Plant Community Analysis along Environmental Gradients in Moist Afromontane Forest of Gerba Dima, South-western Ethiopia

Background: This study was carried out in Gerba Dima Forest, South-Western Ethiopia, to determine the oristic composition, species diversity and community types along environmental gradients. Ninety sample plots having a size of 25 X 25m (625m 2 ) were laid by employing stratied random sampling. Nested plots were used to sample plants of different sizes and different environmental variables. All woody plant species with Diameter at breast height (DBH) ≥ 2.5 cm and height ≥ 1.5m were recorded in 25 m X 25 m plots. Hierarchical (agglomerative) cluster analysis was performed using the free statistical software R version 3.6.1 using package cluster to classify the vegetation into plant community types. Redundancy Analysis (RDA) ordination was used in describing the pattern of plant communities along an environmental gradient. Result: One hundred and eighty plant species belonging to 145 genera, 69 families and comprising of 15 endemic species were recorded. Of these, 52 species (28.9%) were trees, 6 species (3.33%) were Trees/shrubs, 31 species (17.22%) were shrubs, 76 species (42.22%) were herbs, and 15 species (8.33%) were Lianas. Rubiaceae, Acanthaceae and Asteraceae were the richest family each represented by 11 genera and 11 species (6.11%), 9 genera and 11 species (6.11%), 6 genera and 11 species (6.11%), respectively of total oristic composition. Cluster analysis resulted in ve different plant communities and this result was supported by the ordination result. RDA result showed altitude was the main environmental variable in determining the plant communities. The ANOVA test indicated that the ve community types differ signicantly from each other with regard to Electrical Conductivity and Potassium. Conclusions: Description of oristic diversity of species in Gerba forest revealed the presence of high diversity richness. in the study forest shows the potential the

studies [6]. The multidimensional ecological niche space determines the distribution of a species within a community [7]. Physiographic and edaphic factors can determine which plant species will colonize a site since plant species vary in their tolerance and utilization of resources site [8]. These variations have been regarded as a driving force for the coexistence of species in a similar environment [9] and can explain broadscale compositional differences among multiple resource gradients [10,11]. The upper story tree density as an abiotic factor can also affect community composition as understory species differ in their ability to tolerate stresses imposed by competitive trees [12,13]. Moreover, by increasing the abundance of annual and biennial plants, disturbances can affect community composition via favouring stress-tolerant species [14,15].
Information on species composition and diversity of tree species plays a pivotal role not only to understand the structure of a forest community but also in planning and implementation of conservation strategy of the community [16]. Investigation of forest community composition and structure is very useful in understanding the status of tree population, regeneration, and diversity for conservation purposes [17]. Quantitative information on composition, distribution, and abundance of woody species has paramount importance in understanding the form and structure of a forest community and for planning and implementation of conservation strategy of the community.
The recent data on forest resources of Ethiopia reported in [18] puts Ethiopia among countries with a forest cover of 10-30%. According to this report, Ethiopia's forest cover (FAO de nition) is 12.2 million ha (11%). It further indicated that the forest cover shows a decline from 15.11million ha in 1990 to 12.2 million ha in 2010, during which 2.65% of the forest cover was deforested. This study was conducted in the Gerba Dima forest found in South-Western Ethiopia with the aim of investigating the species composition, species diversity, community types and to relate the distribution of plant community types to some environmental parameters.

The study area
This study was carried out in the Gerba Dima forest found in the Illu Aba Bora zone of Oromia regional state of Ethiopia and located between 7 0 45' to 8 0 10' North latitude and 35 0 29 ' to 35 0 50' East. The study forest is bounded by Baro River to the south and west direction whiles three other rivers, namely Bote, Hoyi and Sor cross part of the forest in the east (Figure 1). The geology of the study site is characterized by the Underlying basement rock consisting of intensively folded and faulted Precambrian rocks, overlain by Mesozoic marine strata and Tertiary basalt types [19]. The main soil types of the study area are red or brownish ferrisols derived from the volcanic parent material. Other soil groups in the area include nitosols, acrisols, vertisols, and cambisols soil types exist in the study site [20].
The rainfall data collected from the nearest Gore meteorological station to the study forest indicated that the study area receives very high annual rainfall and characterized by unimodal rainfall pattern, which shows low rainfall in December, January and February, gradually increasing to the peak period in August. The mean annual rainfall of 1854mm while the monthly mean maximum and mean minimum temperature of the area is 27.2 0 C and 13.3 0 C, respectively. The mean annual temperature is 19.2 0 C and with slight variation from year to year [21].

