3.3. Relationship between topographic heterogeneity, soil type and fertility, human-influence factors and richness pattern
Of the total twenty-three variables, the explanatory variables of topographic heterogeneity, specifically ELER and ELER-SD, were the strongest predictors and showed a positive relationship to all richness groups (Table 1). ELER explained 32-36%, whereas ELER-SD explained 30-33% of the total variance in species richness. Most of the soil fertility and nutrient availability variables showed positive and stronger relationship more than soil texture variables, particularly carbon content (ORCDRC), available nitrogen (AN) and CEC were the second-best predictors of conifer richness, and they contributed 24-31%, 14-29%, and 21-26% respectively (Table 1).
Variables of soil texture type, particularly clay (10-27%) and sand (11-23%), contributed much higher than coarse and silt, which explained 8-12% and 2-7%, respectively. Individually, clay and silt soil types showed a negative relationship to all richness groups; while, sand and coarse fractions showed a positive relationship (Table 1). The contribution of the human-influence variable was much lower for all groups showing a negative relationship (Table 1). The proportion of variance using extracted principal components explained by the individual environmental and human-influence predictors showed the significant role of topographic heterogeneity and soil fertility in determining species richness (Fig. 2). Topographic heterogeneity was the strongest predictor of conifer richness followed by soil fertility and nutrient availability for all richness groups.
Hierarchical partitioning suggested that topographic heterogeneity and soil nutrient-fertility had the highest independent effects on all species richness, while they had the highest joint effect on endemic-threatened species richness (Fig. 3a and d). Soil type predictors showed a high joint effect for all species richness groups (Fig. 3). It is worth to note the collinearity between topographic heterogeneity and soil nutrient-fertility variables (see Table S4, Appendix A), which can influence the result interpretation. The combined models produced using stepwise regression (GLM) showed that ELER and soil available nitrogen were consistently significant predictors representing topographic-soil heterogeneity (Table 2). The variance inflation factor (VIF) of each predictor of the four combined models of richness groups was less than 5, which indicates insignificant multicollinearity between the predictors in the models. It is worth to note that when we excluded the human influence predictor (HII) from the models, the same combination of environmental predictors was resulted based on the lowest Akaike information criterion (AIC) indicating to the relevant collaborative interaction between environmental factors; meanwhile, the less importance of the human factor. The developed models moderately predicted the conifer richness for all groups.