Impacts of Lantana camara Invasion on Native Woody Species and Soil Nutrients in the Kavango–Zambezi Transfrontier Conservation Area, Zimbabwe

Abstract

Invasive alien species such as Lantana camara L. impact native species and soil properties, but context-specific effects in transfrontier conservation areas remain poorly understood. Understanding these effects is essential for biodiversity conservation and management. We assessed associations between L. camara presence and native woody species composition and structure, as well as soil nutrients, in protected and communal areas within the Kavango–Zambezi Transfrontier Conservation Area (KAZA TFCA), Zimbabwe. The study hypothesised that invasion effects on vegetation are stronger in communal areas due to higher disturbance, and that soil changes are influenced by land-use intensity. We used stratified random sampling to select 60 plots across invaded and uninvaded sites. Woody vegetation was assessed for species composition and richness, stem density, canopy cover %, height, and diameter at breast height. Soil samples were analysed for nitrogen, organic carbon, phosphorus, potassium, and pH. The presence of L. camara was negatively associated with native species richness, density, height, and canopy cover %, with stronger effects in communal plots. Invaded plots had lower pH (e.g., 6.1 in Park areas) and higher levels of some soil nutrients, particularly phosphorus and organic carbon, though patterns varied by land-use type. These results suggest that anthropogenic disturbance amplifies invasion impacts. We conclude that L. camara reduces native vegetation diversity and structure in this species-rich transfrontier area. Management should prioritise control at communal edges to support woody species resilience, ecosystem services, and biodiversity, with strategies adapted to local land-use conditions.

1. Introduction

Invasive alien species (IAS) rank among the primary drivers of global biodiversity decline [1], with invasive alien plants (IAPs) playing a particularly prominent role. One widespread and highly problematic IAP is Lantana camara L., an evergreen, aromatic shrub of the family Verbenaceae [2]. It thrives in disturbed environments [3] and frequently dominates invaded habitats, strongly suppressing native vegetation.

L. camara has successfully invaded multiple ecosystem types, including grasslands, woodlands, and forests [4], across all continents except Antarctica. Invasions are generally facilitated by habitat disturbance, such that L. camara rarely invades intact forests and cannot persist under dense canopies unless gaps are created [5,6]. The species frequently forms pure stands in open habitats but can also grow within diverse plant communities [7,8]. Its invasive capacity is further enhanced by climate change in some regions [9]. Although studies have documented L. camara’s impacts on native vegetation in African savannas [2,6,8], there is limited information on its effects in transfrontier conservation areas such as the Kavango–Zambezi Transfrontier Conservation Area (KAZA TFCA), where communal and protected lands occur side by side.

Some impacts of L. camara are mediated through allelopathic effects, which hinder regeneration and growth of neighbouring plant species [6,10]. Coupled with its rapid growth, L. camara presence is typically associated with lower native species richness and diversity, plant density, and seedling recruitment [7,11]. For example, Ruwanza [11] demonstrated in northern South Africa that L. camara presence is associated with major shifts in vegetation structure.

The impacts of L. camara on native vegetation also have secondary effects on fauna. For instance, reductions in grazing areas and forage availability have been linked to negative effects on sable antelope (Hippotragus niger) and plains zebra (Equus quagga) in South Africa [12]. In India, habitat alteration associated with L. camara presence has been linked to declines in insectivorous and canopy bird species [13]. In addition to allelopathy and competition, L. camara increases fire risk and intensity due to high flammability from aromatic leaf oils [14] and increased fuel loads [5]. Altered fire regimes, in turn, affect woody species composition and density. Sundaram and Hiremath [15], for example, reported positive feedback between L. camara presence and fire that altered forest composition and ecosystem functioning in India.

Within soil ecosystems, L. camara presence is often associated with changes in soil properties, including pH and nutrient composition, which may favour its own growth [16,17]. Mandiporera et al. [18] showed that L. camara presence is associated with increased soil carbon, phosphorus, and moisture, changes likely to influence native species growth while facilitating further invasion. Other studies have reported increases in calcium, magnesium, potassium, and soil pH in invaded sites [19]. Similarly, Mahla and Mlambo [20] found that L. camara presence was associated with increased soil carbon, nitrogen, and pH.

Despite extensive research on the associations between L. camara presence and vegetation and soil properties, generalisations regarding IAS impacts remain challenging because effects are often site and species-specific [21,22]. Native woody species constitute a significant component of biodiversity and provide essential ecosystem services that support local livelihoods [23]. The Hwange District in western Zimbabwe forms a critical part of the Kavango–Zambezi Transfrontier Conservation Area (KAZA TFCA). Within this region, the spread of L. camara has raised significant concerns among local leadership, scientists, conservation practitioners, land managers, and policymakers [24,25]. In addition, rural communities in this landscape are already navigating complex livelihood challenges linked to environmental change and human–wildlife interactions [26], which may compound the impacts of invasive species. The increasing threat posed by L. camara in such an important ecological zone, therefore, necessitates a deeper understanding of its localised impacts and effective mitigation strategies to safeguard both biodiversity and the livelihoods dependent on these native ecosystems [27].

