Effects of site conditions on regeneration of African mahogany (Khaya anthotheca) in a semi‐deciduous East African tropical rainforest

Restoration of high‐value timber trees in logged tropical forests is indispensable as a conservation strategy and for the enhancement of ecosystem services. Khaya anthotheca is a mahogany tree species of major restoration importance in Uganda. However, the factors affecting the natural regeneration success of this species are poorly understood. We determined the effect of selected site conditions on the abundance of K. anthotheca regeneration in the Budongo Forest Reserve. Two hundred and four 1,000 m2 inventory plots, each nested with four 4 m2 subplots, were systematically established in logged and unlogged sites within the study area. All trees greater than or equal to 10 cm DBH and natural regeneration (<10 cm DBH) were identified and recorded in the plots and subplots, respectively. We characterized plots by the density of conspecifics, light availability, ground cover of undergrowth, litter depth, soil pH, and basal area. Regression models were used to evaluate the variables that influenced regeneration abundance and occurrence. The density of adult conspecifics and soil pH were important. Regeneration abundance and occurrence increased with increasing density of adult conspecifics and decreased with increasing soil pH. Our results highlight that in addition to seed availability, soil pH was important for successful regeneration. This is one of the first K. anthotheca‐specific studies to evaluate more than a few environmental factors influencing the density of its naturally established seedlings in tropical African forests. The abundance and distribution of adult conspecifics in the canopy and soil pH should be considered in K. anthotheca restoration plans.


Introduction
Understanding natural regeneration and its controlling factors is of growing importance, with increasing emphasis placed on it for the restoration of degraded lands (Chazdon & Guariguata 2016;Uriarte & Chazdon 2016;Crouzeilles et al. 2020).Most of the world's remaining tropical forests are degraded due to unsustainable human activities, including poorly conceived logging operations, with a strong selective character (Zimmerman & Kormos 2012;Lewis et al. 2015).
Although degraded, they retain a high portion of their original biodiversity (Gibson et al. 2011;Putz et al. 2012;Edwards et al. 2014) and provide both timber and non-timber forest products (Harrison & Swinfield 2015), thereby playing a huge social and economic role (Akindele & Onyekwelu 2011).They also provide ecosystem services, including carbon sequestration (Laurance & Edwards 2014;Harrison & Swinfield 2015), helping to mitigate climate change (Lewis et al. 2019).Nonetheless, they remain exceptionally vulnerable to a suite of threats, including conversion to other land uses (Petrokofsky et al. 2015;Cerullo & Edwards 2019), particularly those perceived as more lucrative (Fredericksen & Putz 2003;Malhi et al. 2013).
Consequently, the restoration of logged tropical forests is widely advocated as an essential component of the global restoration agenda (Cerullo & Edwards 2019;Ngo Bieng et al. 2021).It is increasingly suggested that the restoration of valuable timber species to degraded forests in which they were formerly abundant, has the potential to not only enhance the ecosystem services they provide but likewise to reverse their further degradation, by increasing their economic value (Harrison & Swinfield 2015;Cerullo & Edwards 2019).In addition, restoration will help preclude the extinction of many threatened high-value timber species (IUCN 2022).
In Uganda, the status of tropical high forests has drastically declined, with an estimate of over 30% classed as degraded (Obua et al. 2010).The State of Uganda's Forestry Report of 2016 indicates that Uganda's natural forests have lost many of their high-value timber trees and may not contain any in the near future (MWE 2016a).Loss of important timber tree species is of major concern as it is associated with a myriad of undesirable outcomes, such as conversion to industrial plantations (Fearnside 1997;Gaveau et al. 2016).The government of Uganda has prioritized forest restoration, pledging in the Bonn Challenge to restore 2.5 million hectares of land to forests (MWE 2016a).The primary target is to restore Uganda's forest cover from the currently estimated less than 10 to 24% of the total land area (MWE 2016b).Boosting the economic value of degraded forests as competitive land uses through the restoration of commercially important species is well recognized in the Forest Landscape Restoration Opportunity Assessment for Uganda (MWE 2016b).
Of the valuable timber trees in Uganda, restoration emphasis has been placed on African mahoganies, including Khaya anthotheca (MWE 2016b).As a highly valued tropical timber tree worldwide, K. anthotheca has been selectively harvested, drastically shrinking its natural stocks (MWE 2016a).With diminishing populations, an important source of timber and revenue will be lost (Held et al. 2010).While K. anthotheca has been grown in plantations to maintain supplies, mahogany plantations worldwide have been bedeviled by the shoot borer (Newton et al. 1993;Opuni-Frimpong et al. 2008a).In addition to plantations, it is generally accepted that restoration in natural forests is indispensable for mahogany conservation and sustainable utilization (Chaikaew et al. 2020).However, promoting the natural regeneration of this species remains a major challenge worldwide (Ward et al. 2008;Owusu et al. 2014).Auto-ecological research is thus necessary to strengthen capabilities to enhance the natural regeneration of K. anthotheca (Ward et al. 2008;Opuni-Frimpong et al. 2008b).
Natural regeneration success is contingent upon a diverse set of biotic and abiotic factors (Puhlick et al. 2012;Bose et al. 2016).These set of environmental factors that are decisive for K. anthotheca regeneration remain poorly understood.Previous studies have mainly investigated the effect of light availability, revealing that seedlings grow better in gaps than in the forest understory (Makana & Thomas 2005;Opuni-Frimpong et al. 2008b).However, poor regeneration in gaps despite high levels of light availability has also been reported (Mwima et al. 2001).With these contradicting results, the effect of light availability remains unclear.The effect of litter has also been studied, but in only a few studies (Makana & Thomas 2005).
To the best of our knowledge, there have been no speciesspecific studies on other factors that may be imperative for K. anthotheca natural regeneration.Previous studies generally suggest that the natural regeneration of K. anthotheca may depend on a host of specific factors, notably light and seed source availability, competition from overstory, forest undergrowth vegetation, litter depth, and soil pH (Mwima et al. 2001;Makana & Thomas 2004, 2005).These factors need to be evaluated to understand their specific effect to inform restorationists on K. anthotheca regeneration in tropical high forests.
In this study, we aimed to analyze the effects of the aforementioned environmental factors on the abundance of K. anthotheca regeneration in Budongo Forest, Uganda, to contribute to better our understanding of the species regeneration ecology.The study complements site-specific knowledge on K. anthotheca regeneration provided by Bahati (2005) to support efforts to restore this highly valued tree species.We hypothesized that the natural regeneration of K. anthotheca is affected specifically by light availability, the abundance of adult conspecifics, litter depth, density of forest undergrowth vegetation, basal area, and soil pH.

