brachystegia portion of Arabuko-Sokoke forest, although small in area, represents an important example of what may be a northern variant of the widespread Miombo forest type. Past selective logging has left an indelible signature through the removal of rarer species, leaving the forest unusually dominated by B. spiciformis. The disturbed site bears this out through its preponderance of unusually large (spared trees) and unusually small B. spiciformis (regeneration) relative to the undisturbed forest. The next four tree species indicate reduced abundance in the disturbed habitat. However, two of these tree species are recovering in the undisturbed habitat. Whereas J. magnistipulata has been completely depleted from the disturbed habitat, but abundant in the undisturbed habitat – active management and restoration may be desirable for this species. For the remaining common tree species, active management may not be necessary. Continued protection promises to maintain this remnant forest, and with time, there will be an opportunity to see how the composition and/or recovery continues.

Acknowledgements

This study was funded by the Wildlife Conservation Society, University of Illinois Provost research award, National Science Foundation Dissertation Improvement Grant (DEB 0309368) and the Conservation Training Consortium. Special thanks go to Mathias Mwavita, the Warden at Arabuko-Sokoke Forest for his logistical help in the field. Particular thanks go to Wellington Kombe who guaranteed the project’s success by his continuous support as guide, field assistant and expert naturalist. Anne Robertson, Geofrey Mashauri and Mathias Ngonyo helped with plant identification. Peter G.H. Frost provided a particularly thorough review of an earlier draft that substantially improved the manuscript.

Ancillary