Introduction

Southern African tropical woodland covers an estimated area of about 265,000 km2 extending from southeastern Angola to northeastern Namibia, across northeastern Botswana, southwestern Zambia and northwestern Zimbabwe (Timberlake et al. 2010). The region is dominated by open woodlands with numerous hardwood species, including Baikiaea plurijuga, Pterocarpus angolensis and Guibourtia coleosperma (Barbosa 1970; Strohbach and Petersen 2007). The study area in Caiundo lies within this ecoregion on deep Kalahari sand in a wide belt along the Angolan–Namibian border. Where other important vegetation subunits are present, they comprise Erythrophleum africanum–Bauhinia urbaniana vegetation unit (Baikiaea–Burkea woodlands, dense and medium dense), Diplorhynchus condylocarpon–Gymnosporia senegalensis unit (Baikiaea–Burkea woodlands, open grasslands), Combretum celastroides–Baikiaea plurijuga woodlands (Baikiaea–Burkea woodlands with thicket-like understory) and the Acacia sieberiana–Piliostigma thonningii vegetation unit (grasslands) located on the floodplains (Frantz et al. 2013; Revermann and Finckh 2013). Dry woodlands support over 60% of people in Africa providing them with wide range of environmental goods and services (Shackleton et al. 2010). However, these benefits are threatened by deforestation and forest degradation. Globally, direct drivers of changes in dry tropical forests and woodlands encompass habitat change and land degradation, climate change and extreme weather events, fire and over exploitation of natural resources (Milles et al. 2006).

The area around Caiundo in southeastern Angola contains relatively intact Baikiaea–Burkea woodland (Revermann and Finckh 2013). However, demand for more agricultural areas in postwar Angola has lead to increased clearing of woodlands (Wallenfang et al. 2015). Illegal logging is a serious problem in nearby northern Namibia (Pröpper and Vollan 2013). Although some studies have addressed the Baikiaea woodlands, hardly any investigations have been carried out in Angola. Thus, much still remains to be understood in terms of regional and local variation in community composition, diversity and especially responses to human disturbance (Zimudzi et al. 2013). With this study we aimed to investigate the diversity and composition of woody species in fallows as compared to mature woodlands and to analyze the population structure of harvestable tree species and regeneration status of woody species.

Materials and methods

Study area

The study area is located along the Cubango/Okavango River (Fig. 1), about 90 km south of the confluence of the Cubango and Cuebe rivers in Mulemba village at an elevation of 1155 m.a.s.l. (Weber 2013). The climate is semi-arid with a pronounced rainy season between November and March and a dry period from May to September. October and April are considered transition months (Diniz 1973). During the period 1971–2000, the mean annual precipitation was 732 mm. Over the period 1950–2009, the annual rainfall showed a high interannual variability without any obvious trend (Weber 2013). Caiundo has an annual mean temperature of 22.5 °C and an average temperature of 26.3 °C in the hottest month, October, and 16.8 °C in the coldest month, July (Diniz 1973; Weber 2013). The long-term annual mean temperature shows a low interannual variability with an increase in temperature since the mid 1960s. On average, 17 frost days per year were recorded between June and September (Weber 2013).

Fig. 1
figure 1

(Reproduced with permission from www.future-okavango.org; Wehberg and Weinzierl 2013)

Location of the Okavango basin in southern Africa and the study site in Caiundo with plots location. The Okavango basin follows the definition of the “The Future Okavango project”.

Full size image

The vegetation in the study area represents mostly intact Baikiaea–Burkea woodlands of southeastern Angola, and it is almost entirely restricted to sandy substrates (Revermann and Finckh 2013). Pure stands of Baikiaea plurijugararely occur; typically, they are associated with Shinziophyton rautanenii, Guibourtia coleosperma, Burkea africana, Pterocarpus angolensis, Dialium englerianum, Erythrophleum africanum, Parinari curatellifolia, Bobgunnia madagascariensis and sparsely with Julbernardia paniculata. The woody tree species range from 5 to 12 m in height, and the shrub layer comprises Baphia massaiensis, Copaifera baumiana, Diplorhynchus condylocarpon, Ochna pulchra, Paropsia brazzaeana, Strychnos spp. and Xylopia tomentosa. Grasses are sparse and wiry, up to 1.5 m tall; characteristic species are Trachypogon spicatus, Schizachyrium jeffreysii, Tristachya superba, Aristida stipitata and Digitaria eriantha (Wallenfang et al. 2015). The vegetation here may represent a reference state, as the human impact is relatively low compared to other areas near the river (Revermann and Finckh 2013). However, the area is heavily impacted by fire with high fire return periods of only a few years (Frantz et al. 2013).

