Factors In�uencing �iana �pecies �ichness and �tructure following Anthropogenic Disturbance in a Tropical Forest , Ghana

e study was conducted to determine the factors that in�uenced liana species richness and structure in forests of different disturbance intensities (high, moderate, and low disturbance forests) in the Southern Scarp Forest Reserve, Ghana. Within each forest, lianas (dbh ≥ 2 cm) were enumerated in six 25 × 20m plots located along transects. Soil physicochemical properties and forest structure were determined within the plots. Liana species richness and abundance were signi�cantly lower in the high disturbance forest (PP P 0P00P) whereas basal area was signi�cantly higher in the low disturbance forest (PP P 0P0PP). Tree abundance and dbh signi�cantly predicted liana species richness and structure in the study (PP P 0P05). On the basis of the importance value index, three main liana communities, each corresponding with a forest type, were identi�ed. Stepwise multiple regression analysis revealed that exchangeable magnesium and calcium, and total exchangeable bases were themain soil variables that affected liana species richness. Liana structure was in�uenced by the above-mentioned soil variables as well as exchangeable potassium and sodium, and pH. e present study has demonstrated that changes in liana species richness and structure following human disturbance may be due to variations in soil properties and forest structure.


Introduction
Lianas are woody climbers that are rooted in the soil and climb other plants to the reach forest canopy [1].ey have signi�cant in�uence on forest ecology and ecosystem function, particularly in tropical forests (cf.[2]).For instance, they help to stabilise the microclimate of the forest �oor by forming a mass of leafy vegetation to close canopy gaps [3].Lianas may help maintain tree diversity in the forest by causing tree falls which could reduce the dominance of tree species [4].Heavy liana loads on trees can cause mechanical damage of the hosts and also reduce their growth rates [5,6].Additionally, lianas could impact negatively on natural regeneration of trees in forest ecosystems [7].In the light of recent reports of increasing liana abundance in tropical forests [8][9][10][11][12][13][14], lianas could modify forest ecosystems through their in�uence on tree regeneration and growth.Understanding the factors that govern liana community assembly in tropical forests is therefore of utmost priority, and necessary in developing forest conservation strategies.
Human disturbance has been identi�ed as one of the main drivers of liana success in tropical forest ecosystems [3,4,11,15,16].Both patterns of increasing and decreasing liana diversity and structure have been reported in tropical forests, although majority of them have centered on the former [17].ough human disturbance could have direct effects on liana communities through species removal, it could also cause changes in certain characteristics of the ecosystem such as forest structure and soil properties, which in turn, could in�uence liana community assembly.For this reason, the indirect effects of human disturbance on liana success following disturbance should be factored in studies that attempt to examine the role of human disturbance in structuring liana communities.Currently, studies that look at the indirect effects of human disturbance on liana success, mediated by changes in forest structure and soil properties, are scarce.ough a number of studies have considered the effects of environmental factors such as total rainfall, seasonality of rainfall, altitude and soil properties on lianas [2,3,18], they were not related to human disturbance.Consequently, information on the factors in�uencing liana species richness and structure following human disturbance is scanty.e current study was, therefore, conducted to determine the factors that affect liana species richness and structure within forests modi�ed by human disturbance in the Southern Scarp Forest Reserve, Ghana.

Study Area. is study was conducted in the Southern
Scarp Forest Reserve (coordinates, latitude 6.55 ∘ and longitude −0.66667 ∘ ) located about 14.71 km to the North-West of Mpraeso near Osubeng, in the Eastern Region of Ghana (Figure 1).e reserve covers an estimated area of about 15,462 hectares.It is the largest forest reserve in the Eastern Region of Ghana and falls within the moist semideciduous forest zone.
e reserve experiences double maxima rainfall pattern since it falls within the west semiequatorial region.e major rainy season starts from April and ends in July whereas the minor rainy season starts from September and ends in October.e forest experiences an annual average rainfall between 1580 mm and 1780 mm, with mean monthly temperature ranging from as high as 30 ∘ C in the dry season to about 26 ∘ C in the wet season.Relative humidity ranges between 75% and 80%.

Plot Selection and Demarcation
. e study was conducted in three forest types with varying disturbance index/intensity (DI): low disturbance forest (DI = 0.07), moderate disturbance forest (DI = 0.14), and high disturbance forest (DI = 0.28).Disturbance index (intensity) was determined for each forest as a percentage of trees that belonged to the pioneer species [19].is index is similar to the "pioneer index" developed by Hawthorne [20,21] and Hawthorne and Abu Juam [22] which has been shown to re�ect disturbance history at various scales in Ghanaian forests [19].Likewise, the disturbance index also incorporates the effects of all types of disturbance throughout recent history as indicated by Bongers et al. [19].Differences in disturbance intensity re�ected in differences in diversity and structure of the forest types.Farming and logging are the major human activities that have taken place in the forest.Two stands were sampled within each forest, and in each stand a transect of length 500 m was located.ree 25 × 20 m 2 plots were placed on alternate sides of each transect at a regular distance of 100 m.Transect sampling was used in the study so as to include all the possible gradients of disturbance within each forest type.