The presence of L. camara within the KAZA TFCA is potentially associated with threats to native woody species, with implications for ecosystem functioning, resilience, and ecosystem services, including livestock fodder, medicinal and food plants, and the ecotourism-based economy [23,28,29]. However, research on L. camara in Zimbabwe remains limited, with most available studies focused on other southern African countries, particularly South Africa [11,18]. Critically, few investigations compare protected (low-disturbance) and communal (high-disturbance) land-use contexts within the same transfrontier landscape, constraining the development of context-specific control strategies in large-scale conservation areas like KAZA TFCA.

To address these gaps, we tested the following hypotheses:

H1. 

L. camara presence is associated with lower native woody species richness, density, height, and canopy cover %, with stronger negative effects in communal areas due to higher anthropogenic disturbance and land-use intensity.

H2. 

Invaded sites exhibit altered soil properties (e.g., elevated organic carbon and phosphorus, lower pH), but these changes are context dependent and primarily modulated by pre-existing land-use conditions rather than direct invasion effects alone.

Consequently, this study aimed to assess the associations between L. camara presence and native woody species composition and structure, as well as soil nutrients, within protected and communal areas of the KAZA TFCA. Specifically, the study addressed two questions:

(i)

What associations exist between L. camara presence and native woody species composition and structure in the KAZA TFCA? And

(ii)

What associations exist between L. camara presence and soil nutrients in the KAZA TFCA?

2. Materials and Methods

2.1. Study Site

The study was conducted in north-western Zimbabwe within a portion of the Kavango–Zambezi Transfrontier Conservation Area (KAZA TFCA), specifically in Zambezi National Park (hereafter, the Park) and the adjacent Ndlovu Communal Area (hereafter, the Communal Area) (Figure 1). The KAZA TFCA, established in 2011, is the world’s largest transfrontier conservation area, spanning approximately 520,000 km² across Angola, Botswana, Namibia, Zambia, and Zimbabwe (Figure 1) [24].

Zambezi National Park covers approximately 56,000 ha along the Zambezi River and is a protected area where human activities are minimal and largely limited to tourism and park management. In contrast, the Communal Area is characterised by multiple land-use activities, including small-scale subsistence agriculture, livestock grazing, fuelwood collection, and settlement expansion, resulting in comparatively higher levels of anthropogenic disturbance [25,26,28]. This contrasting land-use context (protected vs. communal) was deliberately selected to allow direct comparison of invasion effects while capturing variation in anthropogenic pressure and conservation status across the transfrontier landscape.

The study area falls within Agro-ecological Regions IV and V of Zimbabwe and receives low annual rainfall (approximately 400–500 mm), with most precipitation occurring during the summer season (October–April) [29,30]. Mean daily temperatures range from approximately 19 °C in winter to 30 °C in summer [31]. Soils are predominantly Kalahari sands with low fertility due to heavy leaching, although isolated patches of sandy loam occur [32]. Vegetation is mainly deciduous savanna woodland dominated by Afzelia quanzensis, Baikiaea plurijuga, Brachystegia boehmii, Colophospermum mopane, Combretum collinum, and Pterocarpus angolensis [33].

2.2. Sampling Design and Plot Selection

A reconnaissance survey was conducted to identify areas invaded by L. camara within both the Park and the Communal Area. Each area was subsequently stratified into “invaded” (visible L. camara cover present) and “uninvaded” (no visible or presence of L. camara) zones.

A stratified random sampling design was employed to assess differences between invaded and uninvaded sites across the two land-use categories (Park and Communal). The study area was divided into a 1 km × 1 km Universal Transverse Mercator (UTM) grid, within which potential sampling locations were randomly selected using Esri ArcGIS ArcMap version 10.6/ArcGIS Pro (Esri, Redlands, CA, USA) GIS software. Sampling plots measured 25 m × 25 m (~625 m²), a size selected to adequately capture woody vegetation structure, density, and canopy heterogeneity in deciduous savanna woodlands typical of the region [34]. This intermediate scale provides reliable estimates for tree and shrub variables in patchy invasions while remaining logistically feasible, consistent with practices in southern African savanna and invasion ecology research (e.g., comparable to 20 m × 20 m or larger plots used in woody vegetation assessments in semi-arid savannas; larger than the 10 m × 10 m plots often employed for herbaceous or seedling sampling in L. camara studies [12,34]. The research design was specifically selected to capture variation in invasion and land-use while controlling for site-level environmental factors. Invaded and uninvaded plots were paired within each site, allowing improved assessment of both main effects and interactions.

To minimise spatial autocorrelation, sampling plots were separated by a minimum distance of 200 m. Where feasible, invaded and uninvaded plots were paired within the same land-use category (Park or Communal). However, suitable uninvaded patches were limited in the Communal Area owing to extensive L. camara invasion and elevated anthropogenic disturbance. Consequently, the sampling was unbalanced, with more invaded than uninvaded plots overall (Park: n = 12 invaded, n = 6 uninvaded; Communal: n = 34 invaded, n = 8 uninvaded; total n = 60 plots). The smaller number of uninvaded plots reflected the scarcity of ecologically comparable reference sites devoid of L. camara, particularly in the Communal Area. Invaded plots were therefore intentionally oversampled to adequately represent the full gradient of invasion intensity and variability, while uninvaded plots were selected to maximise ecological similarity and serve as robust controls. This design choice mirrors real-world landscape patterns of invasion rather than constituting sampling bias and is consistent with established practices in invasion ecology, where the spatial dominance of the invader frequently constrains the availability of balanced uninvaded comparators [11,35].

