Logging effects on seedling regeneration and diversity in a tropical moist semi-deciduous forest in Ghana

ABSTRACT The effect of time lag after logging on seedling dynamics was assessed in logged and unlogged forest plots located in the Pra-Anum Forest Reserve. Remnant tree attributes were sampled in six plots with sizes of 25 × 25 m in each forest type. Within each forest type, ten subplots with sizes 1 × 1 m were established for sampling seedling regeneration dynamics. Enumerated seedlings were grouped as pioneer, non-pioneer light demander, and shade-tolerant species. Regenerating seedlings were more diverse in the recently logged forest area characterised by Broussonetia papyrifera and Musanga cecropioides, and least in the unlogged forest driven predominantly by Celtis mildbraedii and Baphia nitida. Pioneer species were primarily dominant in the recently logged forest, while shade-tolerant and non-pioneer light-demanding species characterised the logged-over forest. Species richness and canopy openness of remnant stands explained 32 percent of the variations in regeneration richness between the logged and unlogged areas. Twenty-seven percent of the variations in seedling abundance was also explained by remnant trees richness, their size diameter and the degree to which their canopy has been opened. Logging-induced disturbance favours the sporadic emergence of pioneer species in recently disturbed areas which are later replaced by non-pioneer light-demanding or shade-tolerant species as light becomes limiting in the understory.


Introduction
Selective logging is a predominant timber harvesting practice and an important contributor to the economies of most tropical countries (Asner et al., 2009;Poudyal et al., 2018).It involves periodic removal of valuable timber trees of a specified minimum felling diameter (Cazzolla Gatti et al., 2015).The number of trees to be harvested differs from one country to another and at different harvesting cycles (Adam et al., 2006;Poudyal et al., 2018).For instance, in South-East Asia, about eight or more trees are removed per hectare at 35-year cutting intervals (Sist et al., 2003), whereas in Ghana, two to three trees are removed per hectare at 40-year intervals (Adam et al., 2006;Boakye-Dapaah, 1990).Usually, low-intensity logging is expected to result in less damage to the residual forest and enhance regeneration after logging (Ding et al., 2017;Cazzolla Gatti et al., 2015).However, the current selective logging system lacks any standard post-logging silvicultural interventions to enhance natural regeneration which has been cited as a cause for the poor regeneration of forests in Ghana (Swaine et al., 1997).
Natural regeneration is critical for the colonisation and establishment of native tree species, as well as the recovery of forests after disturbances (Crouzeilles et al., 2017).The process aims at encouraging the recruitment of trees, conserving biodiversity, enhancing the stability, as well as resilience of the forest ecosystem (Chazdon et al., 2020;Dyderski & Jagodziński, 2020;Hammond et al., 2021;Latawiec et al., 2016;Liira et al., 2011) and is influenced by both biotic and abiotic factors as well as their interactions (Khaine et al., 2018;Bose et al., 2016).Successful natural regeneration of logged forests in Ghana is attributed to the seeding ability of remnant older trees left un-harvested during logging (Wiafe, 2014).These remnant trees could act as nuclei modifying abiotic conditions, serving as seed sources as well as attracting potential seed dispersals (Manning et al., 2006;Nadkarni & Haber, 2009;Sandor et al., 2014).However, with increasing demand for timber across the tropics associated with a rise in deforestation and increased disturbance to forests, there is concern about the potentials impacts of selective logging on the loss of biodiversity, particularly mother trees of remnant forests.
Notwithstanding the reduced impacts of the selective logging approach, it also associated with disturbance and damage to the residual forests, with implications on soils, biodiversity, and other ecological processes (Asner et al., 2009;Picchio et al., 2020).The extent of these impacts depends on factors such as logging intensity, site differences due to soil, topography and biota (Bahati, 2005;Sist, 2000).For instance, natural regeneration is often promoted by low to medium intensity logging disturbance.In contrast, high-intensity limit regeneration by reducing the overall seedling numbers (Sukhbaatar et al., 2019) or temporally increases pioneer tree species (De Carvalho et al., 2017).On the other hand, the ability of the forest ecosystem to recover depends on the time lag after logging disturbance (Butarbutar et al., 2019;Vasseur, 2012).Yet studies on time lag after logging and recovery of logged forests along the succession trajectory are few in Ghana.
Depending on the species survival strategies and their ecology, regenerating species along the succession strategy could either be pioneer, shade tolerant, or non-pioneer light demanders (Hawthorne, 1995).Pioneer species are known to regenerate in gaps and dominate early succession forest; non-pioneer light demanders prefer shade to gaps, while the shadetolerant species are found in the understory and flourish well in unlogged shaded forests (Hawthorne, 1995;Hawthorne et al., 2012).Pioneer species usually colonise open areas but usually experience a high mortality rate which is replaced by non-pioneer species in the advanced stages of succession (Swaine & Hall, 1988).Understanding the dynamics of natural regeneration following selective logging is crucial for sustainable forest management (Juricka et al., 2019;Wagner et al., 2018) since the recovery and maintenance of desired native species composition are supposed to be enhanced by the current selective logging system.However, there are concerns about the selective system's potential in maintaining the long-term sustainability of forests (Gaui et al., 2019).Selective logging has been reported to negatively impact the structure, composition, and functioning of the forest ecosystem, thus flagged as an agent of forest degradation (Gaui et al., 2019).In Ghana, Hawthorne et al. (2012) reported the unsustainability of the current selective system.If true, it places in doubt the sustainability of the selective system as practices in Ghana.Therefore, studies need to provide an in-depth understanding of the selective system's short and long-term impacts on the forest ecosystem.
Despite the importance of studies on the recovery of forests after logging, few empirical studies in Ghana focused on trends in composition, diversity in shortand long-term following logging and the influence of forest stand attributes on seedling regeneration dynamics.In this study, we evaluated (i) seedling regeneration dynamics with respect to how diversity and composition differed in recently logged, loggedover and unlogged forest areas (ii) the proportion of pioneer, shade tolerant, and non-pioneer light demander in each forest area, and (iii) the influence of stand attributes (diversity, basal area, canopy openness) on seedling regeneration dynamics.