Study Area
We conducted our study in Budongo Forest Reserve (1 35 0 to 1 55 0 N and 31 18 0 to 31 42 0 E; Fig. 1), 1,100 m asl, northwestern Uganda (Plumptre 1996;GoU 2011).The reserve covers an area of about 793 km 2 , making it the largest forest reserve in Uganda (Howard 1991).Budongo is a moist semi-deciduous rainforest (Eggeling 1947), with an average annual temperature of 29 C, and receives approximately 1,410 mm of precipitation annually (GoU 2011).Floristically, it is similar to the rainforests of West Africa and the great Ituri Forest of the Democratic Republic of Congo (Paterson 1991;Sheil 1996).It harbors over 260 species of trees (Sheil 1996), of which Cynometra alexandri is the most dominant species, constituting over 30% of the mature trees (GoU 2011).Other species contribute less than 10% each to the standing crop (GoU 2011).The soils are ferralitic and mostly nutrient-poor (Paterson 1991), characterized by a high pH, and Potassium is considered a limiting nutrient to forest productivity (Manu et al. 2022).Budongo has a rich management history that did affect its structure (Plumptre 1996).It includes wild rubber tapping from Funtumia africana trees in the 1940s, and the treatment of trees considered of low market value with arboricides in the 1950s and 1960s to create gaps to encourage regeneration of valuable timber species (Plumptre & Reynolds 1994).Although timber harvesting started as early as 1910, commercial logging following set management plans commenced in 1945, focusing specifically on mahoganies (Plumptre & Reynolds 1994).Consequently, most of the forest compartments have undergone selective logging and arboricide treatment, except the area that was set aside from the onset of management as a nature reserve (Mwavu et al. 2008).The Budongo Forest, as it stands, consists of selectively logged compartments and unlogged forest areas.This study was carried out in the unlogged forest that occurs in compartment N15 and a selectively logged forest that occurs in compartment N3.Details of these two study forests are well described in Plumptre (1996) andGoU (2011).

