Diversity Inuences Tree Aboveground Carbon Stocks Through Functional Diversity and Functional Dominance in the Dry Evergreen Afromontane Forest of Hararghe Highland, Southeast Ethiopia

Background: The relationship between biodiversity and ecosystem function has increasingly been debated as the key ecological issues behind ecosystem service provision. Still, many experimental and theoritical based studies have reported inconsistent patterns of biodiversity and ecosystem function relationships, supporting either niche complementarity or selection effect hypothesis. In this study, aboveground carbon (AGC) stock used as a proxy for ecosystem function and examined its relationship with species diversity, through functional diversity and functional dominance. It is hypothesized that (i) diversity inuences AGC through functional diversity and functional dominance effects; and (ii) effects of diversity on AGC would be parted for both functional dominance and functional diversity. Community weight mean (CWM) of functional traits (wood density, specic leaf area, and maximum plant height) was calculated to assess functional dominance (selection effects). As for functional diversity (complementarity effects), multi-trait functional diversity (selection effects) indices were computed. The rst hypothesis was tested using structural equation modeling. For the second hypothesis, the effects of environmental variables such as slope, aspect and elevation were tested rst, and separate linear mixed effects models were tted afterward for functional diversity, functional dominance, and both. Results: Results revealed that tree aboveground carbon varied signicantly along the slope gradient. Species diversity (richness) had a positive relationship with aboveground carbon, even when elevation effects were considered. As predicted, diversity effects on aboveground carbon were mediated through functional diversity and functional dominance, suggesting that both the niche complementarity and the selection effects are not exclusively affecting carbon stock. However, the effects were greater for functional diversity than for functional dominance. Furthermore, functional dominance effects were strongly transmitted by CWM of maximum plant height, reecting the importance of forest vertical stratication of diversity and carbon relationship. We therefore argue for stronger complementary effects that would be induced also by complementary light use eciency of tree and species growing in the understory layer. Conclusions: Species diversity (richness) inuences carbon stock through functional diversity and functional dominance. Both the niche complementarity and selection hypotheses are important predictors of carbon stock in the study forest. as we hypothesized that the diversity effects be mediated through both functional diversity and functional dominance. Therefore, we tested the indirect and direct effects of diversity (species richness) on aboveground carbon. Two separate structural equation models were specied representing (a) full mediation: postulate that diversity effects are fully mediated by functional diversity and dominance metrics; and (b) partial mediation: present that there are both direct and indirect diversity effects through functional diversity and functional dominance metrics on aboveground carbon. Due to the presence of multiple measures of functional diversity and functional dominance, stepwise selection techniques were used to select the most relevant functional diversity and functional dominance metrics for the above ground carbon data.

The niche complementarity and selection effect hypotheses (Tilman et al. 1997;Grime 1998) are the two newly proposed ecological hypotheses commonly used to elucidate the role of plant diversity in ecosystem dynamics, processes and functions. The selection effect hypothesis postulates that a higher probability of occurrence of dominant species or traits (as a result of competition) would in uence the functioning of the ecosystem in a diverse group of species (Grime 1998). The niche complementary effect hypothesis states that increasing diversity would promote a wider range of functional traits and provide opportunities for species to use the available resources e ciently, thereby increasing ecosystem function and less competition (Tilman et al. 1997). The niche complementary hypothesis is usually inferred when explaining higher production of biomass and productivity in highly diverse ecosystems (Mensah et al. 2016). Higher biomass, however, may also result from dominant species with strong resource responses and/or strong ecosystem impacts that refers to selection effects (Reich et al. 2001).
The relationship between diversity and ecosystem function (productivity) is inconsistent in whether these are positive, negative or no relationship. For instance, Li  Understanding whether diversity in uences on ecosystem function are more likely mediated through functional diversity than functional dominance, or vice versa, will bring substantial insights into which mechanism is more relevant.
Few studies have addressed the relationships between diversity and ecosystem function in natural dry evergreen Afromontane tropical forests. Using aboveground tree carbon data in a dry evergreen Afromontane forest in Southeast Ethiopia, we examined the relationship between diversity and aboveground carbon stocks through the effects of functional diversity and functional dominance. We hypothesized that (i) diversity in uences tree aboveground carbon stock through both functional diversity and functional dominance effects. However, there are notions that diversity and carbon relationships can be caused by covarying environmental factors (Cavanaugh et al. 2014;Ouyang et al. 2016). Therefore, altitude, slope and aspect were considered as the most environmental factor in this forest and tested for their in uences on tree aboveground carbon stocks. In addition, while accounting for signi cant environmental factors effects, we also hypothesized that (ii) effect of the diversity of tree aboveground carbon stock would be parted for both functional dominance and functional diversity.