The human population density in the study area is relatively low and largely governed by the ease of access to water (Diniz 1973; Frantz et al. 2013). People mostly belong to the Ganguela ethnic group with various subgroups (Diniz 1973). Some significant but isolated and nomadic groups of San people also inhabit the study area. Agriculture, livestock keeping and fishing are the main socio-economic activities. Small businesses were also observed in the most populated villages such as the Savate and Caiundo villages and selling of freshwater fish, bushmeat and other forest goods has become common along the main Katwitwi–Caiundo–Menongue road (personal observations).

Data collection

Sites were selected through informal interviews with traditional authorities of Mulemba village and the wider community, aiming to identify sites with previous agriculture use and known age since abandonment. In contrast to the Miombo region of Cusseque area (Gonçalves et al. 2017), agricultural fields here can be cultivated for longer time spans, and local farmers have reported fields with more than 20 years of continuous cultivation. Thus, identifying suitable fallows for the study was very difficult. However, we were able to identify 10 sites abandoned during the post-electoral conflict (1992–1993). Woodlands were considered as mature if they showed no evidence of past disturbance, e.g., agriculture or wood cutting. Because the western and eastern banks of the river have been reported to have a slightly different species composition (Revermann and Finckh 2013), we included both sides of the river in the sampling scheme.

The fieldwork was carried out in September 2014, using a 20 × 50 m plot design (Felfili et al. 2005). Twenty plots were sampled: 10 in the fallows with an approximate age of ca. 15–20 years and 10 in mature woodlands. Plots were placed randomly in the area of the identified fallows and mature woodlands. On every plot, we measured all individuals with girth equal or above 15 cm, corresponding to a diameter at breast height (DBH) ≥ 5 cm, using a tape measure. Regeneration of woody species was inferred from saplings, which refers to young trees less than 5 cm in DBH, within one central 100 m2 subplot and from size class distribution curves for harvestable woody species.

The species were identified based on the first author’s field experience with Angolan flora and available literature (Mannheimer and Curtis 2009), and by their local names in Ganguela (given by local field guides). Field names were later replaced by scientific names following the identification of voucher specimens, which were deposited in the Herbarium of Lubango (LUBA). We choose to retain the recently published “Checklist of Angolan Vascular Plants” (Figueiredo and Smith 2008) as the standard for scientific names of species and authors, despite some recent changes in botanical nomenclature.

Data analysis

Woody species diversity and composition

Species richness was measured from the total number of species recorded in each stand. Woody species diversity in fallows and mature woodlands was measured by the Shannon–Wiener diversity index (Magurran 2004) calculated as given in Eq. 1. Significant differences in the species diversity and evenness between the fallows and mature woodlands were determined using t tests (α = 0.05). The Shannon–Wiener diversity index H’ was calculated as:

$$ H^{\prime} = - \mathop \sum \limits_{i = 1}^{n} p_{i} {\text{ln }}p_{i} , $$
(1)

where pi is the number of individuals of species in a given plot divided by the total number of individuals in the plot. Species evenness was measured by Pielou’s Evenness Index (Pielou 1966) given in Eq. 2.

$$ E^{\prime} = \frac{{H^{\prime}}}{{H_{ \hbox{max} } }}, $$
(2)

where Hmax is the natural logarithm of S, species richness. Additionally, we calculated diversity profiles for the mature woodlands and fallows. The diversity profile play an important role in terms of diversity comparisons, using α as reference scale parameter (Tóthmérész 1995). The scale parameter α gives the order of Renyi’s diversity with α = 0 refers to species richness, α = 1 gives an index proportional to the Shannon diversity, α = 2 corresponds to the Simpson’s index (Hammer 2012).