Study Species
Khaya anthotheca (Welw.)C.DC. K. anthotheca is a tall timber tree of tropical rainforests belonging to the family Meliaceae (Hawthorne 1998).It is a long-lived tree that grows in medium to low-altitude areas (Hawthorne 1998), growing to a height of about 60 m, with a 300 cm DBH clean bole when mature (Bahati 2005).Individuals can have crowns as large as 20 m in diameter and are characterized by buttresses (Hawthorne 1995).The species is classed as a non-pioneer light demander (Makana & Thomas 2005;Kirika et al. 2010).The timber is of a high grade and is widely sought after in domestic, regional, and international markets (Ofori et al. 2007).Globally K. anthotheca is assessed as "vulnerable" on the IUCN Red List of Threatened Species (Hawthorne 1998;Doucet et al. 2016).After a longstanding debate, the species is now subject to appendix II of CITES.In Uganda, although production statistics are not available, K. anthotheca timber is highly valued in the local markets for its durability and esthetic appearance (Maroyi 2008).The wood is used for veneer, both internal and external joinery, and decorative furniture (Maroyi 2008).Other than timber, this species is of high ecological importance as the bark is eaten by chimpanzees (Newton-Fisher 1999).Intensive exploitation, combined with poor regeneration, has led K. anthotheca to become an endangered tree species in Uganda (MWE 2016a).Conservation emphasis has largely focused on combating illegal and destructive harvesting that is a constant threat to the remaining individuals (MWE 2016a).

Sampling Design
Two forest compartments (N3 and N15) were sampled.Compartment N3 (hereafter "logged forest") was logged in 1947-1952, removing 80 m 3 /ha (Plumptre 1996), whereas N15 (hereafter "unlogged forest") was never formally logged (Plumptre 1996).Although the unlogged forest has experienced some illegal harvesting (Gombya-Ssembajwe et al. 2007), it remains relatively less disturbed.Circular inventory plots of an area of 1,000 m 2 were established at 100-m intervals along five 2.5 km long transects, which had been established in a systematic random manner in each of the compartments for phenology monitoring (Babweteera et al. 2018).In total, we established 204 plots.Each plot was nested with four (4 m 2 ) subplots for regeneration assessment (Fig. 2).In total, we established 816 subplots (420 in logged forest and 396 in unlogged forest).

Assessment of K. anthotheca Regeneration
We searched for K. anthotheca regeneration (<10 cm DBH) within subplots and recorded their height class.Heights were measured to the nearest cm using a pole ruler.Three height classes were used (5-49, 50-99, ≥100 cm).Individuals were also checked for any evidence of herbivory, and if seen, it was recorded.For further analysis, individual regeneration counts from the four subplots were pooled together to form one plot count.Plots with at least one K. anthotheca regeneration individual were categorized as colonized and those with none as uncolonized.We also recorded by species and height class the regeneration of other trees and computed seedling species richness and diversity per plot using the R package "vegan" (Oksanen et al. 2016).

Assessment of Basal Area and Density of Conspecifics
All trees with DBH greater than or equal to 10.0 cm within plots were measured using a diameter tape and identified to species level.Species identification was done in the field by botanists, with reference to Katende et al. (1995).Those that could not be identified in the field were sampled and identified by experts at Budongo Conservation Field Station (BCFS).Species names were cross-referenced with the updated tree list available at BCFS.For further analysis, we computed tree species richness and diversity per plot using the R package "vegan" (Oksanen et al. 2016).Trees were measured following the standard forest mensuration protocols of Malimbwi (1997), unpublished compendium.For instance, individuals with buttresses or malformations were measured or estimated just above the point of columnar formation.Multi-stem trees with forking below 1.3 m were considered independent and measured separately.For further analysis, the basal area of overstory as a proxy for competition, including all tree species, was calculated for each plot and converted into a hectare basis.For each plot, we computed the densities of reproductive conspecifics (≥40 cm DBH), nonreproductive conspecifics (≥10 DBH < 40 cm), and all conspecifics (total of reproductive and none reproductive individuals).Fruit production in K. anthotheca is primarily size-dependent and occurs when individuals attain DBH greater than or equal to 40 cm (Plumptre 1995).The density of trees greater than or equal to 10 cm DBH, all species combined, was also computed for each plot.