Study area
The study was carried out at the Pra-Anum Forest Reserve in Ghana.The reserve lies on latitude 6° 15'00" N and longitude 1°10'00''W in the moist semideciduous forest zone (Figure 1) and falls within a tropical humid climatic zone, which experiences a mean annual rainfall of 1650 mm, and an annual temperature of 26.5 °C (Hall & Swaine, 1981).The altitude of the reserve is about 100 m above sea level (MLNR, 2012).
The soil of the reserve is classified as oxysolochrosol intergrades (Brammer, 1962).Historically, the reserve has served as a timber production area and is now characterised by timber species like Celtis mildbraedii, Cylicodiscus gabunensis, Triplochiton scleroxylon, and Turreanthus africanus (Duah-Gyamfi et al., 2014).Compartments 68 (recently logged), 64 (logged-over), and 49 (unlogged) were selected for enumeration following a review of the forest reserves harvesting schedule at the Oda Forest District Office.

Sampling
Sampling was carried out in August 2020 in three areas of the reserve: a recently-logged (logged areas ≤ five years), logged-over (areas logged in the past twenty years), and unlogged (areas with no history of logging) forest areas.Six plots with sizes of 25 × 25 m were randomly established in each of the three forest areas (18 plots in total) for sampling older trees.Trees with diameter ≥ 10 cm at breast height (dbh, i.e., diameter at 1.3 m) were identified to species level with the assistance of a local botanist and the support of a field manual (Hawthorne & Gyakari, 2006).The family name of each species was obtained from the Global wood density database (Zanne et al., 2009).The diameter of each identified tree was measured with a Vernier calliper, and the basal area (BA) expressed in m 2 /ha was calculated as 0.00007854 × D 2 (where D = diameter at breast height; Nero et al., 2018).Tree heights (the vertical distance from the topmost living or dead part of the tree to the upslope side of the trunk base; Larjavaara and Muller-Landau, 2013) were measured using Nikon Forestry Pro II Laser Rangefinder/ Hypsometer (Nikon, USA).Percentage tree canopy cover and closure of each plot were estimated using the hemispherical photographs technique following Paletto and Tosi (2009) and local tree species richness at each plot was made comparable through rarefaction (Damptey et al., 2021).
Within each of the 18 demarcated plots for sampling older trees, ten 1 × 1 m subplots were established for sampling seedling regeneration dynamics (180 subplots in total).Plots were located systematically to take advantage of light reaching the understory facilitated by canopy openings due to logging induced disturbance.Plants with dbh less than 2 cm and/or less than the height of 1 m were considered valid regeneration (Rahman et al., 2020).Subplot counts of individual seedlings were pulled together for each plot and converted to seedling counts per hectare.Seedlings were classified into three ecological guilds based on their preference for light and/or their ability to grow well under shade.That is, Pioneer trees (Pi) regenerate and do well in gaps and are more dominant in the early successional forest; Non-pioneer light demander (NPLD) are species that prefer shade to canopy gaps, and shade-tolerant (SH) are species that tolerate and do well under shade (Hawthorne, 1995).