2.3. Vegetation Attributes

Field data collection was conducted at the end of the rainy season (April–May 2018) to facilitate accurate plant identification. Native woody species were defined as hard-stemmed, self-supporting plants native to the region with a diameter at breast height (DBH) ≥ 5 cm. Species identification was undertaken using appropriate field guides [36,37]. In each plot, all woody plants were identified, and canopy cover %, DBH, and stem height were measured.

Crown cover was estimated by measuring the longest canopy diameter and the diameter perpendicular to it [36]. Mean canopy diameter was calculated as:

$$\mathrm{D} = {\displaystyle \frac{D_{\mathrm{l}\mathrm{o}\mathrm{n}\mathrm{g}}+D_{\mathrm{p}\mathrm{e}\mathrm{r}\mathrm{p}}}{2}}$$

where D represents the mean crown diameter. Values were subsequently converted to percentages.

Canopy cover (%) was calculated as follows,

$$\left({\displaystyle \frac{\sum _{i=1}^n\pi {\left(\frac{D_i}{2}\right)}^2}{\mathrm{A}}}\right)\times 100$$

where $D_i$ is the canopy diameter (m) of each woody plant in the plot, A is plot area, and the summation is over all woody plants recorded in the 25 m × 25 m plot (area = 625 m²).

Plant height was estimated using a ranging rod, and for multi-stemmed individuals, the height of the tallest stem was recorded [36,38]. DBH was measured using a diameter tape. For multi-stemmed individuals, mean DBH was calculated using the quadratic mean formula [39]:

$${\mathrm{D}\mathrm{B}\mathrm{H}}_{\mathrm{m}\mathrm{e}\mathrm{a}\mathrm{n}}={\displaystyle \frac{\sum _{i=1}^n{d}_i}{n}}$$

where d_i represents the diameter of individual stems (m) and n is the number of stems.

2.4. Soil Attributes

Five soil samples were collected per plot, one from each corner and one from the centre. These were homogenised to form a composite sample [37]. Samples were collected at depths of 10–15 cm and sieved using a 2 mm mesh to remove stones and coarse organic matter.

In total, 60 composite soil samples (one per plot) were collected; however, due to analytical cost constraints, soil chemical analyses were conducted on a subset of 36 samples (20 from the Communal Area and 16 from the Park). Vegetation data were analysed using the full plot dataset, whereas analyses involving soil variables and soil–vegetation relationships are based only on this subset, and this limitation is acknowledged in the Discussion.

Soil analyses included nitrogen (N), phosphorus (P), potassium (K), pH, and organic carbon (OC). Nitrate nitrogen (NO₃⁻–N) was determined colorimetrically using a Shimadzu UV-1800 spectrophotometer (Copenhagen Nanosystems A/S, Copenhagen, Denmark), following extraction with a 1:10 soil:1 M KCl solution. Available phosphorus was measured using the Mehlich-3 extraction followed by colorimetric analysis [40]. Exchangeable potassium was analysed using flame emission spectrometry (Varian AA 200 Varian Inc., Palo Alto, CA, USA) after Mehlich extraction. Organic carbon was determined using the Walkley–Black method [41]. All analyses were conducted at the Department of Applied Chemistry Laboratory, National University of Science and Technology (NUST), Zimbabwe.

2.5. Data Analysis

Species diversity was calculated using the Shannon–Wiener diversity index (H′) in PAST software (version 3.26, Øyvind Hammer, Natural History Museum, University of Oslo, Norway). Species density was calculated as stems per hectare, while species frequency of occupancy was calculated as the proportion of plots in which each species occurred.

Normality and homogeneity of variance were assessed using Shapiro–Wilk and Levene’s tests, respectively. Two-way permutational multivariate analysis of variance (PERMANOVA) was used to examine the effects of site and invasion status on woody species diversity and vegetation attributes. Two-way analysis of variance (ANOVA) was applied to normally distributed variables, while the non-parametric Mann–Whitney U test was used to compare invaded Communal and Park sites for L. camara structural attributes, species richness, and diversity.

A two-way analysis of similarities (ANOSIM) was used to test differences in woody species composition between sites and invasion categories [42]. Non-metric multidimensional scaling (NMDS), based on Bray–Curtis dissimilarities, was used to visualise patterns in species composition [42]. NMDS was performed in two dimensions, and ordinations with final stress ≤ 0.10 were considered a good fit. Analysis of similarity (ANOSIM; 999 permutations) was used to test for significant differences among site groupings. Cluster analysis was employed to illustrate similarities between invaded and uninvaded sites. ANOSIM and cluster analyses were conducted using PRIMER v6 with the PERMANOVA+ add-on (PRIMER-E Ltd., Auckland, New Zealand) while NMDS was performed using the Vegan Community Ecology Package in R (Version 4.3-1, R Core Team, Vienna, Austria) with vegan package (version 2.6-4) [43].

Detrended Correspondence Analysis (DCA) was used to determine gradient lengths and select an appropriate ordination model [44,45]. Canonical Correspondence Analysis (CCA) was then applied to examine relationships between woody species composition and soil variables, as gradient lengths exceeded 3.5 standard deviation units. Forward selection with Monte Carlo permutation tests (999 permutations, p < 0.05) was used to assess model significance and identify key soil variables. Pearson correlation analysis was conducted to examine relationships between soil variables and species diversity metrics. Ordination analyses were conducted in Vegan [43], while correlation analyses were performed using STATISTICA v13 (TIBCO Software Inc., Palo Alto, CA, USA).