Data Collection.
Trees whose diameter at breast height (dbh) was ≥10 cm and lianas with dbh ≥ 2 cm were identi�ed, counted, and their dbh measured.Only lianas that were hosted on trees were included in the sampling.Trees were enumerated regardless of whether they were hosting lianas or not, so that liana infestation per plot could be determined, and the average for the forests calculated.Canopy cover was determined for the plots using a spherical densiometer.Plant identi�cation was carried out by �eld observations with the assistance of a plant taxonomist as well as manuals and Floras [23][24][25][26].Plant identi�cation was later con�rmed with herbarium specimens at the hebarium of Kwame Nkrumah University of Science and Technology, Kumasi, Ghana.

Collection and Analysis of Soil Samples.
Using soil auger, soil samples were collected from the depth of 0-5 cm, 5-15 cm, and 15-25 cm within the 25 × 20 m 2 plots.ere were �ve replicates for each depth in a plot.Soil samples collected were then composited to obtain a single sample per plot.Soil samples were analysed for pH (1 : 1 H 2 O), organic matter, total nitrogen, percent moisture, exchangeable calcium, exchangeable magnesium, exchangeable sodium, exchangeable potassium, exchangeable acidity (Al + H), available potassium, available phosphorous, total exchangeable bases (TEB), effective cation exchange capacity (ECEC), and percent base saturation at the Soil Research Institute of the Council for Scienti�c and Industrial Research, Kwadaso, Ghana, following the methods of Soil Survey Staff [27].( e patterns of liana species dominance were determined by conducting principal component analysis (PCA) on the IVI of the species.
To understand the effects of forest structure on liana species richness and structure, simple linear regression analysis was run between the following pairs of forest structure and liana variables: tree dbh versus liana dbh, tree dbh versus liana abundance, tree abundance versus liana abundance, tree dbh versus liana basal area, canopy cover versus liana species richness, and canopy cover versus liana abundance.Liana species richness and abundance were compared between the forests with analysis of variance (ANOVA).e levels of the soil physicochemical properties were compared among the forests with ANOVA.Fisher's LSD pairwise comparison tests were used to determine differences of means among forest pairs.
Stepwise multiple regression analysis was conducted between liana species richness and structure (abundance, dbh, and basal area), and soil variables to obtain the soil variables that explained variation in liana species richness and structure.e forward selection method was used to eliminate redundant soil variables and reduce collinearity.Pearson's correlation analysis was run to determine the relationships between liana infestation and the various soil properties as well as forest characteristics (tree abundance, tree basal area, and canopy cover).All analyses were performed with the 12th Edition of the GenStat soware (VSN International Ltd., Hemel Hempstead, UK), except that the PCA which was conducted with the CANOCO soware, at a signi�cance level of 5%.