Data analysis
Diversity for both trees and seedlings were estimated with the "DIVERSE" function in PRIMER 7, based on the log-transformed (Log(X + 1) abundance data.Statistically significant differences between the three areas for all plant attributes were evaluated with a one-factorial permutational analysis of variance (PERMANOVA) using Euclidean distances and unrestricted permutation of raw data (N = 9999; Anderson et al., 2008).Forest area was used as a fixed factor with three levels (recently logged, logged-over, and unlogged forest area).In case of significant PERMANOVA results for plant variables compared for areas, a pair-wise PERMANOVA for post-hoc testing was carried.Area patterns in terms of regeneration species composition were displayed with non-Metric Multidimensional Scaling Ordination (nMDS).The most contributing species were further overlaid on the nMDS clusters, while the goodness of fit of the ordination was evaluated using a stress value (Clarke et al., 2014).
The family composition between forest areas was displayed with a shaded plot on a scale between 0 to 100 (0 represents no abundance, 50 represents intermediate, and 100 represents the complete abundance of a particular tree family in a forest area; Opuni-Frimpong et al., 2021).Uniqueness in regeneration seedling composition between the forest areas was evaluated with the similarity percentage analysis (SIMPER) based on Bray-Curtis similarity at a cutoff for low contributions of 70% (Somerfield & Clarke, 2013).Correlation between seedling diversity (richness and abundance) and older tree attributes (richness, diameter at breast height, and percentage canopy openness) were tested with Pearson correlation using the "ggubr" package (Kassambara, 2020)  R Core Team, 2019) were used for all statistical analyses and visualisations.

Diversity of regenerated seedlings
A total abundance of seedling regeneration of 1,576, belonging to 131 species, 101 genera, and 39 families, were recorded.The recently-logged forest area recorded both the highest seedling richness and abundance.The logged-over and the unlogged-forest areas followed respectively in the same order.The recentlylogged and the unlogged-forest areas were characterised by higher percentages of pioneer species (e.g., Broussonetia papyrifera, Musanga cecropioides, Triplochiton scleroxylon) followed by the shadetolerant (e.g., Celtis mildbraedii, Nesogordonia papaverifera, Baphia nitida) and the non-pioneer lightdemanding species (e.g., Corynanthe pachyceras, Pterygota macrocarpa, Sterculia rhinopetala).The logged-over forest area was characterised by the dominance of shade-tolerant species, followed by nonpioneer light-demanding and pioneer species, respectively.We could, however, not assign six (6) species to any of the ecological guilds (Table 1).

Seedling family composition
The unlogged forest is predominantly characterised by Fabaceae and Malvaceae, both contributing cumulatively about 70% to the unlogged forest family composition.The logged-over is also characterised by Malvaceae, Cannabaceae, Fabaceae, and Moraceae, cumulatively contributing about 76% to the family composition of the logged-over.The recently-logged area is dominated by Moraceae, Urticaceae, Malvaceae, and Cannabaceae family composition (Figure 4).

Dissimilarities in seedling regeneration composition between forest areas
An average dissimilarity of 65% seedling regeneration composition was observed between the unlogged and recently-logged forest plots.21 out of the 104 shared species between the two forest areas contributed a little  above 50% to the average dissimilarity between the two forest areas (Table 2).The most discriminating species were Broussonetia papyrifera (6%), Musanga cecropioides (5%), and Baphia nitida (4%).Corynanthe pachyceras, Dialium aubrevillei, Morus mesozygia, and Cola lateritia were completely absent from the unlogged forest area, while only Piptadeniastrum africanum was absent from the recently-logged area.
Between the unlogged and the logged-over areas, 64% average dissimilarity was observed, with 26 out of 102 species contributing about 50% to the average dissimilarity between them (Table 3).Baphia nitida, Celtis mildbraedii, Celtis zenkeri, Trichillia prieureana, and Triplochiton scleroxylon were the most discriminating species.For instance, a greater abundance of Baphia nitida was observed for the unlogged (average abundance = 2.10) than the logged-over (average abundance = 0.67) forest area.Ficus exasperata, Trichilia monadelpha, and Myrianthus arboreus were utterly absent from the logged-over forest area.59% dissimilarity was observed between the recent and the over-logged forest areas, with Broussonetia papyrifera,    Corynanthe pachyceras, Musanga cecropioides, and Dialium aubrevillei contributing the most to species composition between the two areas (Table 4).