Assessment of Light Availability
We quantified light availability using hemispherical photography.At the center of each subplot, 1.3 m above the ground, we took one photograph using an automated fisheye lens camera (Solariscope, Behling SOL300).By automatically analyzing hemispherical photographs according to seven thresholds, the Solariscope provides seven measurements of light availability, notably, indirect site factor, leaf area index, direct site factor, gap fraction, Openness (Opn), total site factor (TSF), and ellipsoidal leaf area distribution.For further analysis, the best photograph among the seven thresholds, the one that showed a clear distinction between sky and non-sky, was visually chosen.

Assessment of Soil pH
We collected soil samples from the upper 20 cm of the mineral soil using a 7 cm diameter extracting auger.One sample was taken from each of the four subplots and composited.The composite samples were processed and transported for analysis in the soil laboratory of the Department of Silviculture and Forest Ecology of Temperate Zones, Göttingen University.Soil pH was determined in distilled water using an electric pH meter (WTM SenTix 41 pH Electrode).

Assessment of Undergrowth Vegetation and Litter Depth
At the subplot level, we determined the cover percentage of species other than tree regeneration.We estimated visually the plot area covered as 0% (0 of the 4 m 2 covered), 25% (1 of the 4 m 2 covered), 50% (2 of the 4 m 2 covered), 75% (3 of the 4 m 2 covered), and 100% (4 of the 4 m 2 covered).For further analysis, the mean percentages were computed to represent the plot.Litter depth in each subplot was determined at four points by inserting a hand ruler through the litter until the mineral soil was removed.The mean litter depth per subplot was calculated using the four readings.Litter depth at plot level was calculated as the mean of the four subplots.

Data Analysis
Regeneration Distribution.Descriptive statistics were employed to summarize data on regeneration occurrence.Particularly, frequencies and sums were computed and presented descriptively.To test whether the distribution of regeneration in the different height classes differed significantly between logged and unlogged forests, differences between observed and expected counts were tested using chi-squared tests (Johnson & Bhattacharyya 2010).Similar tests were done to compare the distribution of colonized and uncolonized plots between the study forests.A nonparametric Wilcoxon test was performed to determine if regeneration abundance in the logged and unlogged forests were statistically different.
Site Characteristics between Khaya anthotheca Colonized and Uncolonized Plots.In the first step, we examined the distribution of all continuous variables using a Shapiro-Wilk normality test (Shapiro & Wilk 1965;Razali & Wah 2011) and homoscedasticity using the Levene test (Johnson & Bhattacharyya 2010).For variables that conformed to the normality and equal variances assumptions, parametric two-sample t-tests were carried out to assess differences in means.For those who did not conform, a nonparametric Wilcoxon test was done to compare differences in group medians.A significance level of p < 0.05 was applied.
Influence of Site Conditions on K. anthotheca Regeneration.A regression model was used to analyze the influence of site factors on the abundance of mahogany regeneration.The response variable was seedling count per plot.Determinants were light availability, litter depth, cover of undergrowth vegetation, basal area, soil pH, and density of conspecifics.Before modeling, multicollinearity, the existence of a correlation between covariates (Imdadullah et al. 2016) was checked by inspecting the correlation matrix of environmental factors (Grewal et al. 2004).We set 0.6 as the threshold, above this value, multicollinearity causes problems in modeling (Grewal et al. 2004).The density of nonreproductive conspecifics (<40 cm) and the density of total conspecifics were highly correlated (r = 0.7).Consequently, the former was not included in further analysis.All measures of light availability were strongly correlated with each other.Consequently, only TSF was used as a proxy for light availability.
To select the correct function for modeling regeneration count per plot as a function of the aforementioned site factors, we tested six generalized regression models using different distributions for count data.Because it is not uncommon that seedling count responses have count distributions with excess zeros (Terwei et al. 2013), models tested included widely used two-part Hurdle models and zero-inflated mixture models (Fortin & DeBlois 2007;Zeileis et al. 2008;Zhang et al. 2012).The models tested included Poisson, zeroinflated Poisson, zero-inflated negative binomial (NB), Hurdle-Poisson, Hurdle NB, and NB regression models.For all six models, the explanatory variables tested were the same, and to ensure consistency, the variables were entered in the same order.The models were fitted using the R package "glmmTMB" (Brooks et al. 2017) and compared through the Bayesian information criterion (BIC).Based on this criterion, the NB model was selected because it performed best as it had the lowest BIC (Table S1).In the subsequent step, we calculated the confidence intervals and validated the model using R packages "ggeffects" (Lüdecke 2018) and "DHARMa" (Hartig 2022), respectively.The stability of the estimated coefficients and their significance were also tested.The NB model adjusted for both overdispersion and zero inflation as both tests were not significant (Fig. S1).The model residuals were uniformly distributed (Fig. S2), indicating no violation of the homoscedasticity assumption.The estimates of the coefficient produced by the NB were meaningful, robust, and in harmony with the data.All statistical analysis was done in R, a software environment for statistical computing and graphics (R Core Team 2019).

