P.Addo-Fordjour·E.K.Boakye·Z.B.Rahmad

Abstract We determined the response of tree community structure to logging disturbance and topography,and the patterns of tree-habitat associations in Tano Offin Forest Reserve,Ghana.We sampled trees in 27 20 m × 20 m plots randomly and equally distributed in three topographic habitats (slope,valley,and hilltop) in each of two forests:logged and unlogged.Two topographic features,altitude and degree of slope,were measured and related with species composition.Overall,there were significant effects of logging and topographic habitat and their interaction on species diversity and composition,with the unlogged forest and valley habitat supporting higher diversity.Tree diversity varied among the topographic habitats in the logged but not in the unlogged forest.There were topographic effects on abundance of individual species but not on tree community abundance and basal area.Logging and its interaction with topographic habitat showed significant effects on tree abundance and basal area.Some species were associated with specific topographic habitats or a combination in the logged and unlogged forests.However,the patterns of habitat associations of the species differed between the logged and unlogged forests.

Keywords Altitude·Community structure·Disturbance·Habitat heterogeneity·Slope·Topographic position

Introduction

Tropical forests provide critical ecological services to sustain human life and wellbeing.Specifically,they store about 46% of terrestrial carbon,thus playing an important role in global carbon storage and sequestration (Soepadmo 1993).They also produce approximately 40% of global oxygen(Beck 2019).Tropical forest ecosystems contain considerable amounts of the world’s biodiversity,harboring over 50% of the world’s terrestrial species (Myers 1988).In different parts of the world,these ecosystems provide tangible services for humans.For example,they provide timber,medicinal plants,fuel and food.Furthermore,they serve as habitats for numerous wildlife species.For instance,almost one million insect species have been identified although a higher number (5–10 million species) is estimated to be present (Lewis and Basset 2007).In spite of their importance,tropical forests are threatened by activities such as urbanization,infrastructure development,mining,logging and farming (Morris 2010).Considering the critical role of tropical forests in sustaining life on earth,there is the urgent need to better manage and conserve them in the face of disturbances so as to enhance their ecological services.

Understanding the patterns of plant community structure and the processes shaping them are key to managing and conserving tropical forests,and forms an important goal in community ecology (Solefack et al.2018).A number of studies have investigated plant community structure,revealing that several ecological factors act together to determine community structure (e.g.,Lu et al.2011).For instance,climatic factors and soil properties affect tree community structure in tropical forests (Zhang et al.2016).Disturbance is also an important factor shaping community assembly(Carreño-Rocabado et al.2012;Feng et al.2014).Equally,topography is an important determinant of tree community structure,although individual topographic variables do not show common effects on tree communities (Homeier et al.2010;Bentos et al.2017;Méndez-Toribio et al.2017).

Forests with different topographic habitats offer great opportunities for diverse tree species and communities to thrive.This is possible because topographic heterogeneity results in heterogeneity in microclimates which influence several aspects of plant vegetation (Mesquita et al.2018).Such microclimates often covary with topography,resulting in topographically related variation in environmental factors which influence vegetation characteristics (Qui et al.2001;Lookingbill and Urban 2005;Zhang et al.2011;Cui and Zheng 2016).For example,soil moisture availability varies with topographic positions (Méndez-Toribio et al.2016),with valleys having higher soil water content than other topographic positions (Yamada et al.2007;Chen et al.2018).Thus,topography can drive tree community structural and compositional heterogeneity,enabling a wider range of species to occur (Kubota et al.2004).For example,Mesquita et al.(2018) recorded strong topographic habitat effects on tree species richness in a subtropical forest in Colorado,USA.Similarly,tree distribution and species composition were affected by topographic variables in a Puerto Rican rain forest (Basnet 1992) and some Asian forests (Wang et al.2017;Jucker et al.2018).Topography also drives species-habitat associations in tropical forests (Guo et al.2017;Liu et al.2018).The existence of topographic heterogeneity within a forest offers species the opportunity to associate with a range of habitats (Guo et al.2017).Altitude,slope angle and slope aspect may serve as barriers to seed dispersal which influence species distribution and habitat associations (Guo et al.2017;Liu et al.2018).Assessment of species-habitat associations enables ecologists to determine how species respond to environmental heterogeneity,making it possible to predict the effects of different habitats on species diversity (Cramer and Willig 2002).In view of the increasing list of threatened forest habitats and species(Wilcox 1995),this assessment would be a useful tool to conserving species and their habitats,and maintain biodiversity in tropical forests.