Vegetation attributes of older stands
Tree species were more abundant in the unlogged forest than the logged forests (both recent and loggedover).However, the opposite trend was recorded for species richness.Dbh and height of stands were as well bigger and higher in the unlogged-forest than the logged-forest areas.A higher percentage of canopy openness characterised the recently-logged forest while the unlogged forest had a more closed canopy structure (Table 5).

Relationship between the diversity of seedling regeneration and older stand attributes
Out of the six possible predictors of seedling regeneration richness (Table 5), only older trees species richness (R = 0.97, P < 0.001; Figure 5A) and canopy openness (R = 0.54, P = 0.02; Figure 5B) correlated positively with seedling richness.Species richness and canopy openness of older stands explained about 32% of the variation in seedling regeneration richness.An increased percentage canopy openness (R = 0.70, P < 0.001; Figure 6A) and species richness (R = 0.65, P = 0.003; Figure 6B) of older stands correlated positively with seedling regeneration abundance.However, the diameter at breast height (dbh) of older stands correlated negatively with seedling regeneration abundance (R = −57, P = 0.013; Figure 6C).Canopy openness, species richness, and diameter of older stands explained 27% of the variations in regeneration seedling abundance.

Discussion
Natural regeneration is an important component of forest dynamics that influences recovery, biodiversity conservation, ecosystem functions, and sustaining the continuity of forest ecosystem services (Chazdon et al., 2020;Khaine et al., 2018).Poor regeneration could imply reduction or cessation in recruitment which threatens the sustainability of the forest ecosystem (Onyekwelu et al., 2021).Our study affirms mature stand richness, size of mother trees (dbh), wider canopy openness as the most significant factors influencing natural regeneration.

Effects of logging on seedling regeneration
The results of this study revealed the positive effect logging induced disturbance has on the initial stage of a community assembly process (Pedro et al., 2016).We recorded a higher seedling regeneration abundance and richness in the recently logged forest than the logged-over and the unlogged forests.Several factors, including soil disturbance, light created by canopy gaps, and resource availability (e,g, viable seeds), might have promoted seedling regeneration after logging.Logging induced disturbances usually create gaps with understory lights required for seedling regeneration or ignite soil microbial activities that supply nutrients and subsequently trigger small-scale forest succession (Harris et al., 2008;Souza et al., 2020).The result of our study is consistent with the work of Slyder et al. (2019) that affirmed how understory community diversity is enhanced by the recruitment of early successional species colonising from seed banks promoted by logging.Hammond and Pokorný (2020) asserted that the creation of gaps as a result of logging introduces light in the understory environment, which facilitates the growth and coexistence of different species and improves overall diversity.
In addition, resource availability in the form of adequate and viable seeds that fall on soils during logging is another possible factor for seedling regeneration in the logged forests.Usually, loggers target larger trees carrying a greater proportion of total seed output in the forest ecosystem (Lindenmayer & Laurance, 2017).When this happens, seeds fall on soils, and if such seeds are viable and soil conditions are favourable, regeneration is initiated.Though some seeds could be dormant, the churning action of skidders during logging exposes such seeds to resources (e.g., light) that initiate their germination processes.

The influence of older stand structure on seedling regeneration
Seedling regeneration richness correlated positively with older tree species richness in our study.The diverse the surrounding vegetation, the higher the seedling regeneration richness.We could attribute this positive relationship to the fact that more diverse mature trees offered numerous seed sources for regeneration.When logging create gaps or disturbed the soil, and other environmental conditions are favourable, viable seeds falling from older trees takes advantage and emerge as seedlings (Girma, 2012;Keeton & Franklin, 2005).The degree to which seedlings emerge depends on the intensity of seed-rain coupled with favourable micro-site conditions (Paluch et al., 2019).
A higher degree of canopy openness went along with higher seedling regeneration abundance and richness.This positive trend has already been confirmed by Olson and Wagner (2011), who attributed resource availability (i.e., light) in the disturbed environment as the major factor responsible for seedling regeneration.According to Feldmann et al. (2020), light transmission to the lower strata influences structuring processes, including seedling regeneration.Logging creates a heterogeneous canopy with different microenvironments that differ in light intensity and different degrees of soil disturbance, advantageous for seedling regeneration (Duah-Gyamfi et al., 2014).
However, the size class distribution of older trees caused a potential trade-off in seedling regeneration abundance in the unlogged and logged-over forest plots.These older trees probably limited regeneration ability with their bigger space occupancy and canopy layers which impeded light transfer in the understory.Besides, the older trees' multi-layered, complex canopy structure may have blocked some seeds from coming into contact with the forest floor (Hammond et al., 2021).Seeds left on canopy layers either got eaten by some birds, rodents or decayed with time.