Sampling method
In this study, a strati ed random sampling design was used to collect vegetation and environmental data [23,1]. Using Arc GIS version 10.3, the study forest was strati ed based on the altitudinal gradient and three types of strata in the form of contour were established. Strata one was distributed between 1500-1800m altitudinal ranges whereas strata two and three were found between 1801-2000m and 2001-2300m altitudinal ranges respectively (Fig. 1). Sample plots were assigned in each contour in the form of Random points Using Arc GIS version 10.3.
Ninety sample plots having a size of 25 x 25m (625m 2 ) along each contour were laid. Nested plots were used to sample plants of different sizes and different environmental variables. All woody plant species with Diameter at breast height (DBH) ≥ 2.5 cm and height ≥ 1.5m were recorded in 25 m x 25 m plots. Within the major plots, ve 3m x 3m subplots (9m 2 ) was used to collect shrubs with dbh < 2.5 cm and > 1.5m height.
Within each 9m 2 subplots, two 1m 2 subplots were used to collect data on the species and abundance of herbaceous plants. Finally, the percent cover of all plant species found within the sample plot was visually estimated and converted to the Braun-Blanquet scale as modi ed by [24]. Every plant species encountered in each plot were recorded. Plant specimens were collected, pressed, dried and brought to the National Herbarium (ETH), Addis Ababa University for taxonomic identi cation. The specimens were determined by comparing with authenticated specimens housed at ETH and by referring to published volumes of Flora of Ethiopia and Eritrea [25, 26, 27, 28, 29, 30, 31, 32, and 33].
For analysing soil variables, soil samples were collected with a soil core sampler from the top 40cm depth within 1m x 1m subplots at the four corners and middle of the quadrat. Composite soil samples from samples collected from the four corners and the middle of quadrats were brought to the soil laboratories of AAU. The soil samples were air-dried, rolled and passed through a 2 mm sieve for laboratory analyses. These soil samples were analysed for pH, electrical conductivity (EC), sodium, potassium, organic matter, total nitrogen, available phosphorus and texture following standard procedures outlined in [36]. The pH and EC were measured using a pH meter and EC meter in the supernatant suspension of 1:2.5 soil-distilled water mixtures.
Available Sodium and Potassium were determined using a ame photometer. Organic matter was determined by the ignition method. The texture was determined on the basis of Bouycous Hydrometer method with the categories sand, silt, and clay (expressed as % weight) while total nitrogen was determined using Kjeldhal method. Available Phosphorus was determined by the Bray-I method and the absorbance of the Bray-I extract is measured at 882 nm in a spectrophotometer.