Although the effects of disturbance and topography on tree community assemblages have been widely reported in tropical regions,there are still some information gaps.Most of the studies either considered only disturbance or topography,without examining the two factors to determine their interaction on tree community structure.Accordingly,information is scarce on how trees respond to disturbances in topographically defined habitats,especially in African tropical forests.Generally,the findings suggest that the patterns of tree species-habitat associations are forest specific but not universal,as species associate differently with different habitats in tropical forests.Since knowledge about specieshabitat associations are needed to conserve species and habitats,it is important that forest specific tree species-habitat association patterns are determined.In tropical forests,disturbance occurs as a common phenomenon,affecting several aspects of biodiversity.As disturbance influences species abundance (Armesto and Pickett 1985;Ofosu-Bamfo et al.2019) and habitat heterogeneity (Mooney 2011),it has the potential to affect species-habitat associations in tropical forests.Depending upon the degree of disturbance,there may be an increase in light availability which promotes lightdemanders,leading to a possible shift in habitat association (Brenes-Arguedas et al.2011).Torimaru et al.(2017)reported that disturbance due to landslides caused deviations from expected species habitat associations in an old-growth beech forest.To obtain accurate information on tree species association with topographic habitats,disturbance effects must be taken into account.This will provide more accurate knowledge on species-habitat associations which could be employed in selecting indicator species for prioritizing habitats for conservation.

The aim of our study was to determine the response of tree community structure to logging disturbance and topography in a tropical upland evergreen forest in Ghana.We also sought to determine the patterns of species association with topographic habitats and how the patterns might be shaped by logging disturbance.Specifically,we hypothesized that:(1) topographic factors and logging,together with their interaction,shape tree community structure,and (2) tree species will associate differently with logged and unlogged topographic habitats.

Methods and materials

Study area

The study was conducted in an upland evergreen forest within Tano Offin Forest Reserve,one of three upland evergreen forests in Ghana lying between latitudes 6°54′ and 6°35′ N and longitudes 1°57′ and 2°17′ W.The Reserve covers an area of 413.92 km2along an elongated mountain range of 200–750 m a.s.l (Addo-Fordjour and Owusu-Boadi 2016).A double-maxima rainfall pattern occurs in the area,with a major rainfall (1700–1850 mm) season from March to July,and a minor one (1000–1250 mm) from August to mid-November.The mountain peaks receive high amounts of rainfall and mist.The dry season usually starts in mid-November and ends in March.Mean relative humidity and temperature range from 87 to 91% and 19–33 °C.

Sampling design and data collection

Tree community structure and habitat association were assessed within three topographically defined habitats (valley,slope and hilltop) located in logged and unlogged forests.The habitats differed in mean altitude (valley:248 m a.s.l.,slope:314 m a.s.l.,hilltop:560 m a.s.l.) and degree of slope (valley:9.7%,slope:30%,hilltop:17%).The two forests share the same parent rock material,soil type and climate and thus we assumed that differences in tree community assemblages between the forests would be due to logging.Data was collected from February to June 2017,10 years following selective logging.The three habitats were exposed to similar logging intensity (20 trees ha−1on average) and only once;no other human disturbance occurred in the logged forest.On all the logged sites,selective logging targeted commercially valuable timber species (about 60 species in Ghana) with a minimum felling diameter at breast height (dbh) of 50 cm.Trees that met this criterion were tagged and randomly selected for felling,starting with larger stems.Due to uniformity of treatment on the logged sites,we assumed that variations in topographic patterns among the habitats in logged forest would not be biased by logging.In each forest,27 20 m × 20 m plots were randomly demarcated and equally distributed over the three topographic habitats (i.e.,9 plots per habitat),resulting in a total of 54 sampling plots.The forests were separated by a distance of 5 km,and the minimum distance separating the habitats in the forests was 1 km.To ensure independence of the plots,a distance of at least 300 m was maintained among them.Trees of diameter at breast height ≥ 5 were identified and recorded with the assistance of a plant taxonomist and the use of identification manuals (Hawthorne 1990;Arbonnier 2004;Poorter et al.2004;Hawthorne and Jongkind 2006).Nomenclature was in accordance with Hawthorne (1990)and Hawthorne and Jongkind (2006).To characterize the topography of the sampling areas,altitude and slope angle were measured using a GPS (Garmin GPSMAP 62) and a clinometer (Suunto SPUME-5/360PC),respectively.

