Figure 4:

Tea mass loss and decomposition metrics showed little variation in response to environmental factors. Points represent plot-level averages of a) relative mass loss in the two litter types,b)  decomposition rate constant k , and c)stabilisation factor S . Trend lines and ribbons show mean and 95% confidence intervals of predicted linear relationships, respectively. Slope, intercept and p- values were extracted from individual sub-models within the Structural Equation Model (see Figure 5).
Environmental variation explained only a small amount of variation in our observed decomposition rate constant (k ) and stabilisation rate (S ). However, we observed a slightly higher explanatory power for microclimate (indirect vegetation/soil influence) compared to plant trait or soil-related explanatory variables (direct vegetation/soil influence; Figures 4b,c). Warmer plots had higherS (p < 0.01) and slightly (non-significant) higher k (p = 0.08), and S was slightly lower in moist soils (p = 0.15). However, neither leaf traits nor soil characteristics explained variation in decomposition metrics (Figures 4b,c). Also, no spatial patterns were visible for either decomposition variable (Figure S4).
Overall, our hypothesised system of topography, vegetation and soils, and microclimate explained little variation in decomposition, reflected in a low overall model fit of the SEM (Fisher’s C = 769.5,df = 106, p < 0.001; Table S3). Explanatory power was highest for leaf economic traits (R2= 0.45) and soil parameters (pH / texture: R2 = 0.28; nutrients: R2 = 0.27; GDD0: R2 = 0.38) but low for decomposition variables (Green Tea mass loss:R2 = 0.16; Rooibos Tea mass loss:R2 = 0.05; k :R2 = 0.08; S :R2 = 0.16; Figure 5). Moreover, SEM relationships between variables besides decomposition only partly reflected our expectations (Figures 1,5). Elevation and solar radiation were the most influential topographic predictors. Higher elevation significantly predicted lower vegetation cover, more acquisitive leaf traits, lower soil nutrient content, and lower soil temperatures (allp < 0.001), as well as higher silt and lower sand content in the soil (p < 0.05). Higher solar radiation was related to higher soil moisture (p < 0.01, with solar radiation being highly negatively correlated with slope (Pearson’s r = -0.93)), as well as to cooler soils, smaller plants, and more acquisitive leaf traits (all p < 0.05). In addition, higher TWI predicted higher soil pH and higher silt rather than sand content (p < 0.01) and higher soil nutrient concentration (p < 0.05). TPI did not predict any variation in vegetation, soil or microclimate variables (Figure 5). Only one of the tested relationships between vegetation and microclimate was significant, as higher vegetation cover predicted lower soil temperatures (p < 0.001). Soil texture did not predict any variation in soil moisture (Figure 5).