Spatial variations of leaf trichomes and their relationships with climatic factors across Eastern Asia
The mean value of trichome density in situ was 459.78 trichome mm-2 within the range of from 325.79 to 552.38 trichome mm-2, and the coefficient of variations of trichome density was 0.16 (Fig. 1). The first two of PCA axes of 3 geographic factors explained 57% and 27% of the total variation in geographic gradients of 44 populations across eastern Asia (Fig. 2). The first axis of PCA (GeoPC1) generally explained longitude and elevation, and the second axis (GeoPC2) mainly related to latitude. The PCA axes of 5 environmental factors accounted for 72% and 16% of the total variation in climate. The first axis of PCA (ClimPC1) mainly explained the variability of precipitation and temperature-related climatic factors, including MAP, AI, and MAT. The second axis of PCA (ClimPC2) best explained the variability of solar radiation (i.e. MMSR).
The piecewise SEM explained 36% of the spatial variation in trichome density. GeoPC1 and GeoPC2 showed indirect effects on trichome density through ClimPC1 and other leaf functional traits, including LMA, SD, SS, and VD. ClimPC1 had both direct and indirect effects on trichome density through four pathways (Fig. 2): one direct from ClimPC1 to trichome density (path coefficient 0.21), three indirect including from ClimPC1 via LMA to trichome density (path coefficient 0.13, 0.64×0.20), from ClimPC1 via SD to trichome density (path coefficient 0.06, 0.17×0.35), and from ClimPC1 through LMA and SD to trichome density (path coefficient 0.08,0.64×0.37×0.35).
Moreover, leaf functional traits could have both direct and indirect effects on trichome density. We found that LMA, SD, and SS had direct effects on trichome density (path coefficient 0.20, 0.35, and -0.15, respectively), whereas LMA and VD had indirect effects on trichome density through SD (path coefficient 0.13, 0.37×0.35) and SS (path coefficient 0.05, -0.31×-0.15), respectively (Fig. 2). ClimPC2 and LA had no significant effects on trichome density resulting in the removal of these pathways from the model. In addition, we further evaluated the correlation of trichome density with climatic factors in the ClimPC1, showing leaf trichome density was positively correlated with AI and negatively with MAT and MAP (Fig. 3).