Data analysis
In this study, hierarchical (agglomerative) cluster analysis was performed using the free statistical software R version 3.6.1 [37] using package cluster to classify the vegetation into plant community types. The similarity ratio with Ward's group linkage method was applied for cluster analysis i.e. to determine plots that can be classi ed into the same groups based on the species abundance data. The decision on the number of groups (clusters) was based on objective methods of obtaining an optimal number of clusters, the Multi Response Permutation Procedures (MRPP) technique (no-difference hypothesis) and the ecological interpretation of the groups conducted in R program. The T and A statistic of MRPP output were used to obtain the number of clusters. The test statistic T describes the separation between the groups. The more negative T value, the stronger the separation. The P-value associated with T is determined by numerical integration of the Pearson type III distribution. The P-value is useful for evaluating how likely an observed difference is due to chance [38]. The agreement statistic A describes within-group homogeneity, compared to the random expectation, and falls between 0 and 1. When all items within-groups are identical A = 1 and 0 if the groups are heterogeneous. In community ecology, A values are commonly below 0.1, and an A value greater 0.3 is fairly high [38].
From the output of the objective method, a sharp bend at the speci c cluster in the plot could be a good indication of the number of clusters in the data [39]. The community types identi ed from the cluster analysis were further re ned in a synoptic table where species occurrences were summarized as synoptic coverabundance values [40]. Dominant species of each community type were identi ed based on their synoptic values and community types were named after one or more dominant species. The identi ed groups were tested for the hypothesis of no difference between the groups (clusters) using nonparametric Multi-Response Permutation Procedure (MRPP). Indicator species analysis was performed in R using package labdsv.
Indicator values were tested for statistical signi cance using a randomization (Monte Carlo) technique. Species richness, evenness, Shannon diversity and evenness indices were computed using the free statistical software R version 3.6.1 [37]. The Shannon diversity index (H') was calculated from the equation: Where pi, is the proportion of individuals found in the i th species. The values of the Shannon diversity index is usually found to fall between 1.5 and 3.5 and only rarely surpasses 4.5 [40,1]. The Shannon evenness index (J) was calculated from the ratio of observed diversity to maximum diversity using the equation: Where Hmax is the maximum level of diversity possible within a given population, which equals ln (number of species). J is normal between 0 and 1, and with 1 representing a situation in which all species are equally abundant [41].
In this study, Redundancy Analysis (RDA) ordination was used in describing the pattern of plant communities along an environmental gradient since the preliminary analysis of the vegetation data using Deterended Correspondence Analysis (DCA) revealed that the longest axis of DCA for the dataset was less than 3 (= 2.22).
Before the application of RDA ordination, environmental variables, which were relatively more important in explaining the species data, were selected using the Monte Carlo technique and function Adonis test for their signi cance. Computation of variance in ation factor (vif) was also conducted to eliminate those environmental variables that are collinear. The community types obtained were subjected to an ANOVA based on environmental variables to nd out whether there are signi cant variations between the groups. Pearson's product-moment correlation coe cient was calculated to evaluate the relationship between the environmental variables.
Based on the information available on the published Floras of Ethiopia and in [42] a total of 15 endemic plant species in 11 families were recorded (Table 1), comprising more than 8.33% of the recorded species. Asteraceae was the rst family having three endemic species, followed by Acanthaceae and Fabaceae (two species each). The remaining eight families have a single species each in the endemic species list. Among the total endemic species, herb, tree, shrub and liana growth forms were represented by 6,3,4,2 species respectively. Out of the 15 endemic species, Crotalaria rosenii and Polyscias farinosa have been included in the IUCN red data list of Ethiopia and Eritrea [42] qualifying for near threatened and vulnerable category respectively. In the Gerba Dima forest, at 625m 2 sample plot, species richness varied from 26 to 59 across the study plots. The Shannon diversity index also varied from 2.92 to 3.83 while evenness ranged from 0.89 to 0.95 in the study plots. The overall mean Shannon diversity index, species richness and evenness of the study area were 3.45, 41 and 0.93 respectively.