Liana Species Richness and
Structure.In all, 57 liana species belonging to 35 genera and 19 families were identi�ed in the study area (Table 1).Liana species richness was signi-�cantly lower in the high disturbance forest than the other forests.(Table 2; df = 2;   .).e most diverse families in the forest reserve were Apocynaceae (9 species), Fabaceae (7 species), and Celastraceae (7 species).e most speciesrich families in the low disturbance forest were Celastraceae, Apocynaceae, and Convolvulaceae which formed 43.5% of the total species in this forest type (Figure 2).e most species-rich families in the moderate disturbance forest were Apocynaceae, Celastraceae, Fabaceae, and Icacinaceae which contributed 56.0% of the species in this forest.In the high disturbance forest, Apocynaceae, Fabaceae, Connaraceae and Celastraceae constituted the most diverse families, contributing 60.0% of the species in this forest.
Altogether, there were 365 individuals of lianas identi�ed within the Southern Scarp Forest Reserve, which translates into 406 individuals/ha.Liana abundance in the high disturbance forest was more than twice as high as that of the low disturbance forest and more than thrice as high as that of the moderate disturbance forest (Table 2).Liana abundance and infestation differed signi�cantly between the forests (Table 2;   .).Mean liana dbh per plot was similar among the forests (Table 2;  = .).However, mean liana basal area per plot was signi�cantly higher in the low disturbance forest compared with the other forests ( = .).Mean liana infestation per plot differed signi�cantly with respect to all the forest pairs (Table 2     accounted for 37.8% of the variation.e �rst two axes clearly separated the three forest types based on the dominance of the species (Figure 3).Additionally, most of the species were separated into three distinct liana communities based on their dominance values (IVI), with each community in a particular forest.For instance, Leptoderris sp., Leptoderris miegei, Agelaea petiolata, Griffonia simplicifolia, and Dalbergiella welwitschii formed part of the liana group in the high disturbance forest.Species such as Millettia latifolia, Cissus petiolata, L. hirsuta, Cercestis afzelii, and Neuropeltis prevosteoides formed part of the liana cluster in the moderate disturbance forest.In the low disturbance forest, species such as Artabotrys insignis, Adenia sp., Salacia staudtiana, Combretum cuspidatum, and Uvaria mocoli formed part of the liana cluster.e PCA revealed another group of species which occurred in between the main liana clusters of forest pairs.ese were species that showed high dominance in two forest pairs.For example, Motandra guineensis was found between liana clusters in the low and high disturbance forests because it had high dominance in these forests.Farquharia elliptica and Acacia kamerunenesis showed high dominance in the moderate and high disturbance forests and hence were located in between the liana clusters of these forests.Finally, Rhaphiostylis sp., C. africanus, and A. rumicifolia recorded high dominance in the moderate and low disturbance forests and were therefore located between the liana groups found in these forest types.Species which occurred in all of the three forests with high IVI in at least one of the forests where distributed close to the center of the PCA diagram.
Most of the soil physicochemical factors differed significantly among the forest types, decreasing with increasing disturbance intensity (Table 3).Various relationships were expressed in the stepwise multiple regression analysis performed between liana species richness and structure, and soil properties (Table 4).Five of the soil variables had signi�cant in�uence on liana species richness and structure in the study (  ).Liana species richness and abundance signi�cantly depended on exchangeable magnesium, exchangeable calcium, and TEB.In addition, liana abundance signi�cantly depended on exchangeable sodium.Liana dbh and basal area were signi�cantly in�uenced by exchangeable magnesium and pH.Furthermore, liana dbh and basal area had signi�cant relationships with exchangeable potassium and exchangeable sodium, respectively.Liana infestation related signi�cantly with some of the soil parameters, namely, pH (  2;   ), exchangeable acidity (  ;   7), % base saturation (  7;   ), available phosphorus (  ;   ), and available potassium (  ;   ).ere was, however, no signi�cant relationship between liana infestation and the rest of the environmental factors as well as forest structural characteristics (Table 5).