Data analysis

Tree diversity was determined and compared at three levels:(1) among the topographic habitats using pooled data of the forests (hereinafter referred to as entire forests),with a view to obtaining the general topographic patterns,(2)among the topographic habitats in each forest to understand specific topographic patterns,and (3) between logged and unlogged forests under each topographic habitat.Diversity was represented by species richness and Shannon diversity index quantified using PAST statistical software,and the differences among treatments were examined through permutation tests.The rest of the statistical analyses were performed in R programme version 4.0.0 (R Core Team 2020) using R Studio version 1.2.5042.Since differences in abundance among samples can influence species richness patterns,individual-based rarefaction-extrapolation analysis was conducted with iNEXT package in R (Hsieh et al.2016) to determine whether the observed species richness was abundance mediated.It also enabled a comparison of sampling effort among the two forests as well as the habitats.Two-way PERMANOVA was run to examine the effects of each forest,topographic habitat and their interaction on tree species composition.In addition,homogeneity of multivariate dispersion (PERMDISP) was run to test for differences in dispersion between the forests and among the topographic habitats.Where both PERMANOVA and PERMDISP were significant,NMDS was used to supplement the interpretation of the patterns in composition among the treatments.PERMANOVA and PERMDISP were conducted withadonisandbetadisperfunctions in the vegan package,respectively.We visualized the patterns of species composition among the forests and the topographic habitats using non-metric multidimensional scaling (NMDS) ordination.The NMDS was run for the entire forest,thus ignoring logging effects,and rather examining topographic patterns in composition.The NMDS was also carried out for the topographic habitats in each forest.Finally,composition between the two forests was compared by pooling topographic data in each forest,thus ignoring topographic effect.We performed the NMDS analysis with themetaMDSfunction in the vegan package using Bray–Curtis distance measure.In addition,the relationship between tree species composition and topographic variables was examined by overlaying altitude and slope angle on the NMDS ordination using theordisurffunction in vegan package.Ordisurf fits smooth environmental surfaces on ordination using thin plate splines in generalized additive models (GAM).

Tree species abundance was compared among the topographic habitats by the means of a Chi square test usingchisq.testfunction in the stats package.Additionally,theaovfunction in the stats package was used to perform two-way ANOVA to assess the effects of logging,topographic habitat and their interaction on tree abundance and basal area.

The indicator value index (IndVal) was used to determine species association with the three topographic habitats in the individual forests using themultipattfunction in the indispecies package (De Cáceres and Legendre 2009;De Cáceres et al.2010;De Cáceres 2013).Significant indicator values were used to infer species associations with each topographic habitat type (Lyons et al.2018).Species areidentified as indicators of a habitat when their frequency of occurrence and/or abundance are high within sites of the habitat (Mouillot et al.2002).Because there were reasonable separating distances among sampling plots in the topographic habitats ensuring their independence,each plot was considered as a site for IndVal analysis.Twenty-seven sampling plots represented the three habitat types,defined and chosen a priori within each forest.Significant associations between the species and habitats were assessed using permutation tests with 10,000 iterations.

Results

Tree species diversity and composition

A total of 189 species belonging to 56 families were identified (Appendix S1).The most species rich families were Fabaceae (28),Euphorbiaceae (17),Rubiaceae (13),Meliaceae (13),and Sterculiaceae (13).There were significant effects of logging and topographic habitat on species diversity (Tables 1,2,3).For the entire forest,the valley habitat supported a significantly higher number of tree species than the other topographic habitats (Table 1;Fig.1 a).Shannon diversity index was highest in the valley followed by the hilltop and the slope habitats,with the differences among all the habitat pairs being significant.When species diversity (i.e.,species richness and Shannon diversity index) was compared among the habitats within each forest,no significant differences were observed in the unlogged forest (Table 2;Fig.1 b).However,within the logged forest,the valley and hilltop habitats had similar species diversity,each of which harbored significantly higher species diversity than the slope habitat (Fig.1 c).There were no significant differences in species diversity between unlogged and logged forests of valley and hilltop habitats (Table 3;Fig.2 a,b),although the slope habitat in the logged forest had significantly lower species diversity than its corresponding habitat in the unlogged forest (Fig.2 c).The patterns of the species richness in all the rarefaction curves were similar to those of the extrapolated species richness,showing that species richness values werenot influenced by abundance of trees.Moreover,the rarefaction and extrapolation curves showed that sampling efforts were similar among the forests and different habitats.