Statistical analyses were performed in R (version 4.3.1) using the packages vegan, agricolae, and multcomp. Data were tested for normality and homogeneity of variances using the Shapiro–Wilk and Levene’s tests, respectively. Variables that met the assumptions of normality and homogeneity of variances were treated as parametric and analysed using two-way analysis of variance (ANOVA), followed by Tukey’s HSD post hoc tests; these included soil chemical and structural variables such as soil pH, potassium, and canopy cover percentage. Variables that violated these assumptions were treated as non-parametric and analysed using Kruskal–Wallis tests, followed by Dunn’s post hoc tests; these included community and density-related variables such as species richness, stem density, and phosphorus concentration. Significance was assessed at p < 0.05. Different lowercase letters within rows in tables indicate statistically significant differences between groups (p < 0.05) based on the respective post hoc tests.

3. Results

3.1. Vegetation Structure and Diversity

Invaded plots were markedly different from uninvaded plots with respect to all investigated variables, and this pattern was consistent across both the Park and Communal areas (

Table 1. Comparison of native woody vegetation structure and diversity metrics between invaded and uninvaded plots across land-use categories (Park and Communal Area) in the KAZA TFCA, Zimbabwe. Values are means ± standard deviation. A total of 60 plots were sampled (invaded: n = 46; uninvaded: n = 14), comprising Park invaded (n = 12), Park uninvaded (n = 6), Communal invaded (n = 34), and Communal uninvaded (n = 8). Statistical comparisons were performed using two-way ANOVA (parametric variables) or Kruskal–Wallis tests (non-parametric variables), followed by Tukey HSD or Dunn’s post hoc tests, respectively (p < 0.05). Different superscript letters within a row indicate statistically significant differences between groups (p < 0.05). ANOVA F-values and p-values are provided for parametric variables; Kruskal–Wallis chi-square and p-values for non-parametric variables.

Variables	Communal Area		Park Area
	Invaded	Uninvaded	Invaded	Uninvaded	Test Statistic Value	p-Value
Native Woody Species
Stem height (m)	5.1 ± 1.6 c	9.5 ± 0.6 ab	8.5 ± 2.0 b	10.7 ± 0.8 a	31.52	0.001
Stem diameter (m)	0.4 ± 0.1 b	0.6 ± 0.02 a	0.7 ± 0.4 a	0.8 ± 0.1 a	24.68	0.001
Density/ha	228 ± 308 b	1 308 ± 432 a	479 ± 284 a	1 561 ± 706 a	18.59	0.002
Canopy cover (%)	6.6 ± 5.9 b	81 ± 14.7 a	9 ± 10.5 b	72.5 ± 9.4 a	7.26	0.01
Shannon–Wiener index	0.4 ± 0.6 c	2.8 ± 0.4 a	1.6 ± 0.5 b	3.2 ± 0.4 a	74.97	0.001
Species richness	5.4 ± 3.3 d	21.3 ± 5.1 b	12.7 ± 4.1 c	33.5 ± 0.3 a	51.02	0.001
L. camara	Invaded	Uninvaded	Invaded	Uninvaded	Z-value	p-value
Stem height (m)	2.2 ± 0.8 a	N/A	1.8 ± 0.1 b	N/A	3.4	0.001
Stem diameter (m)	0.04 ± 0.01 b	N/A	0.07 ± 0.10 a	N/A	1.1	0.04
Density/ha	5 542 ± 4 747 a	N/A	747 ± 567 b	N/A	4.5	0.001
Canopy cover (%)	47.4 ± 27.8 a	N/A	10.3 ±2.0 b	N/A	4.2	0.001

). Invaded plots had significantly lower woody plant stem height, stem diameter, density, canopy cover %, diversity, and species richness than uninvaded plots (Table 1). The most pronounced difference was observed in native species canopy cover %, which declined from 72.5–81% (±9.4–14.7 SE) in uninvaded plots to 6.6–9% (±5.9–10.5 SE) in invaded plots (Table 1). Stem density also showed a marked reduction due to invasion (p < 0.002; F = 18.59), with uninvaded plots in both the Park and Communal areas having higher stem density (1308–1561 stems/ha) than invaded plots (228–479 stems/ha; Table 1).

Woody plant species richness differed significantly, with the highest richness in uninvaded Park plots, followed by uninvaded Communal plots, and then invaded sites (Table 1). Consequently, the Shannon–Wiener index indicated higher diversity in uninvaded plots than in invaded ones. There were no significant differences in species diversity between sites within the same invasion category (Table 1). Comparison of L. camara between the Communal Area and Park revealed approximately seven times higher density (Z = 4.5; p = 0.001) and more than four times greater canopy cover % (Z = 4.2; p = 0.001) in the Communal Area than in the Park. Mean height of L. camara was also approximately double in the Communal Area compared to the Park (Z = 3.4; p = 0.001; Table 1).

3.2. Species Composition and Its Association with the Invasion of L. camara

Fifty native woody species from 21 families were recorded (

Table 2. Frequency of occupancy (%) of native woody species recorded in invaded and uninvaded plots within Park and Communal land-use categories. Frequencies represent the proportion of plots in which each species occurred within each invasion category (invaded: n = 46; uninvaded: n = 14, pooled across land-use types). This table describes patterns of species presence across invasion categories and does not represent species richness or abundance.