Study area
The study was conducted in the Dindin dry evergreen Afromontane forest in the Hararghe highlands of southeast Ethiopia. The geographical location of the study site lies between 40 o 10'40'' to 40 o 18'50'' E and 8 o 33'0'' to 8 o 40'40'' N with elevation ranges between 2,124 and 3,069 m a.s.l. and situated around 336 km southeast of Addis Ababa (Fig.1). Due to the lack of long term climatic data for the study site, the climate estimator software tool, New LocClim (fao 2005; Grieser et al. 2006) was used to produce long term monthly rainfall and temperature data for the study site. The mean annual temperature (MAT) and mean annual precipitation (MAP) at study site was estimated to be 25.6 o C and 804 mm/year, respectively.
Precipitation at study site has a bimodal distribution pattern with a long rainy season lasting from June to October and a short rainy season from April to May (Fig. 2). The soils of study area developed for a wide range of parent materials, including volcanic and mixed limestone and sandstone over a For the purpose of this study, trees are one stemmed woody plants with heights ≥ 5 m. The name of the species was identi ed in the eld whenever possible, but in doubtful cases vouchers were collected and pressed for further identi cation and con rmation at the Addis Ababa University herbarium. Environmental variables such as elevation, slope, aspect and geographical location of each plot were recorded. The elevation was measured by using a GPS garmin; whereas, the slope and aspect were measured by using a compass meter. Where AGB is the aboveground biomass of trees (kg), ρ is the speci c wood density (g cm -3 ), D is the stem diameter at breast height (cm), and H is the total height of trees (m).

Estimation of aboveground biomass carbon
Aboveground biomass carbon was determined by assuming 47% of AGB containing carbon (IPCC 2006). The total AGB carbon for each plot, were calculated by summing up AGB carbon for all species. Carbon stocks were determined for each plot and then extrapolated to tonnes per hectare.

Functional diversity and dominance metrics
Taxonomic diversity was used to measure diversity, at each plot. Species richness was used to characterize the taxonomic diversity (Magurran 2004). Species richness at plot level referring to the number of different species counted in each plot. To quantify functional diversity and functional dominance, three functional traits that are relevant to the ecosystem function of interest (i.e., biomass and carbon storage) were considered, because carbon storage is strongly dependent on wood and foliage structures, in this study, therefore, traits such as wood density (WD), speci c leaf area (SLA), and maximum plant height (PHm) was considered. Data on wood density were extracted from the Global Wood Density Database (Zanne et al. 2009). In case multiple values were available for a single species, the average wood density was used. SLA and PHm were extracted from the TRY database (Kattge et al. 2011;Maire et al. 2015). When the data for a particular species was missing, the average genus wood density and SLA were used.

Statistical analysis Structural equation modeling
Structural equation modeling (SEM) is a powerful, multivariate statistical model found increasingly in ecological studies to test and evaluate multivariate causal relationships. SEMs differ from other modeling techniques as they test the direct and indirect effects on presupposed causal relationships (Fan et al. 2016). This is particularly important, as we hypothesized that the diversity effects would be mediated through both functional diversity and functional dominance. Therefore, we tested the indirect and direct effects of diversity (species richness) on aboveground carbon. Two separate structural equation models were speci ed representing (a) full mediation: postulate that diversity effects are fully mediated by functional diversity and dominance metrics; and (b) partial mediation: present that there are both direct and indirect diversity effects through functional diversity and functional dominance metrics on aboveground carbon. Due to the presence of multiple measures of functional diversity and functional dominance, stepwise selection techniques were used to select the most relevant functional diversity and functional dominance metrics for the above ground carbon data.
The overall t of the SEMs model was evaluated using Goodness of t index (GFI), value should be close to 0.95 or higher (Hu and Bentler 1999), Chi-square (χ2) and standardized root mean residual (SRMR ≤ 0.080). The standardized coe cients were used to make direct comparisons across paths (Grace and Bollen 2005). SEMs were tted in the R statistical software package, using the "sem" function in "lavaan" package and latter, output graph was visualized by using "semPlot" package again in R.