Table 1 Comparison of tree community structure among the topographic habitats in an upland evergreen forest in Ghana

Fig.1 Individual-based rarefied-extrapolated curves for tree species in the habitats for a the entire forest,b unlogged forest,and c logged forest;solid lines are rarefaction curves from the reference sample and dashed lines indicate the extrapolation curves;dots are the observed number of individuals in the forests;95% confidence intervals (shaded regions) were obtained by a bootstrap method based on 50 replications

Fig.2 Individual-based rarefied-extrapolated curves of tree comparing species richness between unlogged and logged forest sites within the a hilltop,b valley and c slope habitats;solid lines are the rarefaction curves from the reference sample and the dashed lines the extrapolation curves;dots represent the observed number of individuals in the forests;95% confidence intervals (shaded regions) were obtained by a bootstrap method based on 50 replications

Table 2 Comparison of tree community structure among topographic habitats in unlogged and logged forests within an upland evergreen forest in Ghana

Table 3 Comparison of tree community structure between unlogged and logged forest sites in each topographic habitat of an upland evergreen forest in Ghana

There were significant effects of logging (PERMANOVA:F Model=4.30,P=0.001) and topographic habitat (FModel=4.01,P=0.001),as well as their interaction(FModel=3.20,P=0.001) on species composition.Post-hoctests revealed that significant differences in species composition occurred among all the habitat pairs (PBioferon=0.003).Differences in multivariate dispersion among the pairs were significant (PERMDISP;P=0.001).The corresponding NMDS ordination showed that valley and slope plots were clustered together,and hence the difference in composition observed in the PERMANOVA may not be real but just due to a dispersion effect (Fig.3 a).In the unlogged forest,species composition in the valley habitat was significantly distinct from composition on the slope (Fig.3 b;F Model=2.44,P=0.003) and hilltop (FModel=4.36,P=0.003).Nevertheless,species composition was similar between the slope and hilltop habitats (FModel=1.75,P=0.054).Multivariate dispersion was not significantly different between the habitats in the unlogged forest (PERMDISP;F2,24=1.25,P=0.369).In the logged forest,NMDS and PERMANOVA showed that species composition differed significantly among the habitats (Fig.3 c;valley vs slope:FModel=4.93,P=0.003;valley vs hilltop:FModel=2.18,P=0.006;slope vs hilltop:FModel=5.84,P=0.003).Differences in multivariate dispersion among the three habitats in the logged forest were not significant (PERMDISP;F2,24=1.09,P=0.352).A comparison of unlogged and logged habitats revealed that they differed considerably in species composition (Fig.4 a–c.PERMANOVA;P=0.001).There were no significant differences in multivariate dispersion among unlogged and logged forest habitats (unlogged valley vs logged valley:F2,16=0.27,P=0.608;unlogged slope vs logged slope:F2,16=0.002,P=0.989;unlogged hilltop vs logged hilltop:F2,24=0.003,P=0.957).The stress values of the NMDS ordination ranged from 0.06 to 0.14.

Fig.3 Non-metric multidimensional scaling (NMDS) ordination comparing liana species composition among the habitats for a the entire forest,b unlogged forest,and c logged forest;ellipses representing 95% confidence interval based on differences in tree species composition are indicated

Fig.4 Non-metric multidimensional scaling (NMDS) ordination comparing tree species composition between unlogged and logged forest sites in the a hilltop,b valley and c slope habitats;ellipses representing 95% confidence interval based on differences in liana species composition are indicated

There was a significant effect of altitude on tree species composition when NMDS ordination was related to GAM contour of altitude (Fig.5 a,REML=325.1,edf=5.78,P=0.001).Generally,there were separation of plots along the altitudinal gradient.Most plots in the valley habitat were separated from those in the hilltop habitat.Some of the slope habitat plots at intermediate altitudes were separated from plots in valley and hilltop habitats.Equally,there was a significant effect of slope degree on species composition in the forest (Fig.5 b,REML=189.2,edf=4.86,P=0.001).The hilltop plots were separated from valley and slope plots along a slope angle gradient,which also separated many of valley plots from slope and hilltop habitats.