Species Diversity and Structural Attributes of the Study Sites
Diversity indices (Simpson's, Shannon's, and exponential Shannon's) showed statistically higher seedling species diversity per plot in the logged than unlogged forest ( p < 0.05; Table 1).However, seedling species richness (Chao 1 estimator) per plot of the logged (11.6 AE 0.26) and unlogged forest (12.09AE 0.29) were not significantly different (t = 1.22; p > 0.05).The density of regeneration, all species combined, was significantly different among the study sites (t = 5.42; p < 0.001), being higher in the unlogged (4.91 ind./m 2 ) than in logged forest (3.12 ind./m 2 ).Tree species richness (Chao 1 estimator) per plot of the logged site (18.55AE 0.35) was significantly higher than in the unlogged site (15.9AE 0.34).Tree species diversity indices (Simpson's, Shannon's, and exponential Shannon's) per plot were higher in the logged than unlogged forest ( p < 0.05; Table 1).Canopy Opn was 1% higher in the logged than unlogged forest, although the difference was not statistically significant ( p > 0.05; Table 1).Soil pH, tree density, and basal area did not significantly differ between the study sites ( p > 0.05; Table 1).

Occurrence and Abundance of Khaya anthotheca Regeneration
Across both study sites, K. anthotheca seedlings occurred in 35% of the sampled plots.A high number of these plots (53%) were colonized by only one individual (Fig. S3).Another 29% and 10% were characterized by two and between 5 and 10 individuals, respectively (Fig. S3).Only 8% of the colonized plots had more than 10 individuals.
Of the plots with regeneration, 56% and 44% were located in the logged and unlogged forests, respectively.There was no significant difference in the number of colonized plots between the logged and unlogged forests (χ 2 = 0.512, p > 0.05).The abundance of K. anthotheca regeneration, regardless of the height class, ranged between 625 ind./ha (0.0625 ind./m 2 ) and 23,125 ind./ha (2.3125 ind./m 2 ), with a mean abundance of 2000 ind./ha (0.2 ind./m 2 ).The regeneration abundance decreased with increasing size, from 6,000 ind./ha in the less than 50 cm class to 2,500 ind./ha in the greater than 100 cm height class.The distribution of K. anthotheca regeneration individuals in the different height classes differed significantly between the unlogged and logged forests (χ 2 = 14.00, p < 0.05; Fig. 3).There were significantly more regeneration individuals in the height class less than 50 cm in the logged than unlogged forest (χ 2 = 5.4772, p < 0.05).A significantly higher number of regeneration individuals in the height category 50-100 cm was found in the unlogged forest (χ 2 = 3.9609, p < 0.05).Regeneration in the height class greater than 100 cm was absent in the logged forest (Fig. 3).There was no significant difference in the abundance of regeneration between the two forests (W = 723.5,p > 0.05).None of the encountered K. anthotheca seedlings had any sign of browsing by small mammals nor damage by insects.

Site Characteristics between K. anthotheca Colonized and Uncolonized Plots
Overall, the studied site conditions in plots with and without K. anthotheca regeneration were similar (Table 2).Only the density of reproductive conspecifics (≥40 cm DBH) differed significantly (W = 3,511, p < 0.05).Colonized plots were characterized by a higher number of mother trees than uncolonized plots, with a mean density of 5.4 ind./ha.Basal area, the density of nonreproductive conspecifics, the density of trees, cover of undergrowth vegetation, the density of other seedlings, and the total density of K. anthotheca conspecifics did not differ significantly ( p > 0.05) between plots with and without regeneration (Table 2).Diversity indices (Simpson's, Shannon's, and exponential Shannon's) revealed no significant differences in both seedling and tree species diversity between colonized and uncolonized plots ( p > 0.05; Table 2).Seedling and tree species richness (Chao 1 estimator) were similar in plots with and without K. anthotheca regeneration (Table 2).Concerning soils, pH did not vary significantly between the sites ( p > 0.05).None of the seven measures of light availability varied significantly between colonized and uncolonized plots ( p > 0.05; Table S2).