Community types and indicator species
Five community types were derived from the hierarchical cluster analysis in combination with Multi-response Permutation Procedures (MRPP) and objective method of the whole data set (Figure 3). From the output of MRPP, the test statistic T value for the ve groups was -38.26 (P < 0.001) and the agreement statistic A was 0.13 while the output of objective method revealed a sharp bend at the fth cluster.
Community 1 (Croton macrostachyus -Bersama abyssinica community) was found in the altitudinal range of 1677-2020 m. a.s.l and slope from at to 50%. Fourteen plots were associated with the community and has 2 indicator species with signi cant indicator values (P < 0.05).
Community 2 (Syzygium guineense -Olea capensis community) was distributed from 1699 to 2240 m a.s.l. and slope ranging from at to 60%. It comprises of 22 plots and twenty species were associated with this community as indicator species where one of the indicator species exhibit signi cant indicator values (P < 0.05). Community 3 (Dracaena afromontana-Pouteria adol -friederici community) was found in the altitudinal range of 1761-2000 m. a.s.l and slope from at to 25%. Thirteen plots were associated with the community community and seven species were associated with this community as indicator species while two of the indicator species showed signi cant indicator values (P < 0.05).
Community 4 (Vepris dainellii -Sche era abyssinica community) was distributed in the altitude range of 1720-2060m a.s.l. and the slope gradient varies at to 60%%. It comprised of 14 plots, eight species were associated with this community as indicator species, while four of the indicator species exhibited signi cant indicator values (P < 0.05).   From computation of vegetation data in the study area Shannon-Weiner diversity and evenness, indices for the ve community types showed the output in Table 4. that out of 14 environmental variables, seven were found to be signi cant in explaining patterns of plant community distribution. From the seven signi cant environmental factors, the vif values of sand and silt were higher than 5. Sand and Silt are highly correlated with at least one of the other variables in the model. One solution in dealing with collinearity is to remove some of the violating variables from the model and thus the one with higher vif value (sand) was eliminated. The result of RDA ordination showed that comparatively, the gradient of altitude and potassium was highly correlated on axis one and gradient of disturbance in axis two. The other factors were correlated with the ve axes with a different value of correlation. The eigenvalue for axis one, two and three were 10.65, 8.06, and 6.32 respectively. Cumulative proportion variance explained by the rst ve RDA axis of the joint biplot was 93.9%. The proportion of variation explained by ve RDA axis also shows a decline towards the successive higher axis (Table 5).  RDA ordination of the study plots of Gerba Dima forest formed ve groups or community based on the species composition. These ve community types were segregated following the arrows of the environmental variables. Community two mostly occur at the higher altitude while species in community one are distributed at the lower altitude and higher EC. Silt, Disturbance and potassium axes were strongly in uencing the distribution of community ve. Organic matter arrow has strongly in uenced the distribution of species in community three and four (Figure 3). The ANOVA test indicated that the ve community types differ signi cantly from each other with regard to EC and K. The result of Tukey's pair-wise comparison test indicates that community 4 and 1 differ signi cantly with respect to Disturbance and K while community 2 and 3 showed signi cant differences with respect to EC.

Floristic Composition and Diversity of Gerba Dima forest
The existence of diversi ed ora of Gerba Dima forest was in line with the general pattern of high species diversity in the tropical montane forests. According to [43], tropical forests are among ecosystems that harbour high species diversity of the globe. East African montane forests of Ethiopia, Kenya, Tanzania and Uganda are among the most diverse and richest African regions with regard to ora composition and endemic plant taxa [44, 45, and 46]. Asteraceae, Acanthaceae, Rubiaceae, Fabaceae and Euphorbiaceae are the ve dominant families, which contribute more than 27% of the total species in the study forest. These dominant families were also reported as top ten species rich families in many Neotropical forests and Asia [43]. Except for Rubiaceae, these families are also among the top ten species rich families in the ora area [47]. The dominance of the above families together with Rubiaceae was also reported in other moist afromontane forests of southwestern Ethiopia [48, 49, and 50]. Thus, the dominance of these families in the Gerba Dima forest agreed to their general dominance in the ora area and tropical forests. The dominance of these families in the study area could be attributed to their successful colonization to the landscape owing to their e cient pollination, dispersal and germination mechanisms [51]. For instance, many species of Asteraceae have umbrella shape structures adapted for air dispersal and increase their opportunity for their successful establishment [51].
Among the growth forms, herbs constitute more than 42% of recorded species. The prevalence of herbs could be attributed to the presence of canopy gap because of anthropogenic disturbance. Disturbance of forest in the form of selective cutting of trees favours the growth of herbaceous species in the forest understory. Under normal circumstances, the forest oor (herbaceous layer) of Afromontane rainforests is usually dark and poor in species composition owing to the closed canopy of the forest that prevents light from reaching the ground [52].
The higher value of Shannon diversity index and evenness indicates that the study forest has high species diversity with more even distribution of the species within the study plots. Species diversity increases when the populations have more even abundances and vice versa [41]. High Shannon evenness in the Gerba Dima forest indicates little dominance by any single species but the repeated coexistence of species over all the plots or sites. Therefore, the implication of evenness values is that, when there is a high evenness value in a given forest, the location of conservation sites might not be of much importance compared to when the evenness value of the forest is low