Linear mixed effects models
Linear mixed-effects models (LMMs) are an important part of statistical models that can be used to analyze correlated data (Galecki and Burzykowski 2013). Prior to the mixed-effects modeling, the signi cant environmental factors and species richness effects on aboveground carbon stock were tested. Environmental factors are expected to have effects on plant composition, growth, and survival (Mensah et al. 2016;Zhang et al. 2014) and for this reason for standing aboveground biomass and carbon stocks.
Here, the focus was given to the factors that are determinant and quanti able in the area, that is, topography (elevation, slope and aspect) (Geldenhuys 2002). The effects of elevation, slope and aspect of aboveground carbon stock were tested using simple linear models in R. In case, these variables showed no signi cant effect on the carbon stock and further analyses was performed and multiple linear regressions were used to test their effects on aboveground carbon storage. For both simple and multiple linear models, Shapiro-Wilk tests (W = 0.985, p-value = 0.739) were used to check for the normality of the square root transformed aboveground carbon data and of the residuals. Further, Breusch-Pagan tests, value in ation factor (VIF) and Durbin-Watson statistics were used to test for heteroskedasticity, multicolinearity and autocorrelation between residuals, respectively.The effect of each diversity component (i.e., functional diversity and functional dominance) on carbon storage was evaluated by tting separate linear mixed-effects models (Zuur et al. 2009). Elevation was considered as random factors and each measure of functional diversity (i.e., FRic, FEve, FDiv, and FDis) and functional dominance (i.e., CWM of WD, SLA, and PHm) as xed effects. Next, mixed effects models were tted to evaluate the individual effect of each functional diversity and functional dominance metric; the combined effects of functional diversity metrics; the combined effects of functional dominance metrics; and combined effects of functional diversity and functional dominance metrics. Further, the effect of each diversity component ( xed effects) on above ground carbon storage was determined using a mixed effects model (Bates et al. 2015) with 'Type III analysis of variance with Satterthwaite's method'. To run mixed effects model package 'lme4' (Bates et al. 2015) was used in R and variables were selected by 'backward selection' using 'cAIC4'package (Saefken et al. 2018) in R. The signi cant effect of xed factors was determined using the "lmer" function of the "lmerTest" package (Kuznetsova et al. 2017) also in R. The signi cance of the random effects was determined using likelihood ratio (LR) test, again in the package "lmerTest". The performance of tted models was evaluated based on the t statistics such as Akaike information criterion (AIC) (Akaike 1974) and the marginal R square, which shows the proportion of variance explained by xed effects (Nakagawa and Schielzeth 2013).