Fig.5 Overlay of a altitude and b degree of slope gradients on NMDS ordination using generalized additive model (GAM)

Tree abundance and basal area

Different sets of species constituted the five most abundant species in each habitat of the unlogged forest;onlyCeltis mildbraediioccurred in more than one habitat.In the unlogged hilltop habitat,the five most abundant species wereTrichilia prieuriana,Calpocalyx brevibracteatus,Monodora myristica,Napoleonaea vogeliiandTriplochiton scleroxylon.They together contributed 26.3% of tree abundance in the habitat.In the unlogged valley habitat,the five most abundant species comprisingTabernaemontana africana,Hymenostegia afzelii,C.mildbraedii,Musanga cecropioidesandRauvolfia vomitoriaaccounted for 27.9% of tree individuals.In the unlogged slope habitat,the five most abundant species wereStrombosia glaucescens,Baphia nitida,C.mildbraedii,Baphia pubescensandNesogordonia papaverifera,which accounted for 30% of tree numbers.For the logged forest,C.mildbraediiandB.nitidaoccurred in more than one habitat.For the logged hilltop,the five most abundant tree species,accounting for 26.9% of tree abundance,wereC.mildbraedii,N.vogelii,S.glaucescens,Rinorea oblongifoliaandC.brevibracteatus.The five most abundant species in the logged valley habitat wereC.mildbraedii,B.nitida,Khaya grandifoliola,Celtis philippensisandCeltis zenkeri,accounting for 19.4% of tree inviduals.In the logged slope habitat,the five most abundant species,contributing 29.6%to abundance wereC.mildbraedii,B.nitida,Olax subscorpioides,Entandrophragma angolenseandT.africana.

Chi square tests showed that the abundance of some species differed significantly among the topographic habitats (Appendix S1;P＜ 0.05).This was the case in the unlogged forest where,for example,the abundance ofAidia genipifl ora,B.nitida,B.pubescens,C.brevibracteatusandC.mildbraediivaried across habitats.Within the logged forest,B.nitida,E.angolense,M.myristica,N.vogeliiandO.subscorpioideswere among the species that showed varying abundance across habitats.

The effect of logging on tree community abundance was significant (F1,48=29.49;P=0.001),even though there was no significant effect of habitat (F2,48=0.29;P=0.752).Specifically,abundance was higher in the unlogged than in the logged forest.The interaction between habitat and logging was significant (F2,48=5.15;P=0.004).For all habitats,unlogged sites harbored higher tree abundance than their corresponding logged sites.Likewise,logging showed a significant influence on basal area (F=36.55,P=0.001),but this was not the case for habitat (F2,48=0.312,P=0.729).The interaction of logging and habitat on basal area was significant (F2,48=7.45,P=0.001).For each logged habitat,basal area was lower than its corresponding unlogged site.

Tree species:topographic habitat association

There were 11 (6.8%) species (out of 161) that significantly associated with the habitats in the unlogged forest (Table 4).Three of them were associated significantly with the hilltop habitat,and one each associated with the valley and slope habitats.There were also significant species association with a combination of habitats;five for valley+slope combination,one for valley+hilltop,and one for slope+hilltop.In the logged forest,nine (7%) of the 128 identified species had significant indicator values in one or two of the topographic habitats (Table 4).Four species were associated with the valley followed by the hilltop with three species and the slope habitat with one species.One species had significant indicator value in association with a combination of valley andslope habitats.Two species,N.vogeliiandM.myristica,had significant indicator values associated with habitats in both logged and unlogged forests.N.vogeliiwas associated with the hilltop habitat in both forests.M.myristicawas associated with a combination of valley and slope habitats in the logged forest,and with hilltop habitat in the unlogged forest.

Table 4 Indicator species analysis of tree species in the topographic habitats of unlogged and logged forests in an upland evergreen forest in Ghana

Discussion

Our study revealed a clear variation in tree diversity among three topographically defined habitats,with the valley habitat supporting the highest diversity.Previous studies also recorded highest tree diversity in valleys compared with other topographic habitats (Homeier et al.2010).In the current study,altitude and degree of slope differed markedly among the three habitats.Differences in such topographic features drive variations in soil nutrients and moisture availability (Qui et al.2001;Zhang et al.2011;Cui and Zheng 2016),which in turn influence tree diversity trends among habitats.By nature of its position and features,valley habitats have higher soil nutrients and moisture levels(Yamada et al.2007;Chen et al.2018),conditions which enhance species diversity (Fujita et al.2009;Cui and Zheng 2016).Our study indicated that slopes harbored the lowest tree diversity.Steep topography results in high leaching of nutrients and erosion,leading to drier and nutrient-poor conditions that are not conducive for plant growth (Geekiyanage et al.2017).The low diversity recorded in the slope habitat may reflect the poor resource nature of the habitat.Our results revealed that logging influenced topographic patterns of species diversity and that the effects remained evident 10 years after logging.Logging on the slope was the major factor responsible for the habitat differences in species diversity in the logged forest as well as in the entire forest.Logging can drive a decline in species diversity on slopes,given that tree harvesting causes rapid degradation of vegetation and soils in such areas (Kideghesho 2015;Regmi et al.2015).Such biotic and abiotic destruction most likely contributed to the substantial reduction in species diversity on the logged slope.There was no evidence of the effect of logging on species diversity on valley and hilltop habitats 10 years later.Post-logging recovery of tree species diversity in these habitats may explain the above-mentioned observation,although it is possible that logging did not affect species diversity in the past.