Species Name	Communal Area		Park Area
	Invaded	Uninvaded	Invaded	Uninvaded
Adansonia digitata	0	0	9	100
Afzelia quanzensis	3	75	17	83
Albizia harveyi	12	100	25	100
Albizia nigricans	6	50	17	33
Antidesma venosum	0	0	9	50
Azanza garckeana	0	0	0	50
Baikiaea plurijuga	3	63	33	33
Bauhinia petersiana	30	100	68	83
Berchemia discolor	0	0	33	83
Brachystegia petersiana	0	0	33	83
Brachystegia spiciformis	12	75	9	50
Bridelia mollis	0	0	25	67
Burkea africana	6	75	0	0
Colophospermum mopane	15	75	50	83
Combretum apiculatum	0	63	9	17
Combretum hereroense	0	0	17	83
Combretum imberbe	53	88	92	100
Combretum molle	44	75	17	33
Corymbia gummifera	0	0	0	67
Dichrostachys cinerea	9	75	50	100
Diospyros lycioides	0	0	17	83
Diospyros mespiliformis	0	0	50	50
Diospyros quiloensis	0	0	42	33
Diplorhynchus condylocarpon	21	100	17	100
Euclea divinorum	38	88	33	100
Ficus natalensis	6	88	17	33
Ficus petersii	0	0	9	50
Friesodielsia obovata	0	0	25	67
Grewia flavescens	24	100	75	100
Gymnosporia senegalensis	9	75	25	50
Hyphaene coriacea	0	0	9	8
Kirkia acuminata	0	0	17	67
Kigelia africana	0	0	9	33
Lantana camara	100	0	100	0
Manilkara mochisia	0	0	9	50
Ozoroa reticulata	0	0	33	67
Peltophorum africanum	0	0	25	83
Philenoptera violacea	0	0	17	50
Phoenix reclinata	0	0	9	83
Piliostigma thonningii	0	0	9	100
Pseudolachnostylis maprouneifolia	0	0	17	50
Pterocarpus rotundifolius	3	100	25	33
Sclerocarya birrea subsp. caffra	18	75	25	33
Senegalia galpinii	32	63	33	67
Senegalia nigrescens	12	88	9	17
Strychnos potatorum	0	0	0	50
Terminalia sericea	21	88	67	100
Terminalia stuhlmannii	0	0	42	67
Trichilia emetica	0	0	17	100
Vachellia erioloba	15	50	25	83
Vachellia karroo	12	63	33	67
Vangueria infausta	24	50	8	100
Ximenia americana	0	0	17	50
Ximenia caffra	0	0	17	33
Ziziphus mucronata	15	88	25	83

). Uninvaded plots in the Park had the most species (48), whereas the invaded plots in the Communal Area had the lowest number (25). Generally, woody species were more abundant in the uninvaded sites; this was particularly so in Communal areas (ANOSIM R = 0.8; p = 0.001) (

Table 3. Analysis of Similarities (ANOSIM) results showing pairwise differences in woody species composition among invasion status and land-use categories, based on Bray–Curtis similarity. Reported values include the Global R statistic and associated significance levels (p-values).

Groups	Global R: 0.62; p = 0.001
	R Statistic	Significance Level (p-Value)
Invaded Communal, Uninvaded Communal	0.8	0.001
Invaded Communal, Invaded Park	0.4	0.002
Invaded Communal, Uninvaded Park	0.9	0.001
Invaded Communal, Invaded Park	0.7	0.001
Uninvaded Communal, Uninvaded Park	0.5	0.001
Invaded Park, Uninvaded Park	0.6	0.002

). The same distribution pattern was evident in the Park, except for species such as Diospyros mespiliformis and Hyphaene coriacea that had equal or greater frequency in invaded plots than in uninvaded plots. Apart from the plots in the Park having more frequently occurring species, the ANOSIM revealed that the differentials between invaded and uninvaded conditions were greater in the Communal plots (Table 3).

The most frequently occurring species, with an occupancy frequency greater than 70% across uninvaded sites, were Afzelia quanzensis, Albizia harveyi, Bauhinia petersiana, Colophospermum mopane, Combretum imberbe, Dichrostachys cinerea, Diplorhynchus condylocarpon, Euclea divinorum, Grewia flavescens, and Terminalia sericea. Only Combretum imberbe had high frequencies (greater than 50%) in the invaded sites (Table 2).

3.3. Species Composition-Ordination and Clustering

Building on the patterns observed in the NMDS ordination (Figure 2), hierarchical cluster analysis identified four clusters based on compositional similarity (Figure 3). Consistent with the NMDS ordination, these clusters were not discrete and showed overlap among invasion and land-use categories. Cluster A was largely composed of invaded Communal sites, whereas Cluster B was primarily associated with invaded Park sites. Cluster C consisted mostly of uninvaded sites from both land-use types, while Cluster D included a mixture of invaded and uninvaded sites and occupied an intermediate position in the ordination space, reflecting transitional assemblages. Both NMDS ordination and hierarchical clustering were derived from the same dissimilarity matrix but capture different aspects of community structure, which may result in slight variation in the placement of transitional samples.