Diversity effects mediated via functional diversity and functional dominance
The results of the structural equation models tted to test the mediated effects of diversity (species richness) on AGC, through functional diversity and functional dominance is summarized in Table 1 and Fig. 3(a). The chi-square value for the "full mediation" model was 24.63 with 14 degrees of freedom and a p value of 0.026 and standardized root mean square residual was 0.076, indicating slight good t.
In the "full mediation" model, species richness revealed a signi cant positive direct effect on the functional richness (R 2 = 0.580; ß = 0.76; p <.001; Table 1), but showed no signi cant positive effect on the AGC (ß = 0.17; p =.374; Table 1). There was also a signi cant positive effect of species richness on functional evenness (R 2 =.194; ß = -0.44; p <. 001; Table 1); the latter, however, exhibited non signi cant positive effect on the AGC. Further, species richness showed a positive signi cant effect on functional dispersion (R 2 = 0.157; ß = 0.40; p <.001; Table 1), however, non signi cant positive effect on AGC was found. There was positive and non signi cant of species richness on functional divergence (ß = 0.12; P = .356; Table 1), which latter, indicates negative and slightly signi cant effect on the AGC (ß = -0.23; P =.084; Table 1) at 0.1, which would suggest that the mediated effects of species richness were mediated through functional divergence only. In addition, no signi cant correlation between functional richness and functional evenness (ß = 0.29; p = .090; Table 1) was found, which would suggest that the mediated effects of species richness were transmitted by both functional richness and functional evenness.
Out of the three functional dominance metrics, the CWM of SLA and CWM of wood density did not retain any signi cant path. Only the CWM of maximum plant height showed signi cant responses to species richness (R 2 = 0.07; ß = 0.26; p = .040; Table 1), but showed negative and non signi cant effect on the AGC (ß = -0.17; p = .612; Table 1). The signi cant residual correlation of CWM of maximum plant height with CWM of wood density (ß = 0.45; p = .003 and CWM of SLA (ß = 0.45; p = .003; Table 1) con rms that the mediated effects of species richness are transmitted via CWM of maximum plant height only. The partial mediation model was tted by only adding a direct path from species richness to AGC to the "full mediation" model. The second model (partial mediation) had chi-square value of 19.58 (DF = 12; p =.076, Fig. 3(b) and Table 1), indicating good t and absence of signi cant deviations between data and model. There are similarities between the two models in terms of signi cance and no signi cant paths (Table 1). However, the "partial mediation" model showed slightly better ts (GFI = 0.947; R2 = 0.22; SRMR = 0.071) than the "full mediation" model (GFI = 0.928; R2 = 0.14; SRMR = 0.076). The signi cant in uence of added causal path from species richness to AGC was found, suggesting an existing true direct effect of diversity on the AGC. Both models recommend that the effects of species richness on aboveground carbon be mediated via functional diversity and functional dominance, which support for both complementarity hypothesis and selection effect hypothesis. Effect of environmental variables on aboveground carbon stocks There were signi cant effects of the environmental variables, particularly the slope which explained 7.41% of the variation of the aboveground carbon (Table 2). However, altitude and aspect gradients showed non-signi cant effect on the aboveground carbon (p = .168 and .701, respectively; Table 2). Additionally, when accounting for the effects of the altitude, signi cant effect of species richness on the aboveground carbon was found (ß = 0.78; p = .034; Table 2).  (Table 3). Both functional divergence (ß = -10.64, p =.485) and functional evenness (ß = 3.82, p =.524) had no signi cant negative and positive effects on the AGC (Table 3). With respect to the combined effects of functional diversity metrics, the signi cant effects of FDis and FDiv were found after backward selection for the nal model ( Table 3). The effects of functional diversity on AGC were therefore shown by a signi cant positive effect of functional dispersion (ß = 21.36; p = <.001; Table 3) and a signi cant negative effect of functional divergence (ß = -39.18; p = .003; Table 3). Both functional dispersion and functional divergence explained 29% of the variance of AGC. Out of the three functional dominance metrics used in this study only CWM of WD showed negative signi cant effects on the AGC (Table 4). However, CWM of SLA (ß = -0.09; p = .250) and CWM of PHm ((ß = 0.28; p = .124) revealed negative and positive with non-signi cant effects on the AGC (Table 4).  Table 4).
The output of mixed effects models of functional diversity and functional dominance showed that the marginal R square (variance explained by xed factors) in the diversity and AGC relationship was greater for functional diversity (29%) than for functional dominance (10.1%) ( Table 5). While analyzing the combined effect of functional diversity and functional dominance on AGC, the result revealed that 22.4% of the deviations of AGC were explained by signi cant effects of functional dispersion and CWM of wood density (Table 5), indicating variation of AGC greater for functional diversity than combined model effects.