There were topographically induced differences in species composition among the habitats,with altitude and degree of slope explaining the variations in composition.This is in agreement with previous studies (Wang et al.2017;Jucker et al.2018).Trees do not respond directly to topography but rather to various environmental factors that covary with topography (Lookingbill and Urban 2005).Given that altitude and slope degree varied among the habitats,the compositional differences observed may most likely be due to topographically induced changes in abiotic factors.Our study generally corroborates the finding that topographic variables play a role in structuring species composition in tropical forests (Wang et al.2017).The findings also show shifts in tree species composition in response to logging,irrespective of the habitat.Similarly,Saiful and Latiff(2014)studied different topographic habitats of a hill dipterocarp forest in Malaysia and reported that logging caused shifts in species composition.Changes in species composition in the logged forest are likely due to tree removal and the associated canopy opening which often facilitate colonization and recruitment of other species (Shima et al.2018).

We did not record evidence of topographic influence on tree community abundance and basal area,even though the abundance of some species varied across the habitats.The topographic variations observed appear to have a more robust effect on the abundance of individual tree species than community abundance and basal area.The disparity in the response of community and species abundance to topography shows that both forms of abundance must be included in studies to achieve a more comprehensive understanding of the effects of topography on plant abundance.Consistent with previous studies,logging affected tree community abundance and basal area significantly (Hall et al.2003;Toyama et al.2015).The reduction of these attributes in the logged forest is due to the removal of trees during logging and to post-logging mortality of residual trees and seedlings,which often occurs after logging (Kasenene and Murphy 1991).The shifts in tree abundance and basal area as well as species diversity and composition clearly confirm that disturbance strongly shapes tree community assembly in tropical forests (Carreño-Rocabado et al.2012;Feng et al.2014).

The indicator species analysis revealed topographic effects on species association with the habitats,or with a combination of them.In each of the forests,distinct sets of species were associated with specific habitats or habitat combinations.Generally,species show varying adaptation to environmental conditions across different habitats (Wang et al.2010).Given this,the habitat association patterns we recorded may denote specific preferences by some species for environmental gradients across the habitats.Stochastic processes such as dispersal limitation due to topography may also account for the differences in species-habitat association observed among the topographic habitats (Guo et al.2017;Liu et al.2018).Tree species-habitat associations differed between the logged and unlogged forests in many respects.The set of species and their associated habitats were different between the two forests.This may be attributed to logging disturbance which often causes changes in species abundance and their association with their habitats.Disturbance was previously reported as a factor causing changes in species-habitat associations in an old-growth beech forest(Torimaru et al.2017).The only similarity in species-habitat association between the logged and unlogged forests lies in the association ofN.vogeliiwith the hilltop habitat in both forests.This species may possess a wider ecological amplitude that allows it to thrive and associate with the hilltop habitats irrespective of logging disturbance.Logging,as a major cause of disturbance in tropical forests,can alter environmental conditions of forest habitats to favour lightdemanding species (Brenes-Arguedas et al.2011;Schleuning et al.2011).Some species that associated with the valley habitat in the logged forest are light demanders (e.g.,C.zenkeri,Entandrophragma cylindricum,K.grandifolia)(Hawthorne 1995;Opuni-Frimpong et al.2013;Ademoh et al.2017).These species were either present in only the logged forest or occurred in both forests,but in higher abundance in the logged forest.Increased light availability due to logging (Inada et al.2017) might have favoured the lightdemanding species (Brenes-Arguedas et al.2011),influencing their habitat association.Overall,our findings show that logging plays an important role in modifying tree specieshabitat association in tropical forests.

Conclusion

The results show that logging and topography influence tree species diversity and composition in the upland evergreen forest in Ghana.Logging was responsible for the topographic variation in species diversity and composition,depicting interactive effects of the two factors.Logging had a strong influence on tree community abundance and basal area,but the effect of topography was not apparent.Some species were associated with topographic habitats in the logged and unlogged forests but the patterns differed.