3.4. Effects of L. camara on Soil Nutrients, Species Composition, and Diversity

There were significant differences in soil nutrients between invaded and uninvaded areas (

Table 4. Soil physicochemical properties of invaded and uninvaded plots in Park and Communal areas. Values represent mean ± standard error (n = 36; Park = 16, Communal = 20). Statistical comparisons were conducted within each land-use category to test differences between invaded and uninvaded plots using one-way ANOVA (or a non-parametric equivalent where assumptions were not met). Different lowercase letters within the same row and land-use category indicate significant differences at p < 0.05 based on Tukey’s HSD post hoc test. Superscripts do not indicate comparisons between land-use categories.

Variable	Communal Area		Park Area		F-Value	p-Value
	Invaded	Uninvaded	Invaded	Uninvaded
pH	6.8 ± 0.5 b	7.6 ± 0.6 a	6.1 ± 0.3 b	7.0 ± 0.8 a	6.3	0.001
NO3−-N (mg/kg)	2.3 ± 1.5 a	1.1 ± 0.7 a	6.7 ± 6.2 a	7.5 ± 13.0 a	1.4	0.30
P (mg/kg)	3.8 ± 5.6 a	0.2 ± 0.4 b	10.7 ± 8.8 a	10.500 ± 0.001 a	3.5	0.03
K (mg/kg)	4040 ± 3453 a	2050 ± 1629 a	1660 ± 523 a	1087 ± 554 a	1.8	0.20
OC (%)	1.8 ± 1.1 a	0.95 ± 0.93 b	1.9 ± 0.84 a	2.9 ± 0.6 a	2.6	0.04

). The most significantly affected variable was pH, which was lower in invaded sites across both areas. Phosphorus showed trends toward higher values in invaded sites, especially in Communal areas where uninvaded levels were notably low, and organic carbon was higher in invaded Communal sites but lower in Park sites. For example, organic carbon was approximately two times higher in the invaded sites than in the uninvaded sites in the Communal Area. Phosphorus was up to 19 times higher in the invaded sites than in the uninvaded sites in the Communal Area (Table 4).

The Monte Carlo permutation test revealed that the CCA model was statistically significant (p < 0.05), indicating that the soil variables had a significant effect on the composition of woody species in both locations. The CCA revealed that the first two axes explained 54% of the total variation between species composition and soil nutrient variables (

Table 5. Summary statistics of Canonical Correspondence Analysis (CCA) showing the relationships between nutrient variables and woody species composition in the Communal and Park sites, based on the first two axes (CCA1 and CCA2). Values in bold indicate soil variables that are statistically significant (p < 0.05) based on permutation tests.

CCA Properties	CCA1	CCA2
Eigen value	0.21	0.24
Variance explained	0.21	0.24
Cumulative variance	0.21	0.54
Soil variables	F-value	p-value
pH	1.52	0.03
NO3−	1.65	0.11
P	1.53	0.01
K	1.02	0.37
OC	1.74	0.03

, Figure 4). The CCA ordination plot revealed that pH and OC were the most important soil variables influencing the distribution of native woody species at the two sites (Table 5, Figure 4). pH and organic carbon (OC) showed negative associations with the invaded Communal sites (Figure 4).

Cluster analysis and NMDS ordination revealed four clusters representing gradients in woody species composition, with considerable overlap among invasion and land-use categories (Figure 2 and Figure 3). Cluster A, largely comprising communal invaded sites, was associated with elevated NO₃ concentrations and disturbance-tolerant taxa, including Lantana camara, Berchemia africana, and Dichrostachys cinerea. Cluster C consisted predominantly of uninvaded sites from both land-use types but included a small subset of communal invaded plots that grouped with uninvaded sites due to high compositional similarity to native savanna states, characterised by lower pH and higher organic carbon. Cluster D represented a mixed/transitional grouping of invaded and uninvaded plots, associated with intermediate assemblages and taxa such as Afzelia quanzensis. In contrast, Cluster B was primarily associated with invaded park sites and species such as Colophospermum mopane and Kigelia africana. Overall, these patterns are supported by ANOSIM (Global R = 0.62, p = 0.001), indicating that while invasion alters woody assemblages, its effects are mediated by land-use history, with some communal sites retaining native-like compositional characteristics despite the presence of L. camara. Species positions in the figures are labelled using abbreviations provided in Table S1 (Supplementary Materials).

Of the ten vegetation attributes considered, only three were significantly associated with soil nutrients (p < 0.05;

Table 6. Spearman rank correlation coefficients (ρ) between soil physicochemical properties and vegetation structural attributes in invaded plots (n = 36). Values represent Spearman’s correlation coefficients. p < 0.05; p < 0.001.

Species Diversity/Vegetation Attributes	pH	NO3−	P	K	OC
Shannon index	0.20	0.15	−0.06	−0.09	0.14
Species richness	0.21	0.28	−0.09	−0.09	0.18
Native woody species density/ha	−0.19	−0.10	−0.10	−0.13	−0.19
Native woody species height (m)	0.13	0.21	−0.04	−0.21	0.12
Native woody species diameter (m)	−0.01	0.09	−0.07	−0.18	−0.04
Native woody species cover (%)	0.28	0.10	−0.19	−0.01	0.17
L. camara density/ha	−0.20	−0.12	−0.08	−0.11	−0.19
L. camara height (m)	−0.27	−0.11	0.15	0.04 *	−0.17
L. camara canopy (%)	−0.17	−0.06	−0.02	−0.03 *	−0.12
L. camara diameter (m)	−0.03	0.02	−0.06	0.04	0.24

Note: Asterisks (*) indicate statistically significant correlations (p < 0.05).