Influence of Site Conditions on K. anthotheca Regeneration Abundance and Occurrence
Among the studied factors, the regression analysis revealed that regeneration abundance depended on two factors: density of adult conspecifics and soil pH.The NB model identified that while soil pH had a significant negative effect on regeneration abundance, the density of adult conspecifics had a significant positive effect (Table S3).The abundance of regeneration  increased with increasing density of adult conspecifics and decreased with increasing soil pH (Fig. 4A & 4B).The probability of regeneration occurrence increased with increasing density of reproductive conspecifics but decreased with increasing soil pH (Fig. 5A & 5B).

Discussion
Understanding the key factors affecting the natural regeneration of Khaya anthotheca, an important tropical timber species, is decisive for successful restoration in selectively logged tropical forests.This study highlights the importance of adult conspecific abundance and soil pH in influencing the regeneration of K. anthotheca.

Effect of Adult Conspecifics
The abundance of adult conspecifics in the canopy positively influenced the abundance of K. anthotheca regeneration.This finding was expected for one core reason.K. anthotheca seeds have no dormancy, precluding the formation of a soil seed bank (Makana & Thomas 2006).Regeneration of this species is thus contingent upon seed rain (Mwima et al. 2001).Moreover, most of the seeds are dispersed within the vicinity of seed trees (Makana & Thomas 2004).Therefore, the high abundance of adults in the overstory leads to higher densities of seed deposition and recruited seedlings.Overall, our results confirm the earlier hypothesis that the abundance of K. anthotheca regeneration increases with an increase in the density of adult conspecifics   S3).All other covariables are fixed to their means.Shaded areas represent 95% CI.
Controls of mahogany regeneration (Plumptre 1995).Our results, however, contradict the negative density effect of adult conspecifics on the establishment, survival, and performance of seedlings (Janzen 1970;Connell 1971).Negative density dependence predicts reduced conspecific regeneration abundance in areas near seed trees (Ghazoul & Sheil 2010;Comita et al. 2014).The effects of this mechanism can be species, scale, and site-dependent (Chapman & Chapman 1996;Cintra 1997;Matthesius et al. 2011) and have also been found to be weaker than environmental conditions (Zhu et al. 2013).A further study could evaluate K. anthotheca regeneration in the context of the Janzen-Connell hypothesis.
Our finding suggests that a high proportion of K. anthotheca potential seed trees in tropical forests is a prerequisite to achieving high natural regeneration densities.In this regard, mapping both the abundance and spatial distribution of potential seed trees should be given major consideration in restoration plans.Phenological observations have shown that high seed production in K. anthotheca takes place when an individual tree has attained DBH greater than or equal to 80 cm (Plumptre 1995).In Uganda's natural forests, most of the large reproductive trees have been severely diminished (MWE 2016a; Mawa et al. 2020).This suggests that these forests have a limited potential to achieve high natural regeneration densities.Human interventions, specifically enrichment, may thus be desirable to regenerate K. anthotheca.Because enrichment with seedlings features high costs (Keefe et al. 2012), targeted seed supplementation may be a more efficient alternative.The few remnant seed trees in Uganda's natural forests should be actively protected and seed production from them prioritized.Further research work on seed-based restoration to restore K. anthotheca is needed to fully evaluate this approach.

Effect of Soil pH
Soil pH is the primary determining factor for the bioavailability of most nutrients and toxic elements (Hong et al. 2018).
Numerous studies have shown strong correlations between pH and soil nutrient concentrations (Bigelow & Canham 2002;Liu et al. 2021).Consequently, an increasing number of studies are using pH as an indicator of the chemical status of the soil (Dyderski et al. 2018;Gentili et al. 2018;Zarfos et al. 2019).Studies have found pH among the numerous soil properties to be of particular importance for tree regeneration (Liu et al. 2011;Annighöfer et al. 2015;Do et al. 2020).In this study, high abundances of natural regeneration were associated with low pH sites.This result seems to contradict the view of better performance of K. anthotheca in fertile and well-drained soil conditions (Katende et al. 1995).In conditions of low soil fertility and high acidity, competition in the regeneration layer is low (Dyderski et al. 2018).Since K. anthotheca is considered to be a poor competitor, it is plausible that in fertile soil conditions, other species dominate and outcompete K. anthotheca (Eggeling 1947).This is supported by our data that shows a higher density of other seedling species in plots without K. anthotheca regeneration.Although we acknowledge the limitations of only testing the effect of pH, our results indicate that successful regeneration of K. anthotheca is likely to be achieved on sites with low soil pH.In more fertile sites, control of interspecific competition may be necessary to successfully regenerate this species.Future studies could evaluate the effect of other soil constituents to better understand the role of soil factors on the regeneration of K. anthotheca.