Discussion
Species diversity promotes aboveground carbon stocks Unlike altitude and aspect, slope showed signi cant effect, and accounted for 5.51% of aboveground carbon variance, evidencing that variations in carbon stocks can result from topological constraints, particularly difference in slope. Consistent with our results, slope has been identi ed as an important environmental gradient that affects tree carbon (de Castilho et al. 2006;Chave et al. 2003). This is because, aboveground carbon is inherently related to wood and biomass production, the effect of slope can be seen as prior impacts of environment on availability of resources (Luizao et al. 2004), which in turn in uence forest dynamics. For instance, steeper slope will speed up nutrients, water runoff, and constrain trees and will also favor highly water and nutrient e cient species against others. Taking this into account, it follows that tree growth and biomass production can be potentially declined at higher . Accordingly, functional diversity and dominance metrics were also examined in this study. While most of these studies tended to compare the relative effects of species richness and other biodiversity components on aboveground carbon stocks, by assuming that these effects were mediated via functional diversity and functional dominance. In this study the structural equation modeling results con rm this assumption and this is, therefore, supports the need to explore beyond species richness to elucidate the mechanisms that drive relationship between diversity and productivity. The results further support the hypothesis that both niche complementarity and selection Both these functional characteristics may measure niche complementarity, and therefore increment of ecosystem processes by functional trait variety. The functional would increase carbon stock because species with different traits would differ in resource use, and would more e ciently use the resources available within the community for higher growth and productivity, indicating the importance of niche complementarity effects in facilitating ecosystem processes (Finegan et al. 2015). Unlike functional richness, functional evenness and functional dispersion, functional divergence did not show any relationship with species richness and show no signi cant in uence on aboveground carbon. According to Laliberte and Legendre (2010), the functional dispersion is the mean distance of each species, weighted by its relative abundances, to the centroid of all species in a community. Therefore, both functional richness and functional dispersion relate to the niche space or volumes of niche space, and functional diversity as measured here could re ect some form of "niche differentiation" (Carroll et al. 2011 CWM of wood density indicates that low wood density species grow faster and expect to store more biomass; therefore, it recommends that conserving and planting low wood density species would likely help to increase the carbon stock. In analysis of combined effects of functional dominance metrics only CWM of speci c leaf area and of maximum plant height were retained in the nal model, with maximum plant height being the signi cant predictor. This is most likely tree height is a key factor for species-speci c or multispecies biomass regressions. Ali (2015) suggest that strong dominance by tall and conservative species, rather than a set of coexisting species with diverse heights and exploitative nature, results in greatest carbon stocks in natural forest ecosystems. Therefore, the positive and signi cant relationship between CWM of maximum plant height and carbon stocks indicates the potential importance of characteristics of dominant and adult trees for ecosystem functioning and productivity, thus supporting the selection effects hypothesis.
Functional diversity and functional dominance partly effects aboveground carbon stocks In testing the proportion of variation explained by the selection effect and niche complementarity mechanisms, functional diversity explained more variation of aboveground carbon than functional dominance (Tables 3 and 4). Unfortunately, this rejects the second hypothesis of this study, and suggests that niche complementarity effects appear to be more important than selection effects. This nding consistent with study by Mensah et al. (2016) supports that functional diversity results in more variance in aboveground carbon than selection effects. In contrary, study by Finegan et al. (2015), Ruiz-Jaen and Potvin's (2011) revealed that selection effects more important for the aboveground biomass and carbon storage in tropical forests. The reason for this discrepancy is that in this study, functional dominance metrics (community weight mean of functional traits) were estimated using species relative abundance, but a study by Ruiz-Jaen and Potvin (2011)  In examining joint effects, the present nding supports the thought that these two hypotheses (niche complementarity and selection effects) are not mutually exclusive, and can contribute to ecosystem functioning. Previous evidence of both complementarity and selection effects on ecosystem function suggests they can also contribute at different proportions at different times of ecosystem services (Fargione et al. 2007). Both complementarity and selection effects mutually promote species coexistence. As reason out by Mensah et al. (2016), these two hypotheses could even be the outcome of interactions of the "relative tness differences" and the "niche differences", whereby some species' populations could be suppressed by dominant competitors, to allow effective utilization of the available resources. In this study the selection effects are strongly mediated through speci c maximum plant height, which indicates the effect of dominant species and suggests a possible competitive exclusion in terms of utilization of resources (refer SEM e.g., sun light).

Conclusion
This study examined the relationship between diversity and carbon stock in dry evergreen Afromontane forest of Southeast Ethiopia and showed that taxonomic diversity (species richness) predicts carbon storage through functional diversity and functional dominance. Further, the study noted that both the niche complementarity and selection hypotheses are important for carbon Stock. On the other hand, carbon stock variations explained by functional diversity (niche complementarity effects) were greater than functional dominance effects (selection effects). Additionally, functional dominance effects were strongly transmitted through the CWM of maximum plant height, indicating the importance of forest vertical strati cation for diversity and carbon relationship. Therefore, complementary effects would be induced also by complementary light use e ciency of species and trees growing in the understory layer. This study recommends that future research on the relation between diversity and forest carbon to be

Declarations
Ethics approval and consent to participate Not applicable.

Consent for publication
Not applicable.

Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Competing interests
We have no competing interests.

Funding
Funding for this research was obtained from Oda Bultum University, Ethiopia.