). Native woody species DBH significantly lower with increasing pH and OC, whereas L. camara height increased with increasing K. Conversely, K negatively affected the canopy cover % of L. camara (Table 6).

Interaction effects were detected for some variables, indicating context-dependent responses in vegetation structure and soil properties (

Table 7. Two-way ANOVA results testing the effects of invasion status (Invaded vs. Uninvaded), site (Communal vs. Park), and their interaction (Invasion × Site) on selected native woody vegetation and soil variables. Values represent F-statistics and associated p-values. Asterisks indicate levels of statistical significance (p < 0.05; p < 0.01; p < 0.001).

Variable	Invasion (F, p)	Site (F, p)	Invasion × Site (F, p)
Canopy cover (%)	1226.6, p < 0.001	0.20, p = 0.652	0.22, p < 0.001
Species richness	445.5, p < 0.001	116.6, p < 0.001	0.04, p = 0.85
Stem density (ha−1)	203.2, p < 0.001	16.9, p < 0.001	0.51, p = 0.48
Soil pH	43.5, p < 0.001	15.3, p < 0.001	0.01, p = 0.93
Soil organic carbon (%)	0.41, p = 0.52	9.11, p = 0.005 **	4.23, p = 0.048 *
Soil available p (mg kg−1)	4.36, p = 0.045 *	39.1, p < 0.001	0.03, p = 0.86

Note: Asterisks indicate levels of statistical significance: * p < 0.05, ** p < 0.01.

).

4. Discussion

4.1. Effects of L. camara Invasion on Vegetation Structure and Diversity

Invaded sites exhibited significantly lower native woody species richness, Shannon–Wiener diversity, stem density, height, diameter at breast height, and canopy cover % compared to uninvaded sites in both land-use categories (Table 1). These patterns support H1, demonstrating that L. camara presence is associated with shifts in native woody structure and reduced diversity, with evidence of stronger suppression in communal areas due to higher anthropogenic disturbance. The significant interactions for canopy cover and soil organic carbon (Table 7) indicate that land-use context modulates some invasion associations [6,10,25,46]. These results are consistent with previous studies reporting reduced native plant diversity and structural attributes in L. camara–invaded sites [11,47,48], including Ruwanza [11], which documented similar declines in South African invaded sites.

In this study, the significant reductions in richness, diversity, and structural variables support the association between L. camara presence and suppression of native woody species, potentially through competition for resources or allelopathic effects [6,48]. However, the lower values observed in uninvaded Communal sites relative to uninvaded Park sites suggest that land-use disturbance may also contribute to lower vegetation attributes independently of invasion [18,25].

These findings have broader ecological implications for savanna woodlands in transfrontier conservation areas: prolonged suppression of woody structure and diversity in disturbed communal zones could reduce ecosystem resilience, limit recruitment of key canopy species, and alter habitat quality for wildlife dependent on diverse woody cover.

4.2. Effects of Invasion on Species Composition

Woody species composition differed significantly between invaded and uninvaded sites, with stronger separation in the Communal Area (ANOSIM R = 0.8, p = 0.001; Table 3). Uninvaded sites had higher species richness (48 species in uninvaded Park plots vs. 25 in invaded Communal plots; Table 2) and greater frequency of occurrence for many native species such as Afzelia quanzensis and Colophospermum mopane (>70% in uninvaded sites). Only Combretum imberbe maintained high frequency (>50%) in invaded sites.

NMDS ordination and cluster analysis revealed clear groupings by invasion status and land use (Figure 2 and Figure 3), consistent with findings that L. camara presence is associated with altered community composition [26]. The more pronounced differences in Communal areas likely reflect higher invasion intensity there. These shifts support H1 by indicating competitive exclusion or lower recruitment of less tolerant natives in invaded sites, although pre-existing site conditions cannot be ruled out as a contributing factor [11,22,49].

From a broader perspective, such compositional shifts may reduce functional diversity (e.g., loss of large-seeded canopy species important for seed dispersal and carbon storage), with cascading effects on ecosystem services like fodder availability and ecotourism value in the KAZA TFCA.

4.3. Effects of L. camara on Soil Nutrients

The soil patterns observed in this study indicate that the influence of L. camara on edaphic conditions is neither uniform nor universally transformative but instead mediated by land-use intensity and local disturbance regimes [8,11]. Rather than producing consistent nutrient enrichment or depletion, invasion appears to interact with pre-existing soil characteristics and management histories, resulting in heterogeneous outcomes across the landscape [18,20,50]. This variability suggests that L. camara functions less as a sole driver of soil change and more as a reinforcing agent within already altered or disturbed environments.

The tendency toward lower soil pH in invaded communal areas points to potential acidification processes associated with dense shrub thickets, possibly linked to increased litter accumulation, altered microbial activity, or shifts in decomposition pathways [16,19]. Such mechanisms are widely recognised in invasion ecology, where changes in litter quality and root exudates can influence soil chemical balances [16]. However, the absence of comparable trends across all nutrients and land-use types indicates that these processes are not universally expressed and may depend on disturbance frequency, grazing pressure, and vegetation turnover [8]. In this context, soil responses appear to reflect an interaction between invasion and anthropogenic land use rather than a direct and uniform ecological transformation [11,18].