Plant community types in Gerba Dima forest
The output of Multi-response Permutation Procedures (MRPP) results in T statistics having more negative value with signi cant P-value (T= -38.26, P< 0.001) and an agreement statistic A (0.13) con rming the distinctness of clusters. The test statistic T describes the separation between the groups. The more negative T value, the stronger the separation. From the result of this study, the null hypothesis of no difference among groups can be rejected. The ve groups occupy different regions of species space, as shown by the strong chance correction within the group (A) and test statistic (T) and thus con rm the existence of 5 distinct plant communities in the Gerba Dima forest [53]. The ve plant communities showed a slight variation in their species richness, diversity and evenness. Relatively community types 1, 2 and 5 were the richest with respect to species richness and diversity while community types 3 and 4 the lowest. The differences in species richness among the ve communities could mainly be attributed to the dissimilarities of the communities in terms of location, altitude, human impact, rainfall, and other biotic and abiotic factors. According to [54], different altitudes and slopes in uence species richness and dispersion behaviour of tree species. Altitude and climatic variables like temperature and rainfall are also other determinant factors that affect species richness [55].

Plant community -Environmental variables relationship
In the current study, the multivariate analyses (both Ordination and cluster analysis) were consistent in showing the patterns of oristic grouping within the studied forest and hence the two methods are complementary. The variable with the highest score (0.88) associated with axis one was the altitude.
Therefore, altitude was the most important variable in weighting axis one and to interpret or explain the axis.
Similar studies conducted in other Afromontane forests of Ethiopia also con rm the importance of altitude as a major determinant of vegetation distribution along altitudinal gradients [56, 57, and 58]. Altitudinal change leads to changes in humidity, temperature, soil type, and other factors that in uence the growth and development of plants which in turn determine the patterns of vegetation distribution [59,60].
Potassium followed by altitude was also the most important constraining variable in weighing axis one in the ordination. In the sandy soil, plant-soil feedback effects were most strongly correlated with potassium.
Although most studies investigating abiotic plant-soil interactions have focused on nitrogen and phosphorus dynamics, in sandy soils with little clay content, potassium could be a limiting factor for plant growth [61,62].
In particular, a growth of forbs can be highly dependent on potassium [61] and hence potassium at least affects the distribution of these species. In the same way, the disturbance was the most important variable in weighting axis two. Disturbance affects the distribution of plant communities by hampering natural regeneration and seedling establishment in tropical forests [63]. Disturbance also favours the growth of herbaceous plant species by improving the availability of light conditions in the ground layer as it widens the canopy gap [64] and thus affects the distribution of communities with these species. An analysis of variance (ANOVA) performed to see any signi cant variation among the community types of Gerba Dima forest with respect to non-collinear signi cant environmental variables indicated that the ve community types differ signi cantly from each other with regard to EC and K. Similarly, result of Tukey's pair-wise comparison test indicates that community 4 and 1 differ signi cantly with respect to Disturbance and K while community 2 and 3 showed signi cant difference with respect to EC.

Conclusions
Description of the oristic diversity of species in the Gerba Dima forest revealed the presence of high species diversity and richness. Of the species recorded in this forest, 15 (8.3%) species were endemic to Ethiopia. However, the percentage of endemic species in the study forest is lower than the proportions generally expected in the Afromontane forest of Ethiopia and this is attributed to the low endemicity feature of forests in South-western Ethiopia. In this study, ve community types were identi ed and altitude was the major environmental variable in determining the community types. The existence of high species diversity and a number of endemic plant species in the study forest shows the potential of the area for biodiversity conservation. Thus, all Stakeholders including Oromia Forest and wildlife enterprise (OFWE) and the regional government should work to designate the forest as a biosphere reserve and being registered under UNESCO.

Availability of data and materials
We have also included part of the data used in this research and attached as Additional les 1 and 2.

Competing interests
The authors declare that they have no competing interests.

Funding
This study was supported by International Foundation for Science (Grant Number D/5481-1). The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Authors Contribution
All authors have made substantive intellectual contributions to this manuscript. AD is made substantial contributions to conception and design, or acquisition of data, or analysis and interpretation of data and also been involved in drafting the manuscript or revising it critically for important intellectual content. TS is also made substantial contributions to conception and design of data, or analysis and interpretation of data but not involved in data collection or acquisition of data. BW has been involved in drafting the manuscript or revising it critically for important intellectual content. All authors read and approved the nal manuscript.