Effect of Light Availability
Light availability is widely regarded as the most decisive factor for the successful regeneration of African mahoganies (Makana & Thomas 2004;Bahati 2005).In this study, we found no significant positive effect of light availability on regeneration abundance.K. anthotheca is considered a late successional species (Kirika et al. 2010), able to establish well in both closed and open canopy forest conditions (Mwima et al. 2001;Figure 5. Predicted probability of occurrence of natural regeneration of Khaya anthotheca depending on the density of adult conspecifics (A) and soil pH (B) according to the negative binomial model (Table S3).All other covariables are fixed to their means.Shaded areas represent 95% CI.Makana & Thomas 2005).This suggests that light availability is not a decisive factor for the early establishment of this species.Probably for that reason, we found no significant effect of light availability on regeneration density.The seedlings found in our study significantly comprised an ontogenetic stage characterized by seedlings with heights less than 50 cm.These are known to be established under a wide range of light conditions.According to other studies, the light availability requirement for K. anthotheca appears to change with the ontogenetic stage (Augspurger 1984;Poorter et al. 2005).It is reported that less than 2% of full sunlight irradiance is optimum for K. anthotheca seed germination (Kyereh et al. 1999).At high light levels, as found in large canopy gaps, Makana and Thomas (2005) caution that germination of this species may be severely depressed.This is ascribed to K. anthotheca flat seeds that are vulnerable to desiccation.After attaining a height of about 1 m, seedlings of K. anthotheca require more than 20% full light to recruit successfully to a large-size class (Hawthorne 1995;Agyeman et al. 1999).Given this, we suggest that adapting the light requirement to the seed and seedling stages of K. anthotheca be considered an essential aspect in the restoration of the species.Closed canopy conditions may be preferred for early establishment.Thereafter, to promote seedling development, light availability should be gradually increased by reducing overstory stocking.
In conclusion, K. anthotheca is a mahogany species of major economic and ecological importance in Uganda.Regeneration of this species is an important means of restoring degraded forests, allowing recovery of ecosystem functions and improvement of the economic value of logged forests.We evaluated how selected environmental factors affect its regeneration.We found that soil pH and the presence of seed trees were relevant factors.Our study also showed that light availability may not be a decisive factor for the early establishment of this species.Restorationists should thus pay attention to the various imperative environmental factors to successfully regenerate this species.

Figure 1 .
Figure 1.Map of Budongo Forest Reserve and the location of the two study forests.

Figure 2 .
Figure 2. Plot design (Only illustrative and not drawn to scale): Circular plot of 1,000 m 2 for tree assessment (≥10 cm DBH) and four 4 m 2 regeneration (<10 cm DBH) plots, two in each cardinal direction.

Figure 3 .
Figure 3. Distribution of Khaya anthotheca regeneration by height class (cm) in the logged and unlogged forests within Budongo Forest Reserve, Uganda.Error bars show AE SE.

Figure 4 .
Figure 4. Predicted regeneration abundance of Khaya anthotheca depending on the density of reproductive conspecifics (A) and soil pH (B).Predictions according to the negative binomial model (TableS3).All other covariables are fixed to their means.Shaded areas represent 95% CI.

Table 1 .
Biotic and abiotic characteristics of logged and unlogged forest sites within the Budongo Forest Reserve, Uganda.The mean and standard error (SE), t value (t), and p value (p) are shown for each variable.Bold p values indicate variables that differed significantly.

Table 2 .
Biotic and abiotic attributes of plots with and without Khaya anthotheca regeneration in Budongo Forest Reserve, Uganda.For each variable, the median value, mean, and standard error (SE) are shown.Variables, litter depth, basal area, and stocking were compared using t-tests because they met all the required assumptions.Density of reproductive, nonreproductive, and total conspecifics, cover of undergrowth vegetation, and soil pH were assessed with Wilcoxon tests.