Species dominance in logged and unlogged forest plots
The study revealed Broussonetia papyrifera, Musanga cecropioides, Celtis mildbraedii, Nesogordonia papaverifera, and Celtis zenkeri to be more dominant in the recently-logged forest plots compared to the loggedover and unlogged-forest areas.B. papyrifera, M. cecropioides, and C. zenkeri are pioneer species that are widely distributed in adverse environments (Zhang et al., 2020).In fact, B. papyrifera and M. cecropioides are indicator species and the occurrence of which signifies disturbance and degradation of a particular ecosystem (Cazzolla Gatti et al., 2017).Their dominance and inflation in their numbers confirm the fact that the area has suffered a myriad of disturbances in the past (Seidler & Bawa, 2013).These past disturbance events induced by logging (e.g., harvesting and skidding) in our study may have promoted the emergence and growth of these species, which finds favour for growth in light-rich open microhabitats (Seidler, 2017;Swaine & Agyeman, 2008).Ordinarily, the dominance of non-pioneer shadebearer species like C. mildbraedii, and N. papaverifera were not expected in the recently-logged forest plots because of their preference for shade.However, the dominance ability of such shade-bearer species could indicate the intrinsic reactions of such species to changes in the microenvironment (e.g., light) of the area with time (Duah-Gyamfi et al., 2014).Usually, canopy gaps in the tropics could favour the emergence of both pioneers and shade-bearers until the guild with an inherently long lifetime replaces the short-lived ones in the later stage of succession (Kuusipalo et al., 1997;Swaine & Hall, 1988).
The unlogged forest plots were also dominated by the abundance of non-pioneer light-demanding tree species like Piptadeniastrum africanum and Albizia zygia.Surprisingly, another pioneer species (Ceiba pentandra) was rather dominant in the closed canopy unlogged forest than the logged forests.Kyereh et al. (1999) observed that pioneer species germinate even under shade in forests, but they die or do not survive long because of negative carbon balance.

Conclusion
Canopy gaps, as well as adequate viable seeds falling from remnant trees, facilitated the successful establishment of seedlings in the recently logged forest.However, the bigger sizes of stands and the closed canopy nature of the logged-over and the unlogged forests limited the amount of light in the understory, minimising seedling diversity.Understanding the dynamics of natural regeneration is crucial for sustainable forest management, while the recovery and maintenance of desired native species composition will ensure the sustainability of forest ecosystems, particularly in the tropics where logging induced disturbance is prevalent.

Figure 2 .
Figure 2. Box plots of seedling regeneration A) abundance and B) richness in unlogged, recently and logged-over forest areas.The line in the box represents the median value; the box limits are the 25th and 75th percentiles, error bars show 10th and 90th percentiles on a log scale.Different letters indicate significant differences between the forest areas according to pair-wise PERMANOVA.

Figure 3 .
Figure 3. Non-Metric Multidimensional Scaling (n-MDS) Ordination of seedling regeneration composition for unlogged (green), recently (red), and logged-over (blue) forest areas with a stress value of 0.17.Highly correlated species (correlation at 0.7) for each forest area are overlaid.The distance apart between the individual forest plots reflect the degree of dissimilarity of species composition among them.The length and direction of vectors represent the strength and direction of the association.The circle indicates a maximum vector length corresponding to a Pearson correlation coefficient of 1.0.

Figure 4 .
Figure 4. Shade-plot of family composition for logged, recently logged and logged-over forest areas.The light-scale reflects a lesser abundance, while the deep scale reflects a higher abundance of families for each forest area.

Figure 5 .
Figure 5. Relationship between seedling regeneration richness and A) older trees species richness, or B) Older trees canopy openness (%) across logged and unlogged forest plots.

Figure 6 .
Figure 6.Relationship between seedling regeneration abundance and A) older trees canopy openness, or B) older trees species richness or C) older trees mean dbh across logged and unlogged-forest plots.

Table 1 .
Diversity and the ecological guild of species for the three study forest plots.

Table 5 .
Older stands attributes of logged and unlogged forest plots.