Importance value index is an important parameter that indicates the ecological significance of species in a given ecosystem and is being widely used to assess the floristic composition in forest systems, including African savanna and woodlands. The woody species composition of the plots was assessed by the IVI (Eq. 3), which is a summation of the relative values of frequency, density and dominance (Curtis and McIntosh 1951). A non-parametric sign test was used to test differences in IVI between fallows and mature woodlands of most abundant species:

$$ I_{\text{vI}} = \frac{{\left( {R_{\text{F}} + R_{\text{D}} + R_{\text{Do}} } \right)}}{3} $$
(3)

where RF is the relative frequency of species obtained from absolute frequency, dividing the number of sampling units (100 m2 subplots) in which the species occurs, by the total number of sampling units; RD is the relative density of species obtained from absolute density calculated from the total number of individual of a species present in a plot divided by the total area sampled (0.1 ha) and RDo is the relative dominance of species obtained by dividing the total basal area for a given species by the total basal area of all species per plot.

The similarity of woody species between fallows and mature woodlands was calculated from Jaccard similarity index (J’, Eq. 4), as it is a useful measure to determine the extent of overlap of tree species between two forest communities in this case forest stages (Kalaba et al. 2013).

$$ J^{\prime} = \frac{A}{{\left( {A + B + C} \right)}}, $$
(4)

where A is the number of species found in both communities, B is the species in community A and not in B, C is the species in B and not in A (Chidumayo 1997). The index assumes values between 0 and 1, with value 1 indicating that all species co-occur (Jaccard 1912).

To investigate the variation within the species composition, using the abundance of species, we calculated detrended correspondence analysis (DCA). This ordination method is widely used in ecological data because it avoids the arch effect caused by unimodal species response curves by detrending, and rescaling in order to preserve ecological distances within the interactive calculation process (Hill and Gauch Jr. 1980). The axis in DCA are scaled into units of the average standard deviation (SD units), with the length of the first axis referring to the heterogeneity or homogeneity of the data set (Lepš and Šmilauer 2003).

Population structure and regeneration of woody species

Traditionally, forest inventory practices have concentrated on the structural characteristics that are important for timber management (Graz 2004). However, the analysis of population structures provides a deeper insight when assessed for each individual tree species, providing specific information on population dynamics and recruitment, which facilitates prediction of future states of the species in the forest community (Didita et al. 2010).

The population structure in this study was assessed only for Baikiaea plurijuga, Guibourtia coleosperma, Pterocarpus angolensis and Schinziophyton rautanenii that are normally harvested due to their potential economic value. we constructed bar charts using the number of individuals (y-axis) and the DBH, categorized into 13 diameter classes (x-axis), with size class intervals of 5 cm (MacLaren and MacDonald 2003; Neelo et al. 2015). The regeneration status of woody species can be determined from the profile. An inverted J-shaped size class distribution indicates a stable population (Bin et al. 2012), whereas a unimodal J-shape distribution with fewer juveniles than adults is taken as evidence of population decline (Condit et al. 1998).

Under dry tropical conditions, vegetative regeneration is reported to be more effective than seed regeneration, which is more subject to negative anthropogenic effects. Regeneration of woody species was characterized from sapling density, where density is the number of recruits in a unit area (Yang et al. 2014).

Results

Species diversity

A total of 718 individuals were measured, 336 in fallows sites and 382 in mature woodlands sites, corresponding to 30 (12.8 ± 1.9) woody species in fallows and 28 (10.2 ± 1.6) in mature woodlands. We recorded 34 woody species in 32 genera, belonging to 14 families; only one species could not be identified (Table 1). The most diverse family was Fabaceae (12), followed by Euphorbiaceae (5), Combretaceae (3); Annonaceae, Anacardiaceae and Strychnaceae (2); and Apocynaceae, Burseraceae, Celastraceae, Ochnaceae, Polygalaceae, Rhamnaceae, Rubiaceae and Sterculiaceae (1).

Table 1 Stand structure of woody species and importance values index (IVI) of fallows and mature woodlands vegetation plots
Full size table

Species diversity measured by the Shannon diversity index was 2.98 (2.26 ± 0.15) in fallow sites and 2.68 (1.90 ± 0.39) in mature woodlands and was significantly higher in fallows compared to mature woodlands (p = 0.012). Diversity profiles indicated that the fallow plots had greater diversity than in mature woodlands plots, since the diversity profile curve of fallows was higher than that of mature woodlands and the curves did not intersect (Fig. 2).

