Functional traits are increasingly being used to quantify ecosystem responses and services, and to better classify species. Typically, these studies rely on trait means or community level trait averages to quantify functional change. Recent research, however, suggests that intraspecific variation in functional traits may play a significant role in ecosystem services and response to change. High trait variation in a species may alter its functional role within an ecosystem, or provide a buffer against ecosystem change. Furthermore, ecosystem services and responses may be over- or underestimated when intraspecific trait variation is ignored. Numerous methods for quantifying intraspecific variation have been proposed, but have not been critically compared. We evaluated the consistency and usefulness of these methods to characterize two commonly studied plant traits, specific leaf area (SLA) and plant diameter. Traits were collected for 8 populations (165 plants) of the common and widespread weed Hypochaeris radicata. We used several methods to quantify intraspecific variation within and between the populations, including standard deviation of the mean (SD), coefficients of variation (CV), variance partitioning using ANOVA and linear modeling, and a permutational method for assessing dispersion (permDISP).
Results/Conclusions
Intra- and inter-population patterns varied between traits and amongst quantification methods. Overall, there was a 25-fold difference in SLA (0.018 to 0.461 cm2/mg) and a 23-fold difference in diameter (59.1 to 2.5 cm) among individuals. Populations differed significantly in average SLA (P = 0.0001); inter-population differences accounted for 44% of that variation. Populations also differed from published trait values; measured averages ranged from 0.053 to 0.226 cm2/mg, while means reported in the LEDA database ranged from 0.196 to 0.326 cm2/mg. The CV for SLA ranged from 18 to 59% within populations, and was 53% across all populations, but permDISP indicated that dispersion did not differ among populations. Average plant diameter also differed significantly (P =0.0001) with inter-population differences accounting for 55% of the in this trait. The CV for diameter ranged from 26% to 45% within populations, and 61% across all populations. Populations differed in dispersion around the mean (permDISP; P = 0.001). The magnitude of variation present in these populations, as well as the conflicting quantifications of variation (e.g. large differences in CV for SLA, but no significant detected by permDISP) indicate the need for a more accurate method for quantifying variation. We suggest probability and distribution-based methods for quantifying and incorporating intraspecific variation into community level analyses and predictive modeling.