The multivariate analysis (Table 5) further supports this interpretation by showing that soil variables contribute to vegetation structuring at the community level, yet exert limited influence on individual species attributes. This pattern implies that while soil conditions help shape overall compositional gradients, they do not singularly determine invasion success or native species performance. Instead, invasion dynamics likely emerge from a combination of above-ground competition, disturbance regimes, and microhabitat variability, with soil nutrients acting as one component within a broader ecological network [8]. The selective nature of these associations supports the importance of considering both biotic and abiotic drivers when interpreting invasion impacts.

Taken together, the findings provide partial support for the hypothesis that L. camara alters soil properties, but they also indicate that such alterations are context dependent and often subtle [20]. The relatively limited number of strong nutrient shifts suggests that invasion may frequently occur in sites where soil conditions are already conducive to establishment, rather than consistently inducing major chemical transformations after colonisation [20]. This differentiation is important, as it shifts the narrative from invasion as a purely causal agent to invasion as a process embedded within existing environmental gradients and land-use legacies.

From a management perspective within the KAZA TFCA, these insights emphasise the need for spatially differentiated and adaptive strategies. Communal landscapes where disturbance pressures and invasion intensity are higher may require more proactive monitoring, community-based control initiatives, and post-clearing ecological restoration to prevent further woody species decline and maintain soil functionality. In contrast, protected areas appear to exhibit greater ecological buffering capacity, suggesting that natural recovery processes may be more effective once invasive pressure is low. Overall, the soil–invasion relationship observed here reinforces the importance of integrating land-use heterogeneity, disturbance history, and local ecological context into conservation planning, rather than assuming uniform below-ground impacts of invasive shrub encroachment across transfrontier conservation systems.

4.4. Limitations and Interpretation of Cause–Effect Relationships

The significant associations between L. camara presence and lower native woody species richness, diversity, and structural attributes are clear (Table 1 and Table 3), but it remains uncertain whether presence directly caused these changes or whether pre-existing site conditions (e.g., higher disturbance, grazing, or soil properties in communal areas) facilitated L. camara establishment while independently affecting native vegetation and soils. The observed associations between the presence of L. camara and reduced native woody species density, height, and canopy cover should be interpreted with caution because the cross-sectional study design does not allow causal inference.

Although reductions in native woody structure are consistent with suppression by L. camara, pre-existing canopy openness or disturbance may have facilitated invasion, rather than being solely a consequence of it. As Sundaram and Hiremath [15] demonstrated across multiple forest types, canopy gaps and anthropogenic disturbance are key drivers of L. camara invasibility. The significant interaction between invasion status and site (Table 7) further indicates that land-use context modulates these relationships, but the cross-sectional design prevents full disentangling of cause and effect. Consistent patterns across the NMDS ordination (Figure 2), dendrogram (Figure 3), and CCA biplot (Figure 4) highlight the association between invasion and altered woody composition, warranting further investigation into causal mechanisms. Future longitudinal or experimental studies are needed to clarify these feedbacks.

Canopy gaps and disturbed conditions are widely reported as important drivers of L. camara invasibility in heterogeneous landscapes [15]. The significant interactions for canopy cover and soil organic carbon (Table 7) indicate that land-use context modulates some invasion associations, further suggesting that the strength and nature of these relationships depend on site-specific conditions [15]. The stronger separation between invaded and uninvaded plots within sites, reflected in pairwise R values ranging from 0.4 to 0.9, compared with differences between land-use types, suggests that invasion status plays a major role in structuring vegetation patterns. However, site-specific disturbance likely influences the magnitude of these associations, which is consistent with the higher anthropogenic pressure observed in communal lands. The unequal sampling (more invaded than uninvaded plots) reflects real landscape patterns but may have low statistical power for some comparisons. Future studies employing paired invaded–uninvaded designs, longitudinal monitoring, or manipulative experiments would help better disentangle cause and effect.

5. Conclusions

The results indicate that the presence of L. camara is associated with reduced native woody vegetation structure and diversity in both Communal and Park areas of the KAZA TFCA (Table 1; Figure 2). Invaded plots showed lower canopy cover (6.6–9.0% vs. 72.5–81.0% in uninvaded plots) and stem density (228–479 stems/ha vs. 1308–1561 stems/ha), together with shifts in species composition (Clusters A and B; Figure 2 and Figure 3). Uninvaded plots (Cluster C) supported more diverse assemblages dominated by native species such as Combretum imberbe, Brachystegia spiciformis, and Kigelia africana. Soil properties showed variable patterns, with trends toward lower pH and higher organic carbon and phosphorus in invaded communal sites (Figure 4), though these were inconsistent and likely modulated by land-use context.

These findings contribute to understanding context-specific effects of L. camara in transfrontier conservation landscapes, where communal areas appear more vulnerable to invasion-associated changes, possibly due to higher anthropogenic disturbances. However, the cross-sectional design limits inference of causation, and pre-existing conditions may have influenced invasion success [15]. Longitudinal or experimental studies are needed to clarify these relationships.

The results highlight the need for targeted management in disturbed communal zones to protect native woody species and ecosystem services such as fodder, medicinal plants, and habitat. Monitoring and early intervention in protected areas could limit edge spread. Adaptive policies tailored to land-use systems would support conservation goals in the KAZA TFCA.