Discussion
e Southern Scarp Forest Reserve exhibited a higher species richness (56 species ≈ 62 species/ha) in relation to other forest reserves in Ghana.For instance, in the Bobiri Forest Reserve, Ghana, a total of 27 species (≈18 species/ha) was recorded [17] whereas a much lower species richness (27 species ≈ 9 species/ha) was reported in the Tinte Bepo Forest Reserve, Ghana [15].Moreover, liana species richness in this study was higher than those reported from outside Ghana: 51 species/ha in Bolivia [28] and 44 species/ha in China [29].A similar trend was observed with regard to liana abundance.In the present study, liana density of 406 individuals/ha obtained was higher than what was reported in the Tinte Bepo and Bobiri forest reserves (253 individuals/ha and 213 individuals/ha, resp.)[15,17].In the Democratic Republic of Congo, liana densities of 54 and 239 individuals/ha were recorded in a monodominant and mixed stands, respectively, in Ituri Forest [30].Nevertheless, liana density in this study did not compare favourably with the results from other tropical forests: 531 individuals/ha [31] and 3569 individuals/ha [32].e effect of disturbance on liana species richness and abundance was apparent in the current study.Generally, the high disturbance forest had signi�cantly higher liana species richness and abundance, con�rming the widely held �nding that disturbance promotes liana success [33][34][35][36].is pattern has been attributed to the ability of more disturbance forests to provide more favourable conditions for lianas than low disturbance forests [37][38][39].Apocynaceae and Celastraceae maintained high species richness in all the forests indicating the ability of their species to thrive in forests with varied disturbance intensities.
e forest structural characteristics expressed mixed relationships with liana species richness and structure in the study.Tree abundance signi�cantly predicted liana abundance in the study.is �nding provides support for other studies which reported of signi�cant dependence of liana abundance on abundance of trees [15,40].e absence of signi�cant relationship between liana abundance and canopy cover may be related to the ability of liana species in the forest to regenerate and grow well in shade and canopy gaps [39], leading to no clear cut pattern.Generally, the average level of liana infestation (48.8%) recorded in the Southern Scarp Forest Reserve was lower than what has been reported in other forests [2,17].e level of liana infestation increased signi�cantly with disturbance intensity, showing that high levels of disturbance could be more favourable to liana colonisation.ese �ndings indicate that different disturbance intensities could create differences in structure of forests, which could in turn, predict liana species richness and structure.
e patterns of dominance expressed by liana species indicated that liana communities in the three forests were distinct from one another.Unique species constituted the �ve most dominant species in each forest type with the exception of C. africanus which occurred in both the low and moderate disturbance forests.Furthermore, liana species were clustered into three different groups, one in each forest type.ese patterns clearly demonstrated that liana species dominance responded differently to different disturbance intensities in the study area [17].e equitability of distribution of IVI among liana species increased with disturbance intensity.For instance, the �rst ten species in the low disturbance forest accounted for 80% of the IVI of the species, whereas the �rst ten species in the moderate and high disturbance forests accounted for 67.9% and 62.1% of the IVI of the species, respectively.Disturbance therefore appeared to have reduced dominance of some species in the forests.is has the ability to enhance diversity of liana species in the forest reserve.
e �ndings of the current study have brought to the fore the important role soil properties can play in determining liana species richness and structure following human disturbance.Hitherto soil properties had been considered to be of less importance in determining liana species diversity and structure in some studies [2,[41][42][43].Soil pH and magnesium signi�cantly affected liana species richness and structure in this study, providing support for a previous study [44] which had a similar result.Soil pH and magnesium have been found to in�uence ecological amplitudes of lianas [45].In the light of this, it is possible that soil pH and magnesium affected liana species richness and structure by in�uencing the ecological amplitude of liana species (cf.[44]).Soil pH affects the availability of nutrients to plants by in�uencing nutrient solubility, microbial activity, and soil physical properties (cf.[44]).Low soil pH values were recorded in this study, and at such levels, nutrients availability becomes limited [46].us, the in�uence of soil pH on liana species richness and structure might have occurred through its regulation of nutrients available to lianas in the forest reserve [44].Generally, soil exchangeable cations exerted a greater in�uence on liana species richness and structure in the study compared to the other soil properties.Possibly, soil exchangeable cations in�uenced liana assemblage by controlling vital processes in lianas such as stomatal movement regulation, maintenance of plant turgor, phloem transport, protein synthesis, chlorophyll synthesis, and the activation of many enzymes, all of which are involved in growth and development of plants [47].Whereas forest structure did not affect liana infestation signi�cantly, some of the soil properties related signi�cantly with liana infestation, con�rming the hypothesis of Homeier et al. [2] that soil properties could be more decisive in liana infestation rate.Soil properties might have affected liana infestation by causing variation in liana abundance within the plots.In general, there were signi�cant variations in soil physicochemical factors among the forest types which re�ect the effects of human disturbance on soil fertility.In essence, human disturbance contributed to differences in soil physicochemical factors in the forests which in turn, in�uenced liana species richness and structure.us, apart from the direct effects of human disturbance on liana species richness and structure, it also had indirect in�uence on liana success through soil properties.
In conclusion, human disturbance had signi�cant in�uence on liana species richness and structure in the forest types.Most of the soil physicochemical factors varied significantly among the forest types and related signi�cantly with liana species richness and structure.In addition, liana species richness and structure depended signi�cantly on some forest structural attributes.Human disturbance resulted in distinct liana communities within the forests on the basis of IVI of the species.e study revealed that following human disturbance, liana species richness and structure may alter as a result of changes in forest structure and soil properties caused by the disturbance.

F 1 :
A section of the Southern Scarp Forest Reserve showing the study site.

2. 5 .
Statistical Analyses.Importance value index (IVI) of lianas was calculated using the following equation: IVI = Relative density + Relative basal area + Relative frequency.
T 2: Summary of liana characteristics in the low disturbance (LDF), moderate disturbance (MDF), and high disturbance (HDF) forests in the Southern Scarp Forest Reserve.
(RD: relative density; RF: relative frequency; RBA: relative basal area; IVI: importance value index).Mean values within the same row that have different superscripts are signi�cantly different (  0.05).
T 4: Multiple regression (stepwise) analysis of the effects of environmental variables on liana species richness and structure (abundance, dbh, and basal area) in the Southern Scarp Forest Reserve.e �nal model included only those variables which made signi�cant in�uence on the dependent variables.

T 5 :
Relationships between liana infestation and soil and forest structural characteristics in the Southern Scarp Forest Reserve.Pearson's correlation coefficient () and  values are provided.