Quantifying the relationship linking the community-weighted means of plant traits and soil fertility.

Is it possible to generalize relationships between certain plant traits and soil fertility? In particular, are there quantitative relationships between community-weighted mean (CWM) trait values of leaf dry-matter content (LDMC), specific leaf area (SLA), plant height, and Grime's competitor-stress tolerator-ruderal (CSR) strategy scores and the generalized soil fertility, FG (i.e., the capacity of a soil to produce biomass when all nonsoil variables are held constant) that are generalizable across different species assemblages and geographical areas? We assessed FG in 21 sites in southern Quebec and 7 sites in southern France using a previously published method based on structural equation modeling. We then determined the CWM values of LDMC, SLA, plant height, and CSR scores in the 21 Quebec sites to obtain quantitative relationships between FG and these CWM traits. Using these relationships, we independently tested the generality of the trait-fertility relationships using data from French sites. The relationships between FG and the CWM traits were nonlinear, but displayed the expected qualitative trends as reported in the literature. In particular, the S score and CWM LDMC decreased with increasing soil fertility, and the R score and CWM SLA increased. CWM traits were more strongly correlated to measures of FG (r2 up to 0.48) than to measures of other soil characteristics (r2 up to 0.17 for nitrogen flux). Importantly, the independently tested French FG -trait relationships showed no significant deviations from these quantitative relationships. Further investigation is necessary to confirm if the same trend applies to other regions and or ecosystems.

The measurement and quantification of generalized gradients of soil fertility relevant to plant community ecology.

We propose an operational definition of soil "fertility" that is applicable to plant community ecology and develop a method of measuring and quantifying it, using structural equations modeling, that is generalizable to soils in different regions whose fertility has different causes. To do this, we used structural equation modeling (SEM). The measurement submodel predicts the latent "generalized fertility," FG , of a soil using four indicator variables: the relative growth rates of Festuca rubra, Trifolium pratense, Triticum aestivum, and Arabidopsis thaliana. The direct causes of FG in this study were the supply rates of NO3- , P, and K as well as three indirect causes consisting of three physical soil properties, but these can change between studies. The model was calibrated using 76 grassland soils from southern Quebec, Canada and independently tested using aboveground net primary productivity (NPP) of the natural vegetation over two growing seasons. Both the measurement submodel and the full SEM fit the data well. The FG values predicted 51% of the variance in NPP and were a better predictor than any other single variable, including the actual nutrient flux rates. Furthermore, this model can be applied to grassland soils anywhere because of its modular nature in which the causes and effects of soil fertility are clearly separated.