Fig. 2
figure 2

Diversity profiles using Renyi’s diversity in the fallows (red) and mature woodlands (blue) in the Caiundo study area

Full size image

The overall evenness was 0.88 (0.89 ± 0.03) in fallows sites and 0.80 (0.28 ± 0.12) in mature woodlands, and no significant difference was found between the two stages (p = 0.098), indicating that species abundance distributions were more or less equal in both stages, fallows and mature stands.

Species composition

Twenty-three woody species were recorded in both sites. Seven woody species occurred only in the fallows, while only five were unique to the mature woodlands (Table 1). The most dominant species in both stages based on their IVI were Baikiaea plurijuga, Combretum collinum, Diplorhynchus condylocarpon, Erythrophleum africanum, Ochna pulchra, Pterocarpus angolensis and Schinziophyton rautanenii. Baphia massaiensis subsp. obovata var. obovata, Terminalia sericea and Combretum zeyheri characterized the fallows sites, while the mature woodlands were characterized by the dominance of Burkea africana, Dialium englerianum and Pseudolachnostylis maprouneifolia var. dekindtii. No significant difference (p = 0.754) was found in IVI of most-abundant species between fallows and mature woodlands.

The Jaccard similarity index (J’) indicated high similarity of woody species composition of fallows and mature woodland sites (0.66). This result was also supported by the detrended correspondence analysis (DCA) showing no separation in species composition among the fallows and mature woodlands (Fig. 3). The first and second axes lengths equalled 4.48 and 5.40 and eigenvalues of 0.44 and 0.24, respectively, giving an indication of similar species composition among the two stand stages.

Fig. 3
figure 3

Detrended correspondence analysis (DCA) ordination showing the vegetation sample plots marked by convex hulls in both stand stages, fallows (red dots) and mature woodlands (blue dots)

Full size image

Population structure and regeneration status of woody species

The population structure of the four most harvested woody species had three general size class distribution patterns in the fallows. An irregular J-shape for B. plurijuga resulted from few or no stems in the second and third diameter classes, and a bell-shape for P. angolensis resulted from having more stems in the middle diameter classes and fewer in the lower diameter classes. An interrupted, inverted J-shape was found for S. rautanenii, with more stems in the lower diameter classes and fewer in the higher diameter classes, but with few or no stems in the intermediate diameter classes (Fig. 4). In the mature woodlands, the population structure for B. plurijuga followed a bell-shaped curve and an interrupted inverted J-shape for P. angolensis and S. rautanenii (Fig. 5). For the fourth species with commercially important timber, G. coleosperma, there were too few individuals measured, making it difficult to conclude anything about the population status of the species in either the fallows or mature woodlands. In the fallows, the saplings had 104 stems 0.1 ha−1, ranging from 4 to 21 stems, but only 43 stems 0.1 ha−1, with a range between 2 and 16 stems in the mature woodlands. In the fallows, Baikiaea plurijuga, Diplorhynchus condylocarpon and Schinziophyton rautanenii had the highest density of saplings with 11 individuals (ind.), followed by Combretum zeyheri (10 ind.); Baphia massaiensis subsp. obovata var. obovata and Terminalia sericea (9 ind.); Combretum collinum and Pterocarpus angolensis (7 ind.); Burkea africana (6 ind.); Ochna pulchra, Bauhinia urbaniana, Bobgunnia madagascariensis, Dialium englerianum, Hymenocardia acida var. acida, Pseudolachnostylis maprouneifolia var. dekindtii, Vangueriopsis lanciflora and Xylopia tomentosa (2 ind.); and Guibourtia coleosperma, Hexalobus monopetalus and Oldfieldia dactylophylla (1 ind.).

Fig. 4
figure 4

Size class distribution of three harvestable woody species in the fallows sites

Full size image
Fig. 5
figure 5

Size class distribution of three harvestable woody species in mature woodlands

Full size image

In contrast, in the mature woodland, the most common saplings belonged to D. condylocarpon (10 ind.); O. pulchra and P. angolensis (6 ind.); Strychnos pungens (4 ind.), B. plurijuga, B. africana, C. collinum, Dombeya burgessiae, Erythrophleum africanum, Hymenocardia acida var. acida and O. dactylophylla (2 ind.); and B. madagascariensis, D. englerianum and S. rautanenii (1 ind.).

Discussion

Woody species diversity and composition

The canopy tree species Burkea africana, Erythrophleum africanum, Guibourtia coleosperma and Pterocarpus angolensis occurred in almost all plant communities as did the species of the lower tree layer Combretum collinum, Diplorhynchus condylocarpon, Ochna pulchra, and Pseudolachnostylis maprouneifolia var. dekindtii. These species can therefore be considered characteristic for the woody vegetation of the Cubango Basin as a whole.

Fallow sites varied slightly in species diversity as measured by the Shannon diversity index compared to mature woodlands, probably caused by the relatively high species richness in the fallows sites; this index may increase as richness also increases for a given pattern of evenness (Buzas and Hayek 2005; Colwell 2009), but also because the latter stages of abandonment are characterized by a reduction of the dominance of early-successional and fire resistant species and an increase in slow-growing, fire-tender tree species commonly associated with old-growth habitats (McNicol et al. 2015).

The overall species diversity in the fallows and mature stages was higher than reported for other woodlands of the ecoregion (Nduwayezu et al. 2015; Neelo et al. 2015), although these other studies used a smaller sample plot size, considered the minimal sampling area for upland forests and woodlands (Kent and Coker 1992). Our results are therefore not directly comparable as the size of sample plots used may have a strong effect on the obtained results. Comparing diversity across different studies may also be an inadequate approach, since species diversity is dependent also on local and environmental conditions. The use of larger sampling areas is to be encouraged, as frequency of species and diversity levels increase with sampling effort (Kacholi 2014).

Evenness measures the homogeneity of abundances in a sample or in the community (Colwell 2009). The overall evenness values found in the present study were more or less similar in the fallows and mature sites, implying that individuals of different species have moderately similar abundances in both stages (Neelo et al. 2015). D. condylocarpon and Schinziophyton rautanenii had the highest IVI. In general, these species are not commonly harvested for timber in the studied area. D. condylocarpon was reported to be harvested occasionally for firewood and medicinal uses, while S. rautanenii is harvested for multiple uses, including making crafts and other utensils (Kissanga 2016). Attention should also be paid to other species with a lower IVI, as information gathered locally cites most of these woody species to be used by local communities, providing them with a wide range of goods such as wild fruits and medicinal products. The majority of the fallows in Caiundo are in old floodplains, one of the preferred areas for cultivation. These fallows are currently in a stage of intermediate shrubland (Gröngröft et al. 2013). The abandonment of these fields for long periods may have been due to security issues during the long civil war in the recent past. These areas are strongly affected by fire, since most of the fallow sites and mature woodlands are burned annually during the dry season. Fire may thus explain the lower density of hardwood species (i.e. B. plurijuga, G. coleospema and P. angolensis) in the fallows. Disturbance caused by humans also profoundly contributes to the degradation of natural vegetation in Baikiaea woodlands, generally resulting in secondary formations as observed in the fallows sites, that are dominated by Acacia erioloba, Combretum collinum and Terminalia sericea, with or without Baikiaea plurijuga. Pure stands of B. plurijuga were also only rarely recorded in northern Namibia (De Cauwer et al. 2016).

Species composition similarity measured by the Jaccard similarity index was high, showing that the two woodland stages had almost the same species composition, is in accordance with the DCA, that indicated a strong overlap in species composition and distribution. A similar study in the nearby communal area of Savate also demonstrated the same trend (Wallenfang et al. 2015).

Population structure and regeneration status of woody species

Classifying tree distribution according to size classes is one of the most used methods in plant population biology because it is a useful indicator of population structure and dynamics in forest systems (Bin et al. 2012). The population structure analysis of harvestable woody species in general indicated that there were fewer larger trees compared to smaller ones. The large proportion of small size classes found for P. angolensis and S. rautanenii in the fallows compared to large trees may represent a viable recruitment of the woody species (Lykke 1998). B. plurijuga had low recruitment, probably due to disturbance caused by fire frequency and intensity, as the area is usually burned annually (Frantz et al. 2013). Intentional burning negatively affects woodlands, as it leads to reduced basal area and density of harvestable trees, also decreasing shoot production and growth rates (Gambiza et al. 2000; Ncube and Mufandaedza 2013).

Natural regeneration is defined as the renewal of forest stands by natural seedling, sprouting, suckering or by layering seeds that may be deposited by wind, birds or mammals (Pardos et al. 2005). In dry tropical forests, edaphic and climatic constraints are the main limiting factors of tree growth and seedlings survival rates. In these habitats, shoots perform better than seedlings because they obtain water, nutrients and carbohydrates directly from the well-developed root system of the mother tree (Lévesque et al. 2011).

Baikiaea plurijuga may also regenerate easily from seeds, developing a root system that can grow fast during the first rains, reaching about 1.5 m after three seasons (Calvert 1986). However, its seeds and seedlings are eaten by rodents, birds, duikers and monkeys, limiting its natural regeneration.

The low proportion of saplings recorded in Caiundo may be related to fire frequency and seasonality, since fires in the Angolan/Namibian border mainly occur in August and September, the late season for fire. Fires in this season are generally hotter, consuming more biomass and affecting progressively larger stems and emerging shoots (Stellmes et al. 2013; Burger et al. 1993). Another study suggests that more intense fires, combined with increasing competition from grasses and shrubs, may severely hamper regeneration of Baikiaea woodlands after disturbance.

G. coleosperma had fewer larger stems and consequently fewer or no smaller stems in both stand stages, supporting the view that this species together with P. angolensis is subjected to selective logging as also reported in other studies of the ecoregion (Syampungani 2008; Phiri et al. 2012). Past disturbance history caused by massive timber exploitation may also explain the few large-stemmed individuals of G. coleosperma found, as this species is reported to show slow growth rates, with a mean annual increase in bole diameter of about 3 mm, and survival rates of only 5%. Both species were also reported as the most harvested and exploited timber species in the Cuando-Cubango Province at least during the colonial era (Baptista 2014).

P. angolensis exhibited stable population based on the higher number of individuals in lower diameter classes compared to larger stems, both in fallows and mature woodlands, which is an indication of a good recruitment. The species may also regenerate from seeds, but only 2% of seeds, mainly dispersed by wind, is reported to germinate. The young P. angolensis trees often remain in a suffrutex stages that may last up to 14 years (Thunström 2012), due to annual dieback of seedlings every dry season, a phenomenon characteristic of most woodland species (Boaler 1967).

Regarding S. rautanenii, little is known about seedling development and seedling survival rates. Experiments conducted in neighbouring Namibia suggest that S. rautanenii seedlings may grow faster in light and moderate shade conditions, with germination rates around 26% (Graz 2003; Rønne and Jøker 2006), implying that environments totally open as encountered in the studied area, may facilitate the germination and establishment of seedlings, although frequent fires reduce the ability of species to emerge. Another important factor to consider is that the plant is fire sensitive due to its light and soft wood, and it is mainly harvested for floats, canoes, crafts and other utensils (Rønne and Jøker 2006), which may explain the absence of larger stems in the population in the fallows and the mature woodlands.

Conclusions

We conclude that the Caiundo woodlands are dominated by hardwood species characteristic of the tropical woodlands of southern Africa. Species diversity was higher in fallows compared to mature woodlands. The fallows reached high similarity in terms of species composition compared to mature woodlands. The population structure showed three different patterns in the fallows and mature woodlands. The abundance of young trees in the lower diameter classes (which characterize most of the natural forests systems) indicates that regeneration is taking place. However, the density of saplings maybe strongly affected by fire frequency and intensity, which also may negatively influence seed germination and seedling survival rates. On the other hand, the absence of larger stems is an indication that the preferred timber species are subject to selective logging in the studied area. Controlling the harvest of these targeted woody species and preventing fire in the woodlands for long periods, long enough to allow the establishment of tree saplings, is probably the best way to promote regeneration of woody species and population stability of the targeted species.