Why trait-based ecology is full of surprises

Noémie Pichon discusses her recent article: ‘Intraspecific trait changes have large impacts on community functional composition but do not affect ecosystem function‘. Find out more about this new research into intraspecific trait variation and its importance for understanding community functional composition.

Trait variation is as large within as between species, but it doesn’t mean the same for ecosystem functioning. This work is the last chapter of my PhD thesis. Although this correctly implies that it was written with slightly more time pressure than the first two, I’m particularly fond of it, and I especially like the way the discussion turned out. I think I liked it because I didn’t expect the results.

Here is the background: plant functional traits, such as specific leaf area (SLA) and leaf dry matter content (LDMC), are used as proxies for growth strategies. For instance, high SLA and low LDMC species are rather on the acquisitive/fast side of the leaf economics spectrum, whereas low SLA and high LDMC species are on the conservative/slow side. These plant traits are used sometimes to test how the plant community responds to its environment, for example community mean SLA increases with nitrogen enrichment. Sometimes they are used to test the effect of species strategies on an ecosystem, e.g. high mean SLA communities produce more aboveground biomass. However, in both cases, to calculate trait responses or effects, we rarely consider the amount of variation within species. This is because extensive trait collecting is time and resource consuming, and because we believed, until recently, that intraspecific variation was smaller than interspecific variation, and therefore did not matter much in calculating the response or the effect of community mean trait values. However, more and more studies have shown that the amount of intraspecific variation in functional traits could be as large as interspecific variation.

This is where we started the journey with this paper: how much do plant communities respond to environmental condition, due to a shift in traits relative abundance, or due to a shift in trait values, i.e. does mean SLA increase because nitrogen enrichment favours the establishment of high SLA species, or because the species are producing leaves with a higher SLA? And how do these changes translate into biomass production?

*Measure of species percentage cover on the PaNDiv experiment. Picture: Hugo Vincent.*

I started to doubt about the importance of intraspecific variation for biomass production after calculating the effect of mean SLA on biomass, with three different ways to calculate community SLA, in a similar way to Roscher et al. (2018). These were: one with equally weighted monoculture trait values, one with abundance weighted monoculture trait values, and one with abundance weighted plot-specific trait values. After many hours checking for possible mistakes in the data or in the code, I had to admit that, just like in the Jena experiment, the second model was better at explaining biomass production: intraspecific variation did not matter.

*Sky view of the PaNDiv experiment. Picture: Hugo Vincent.*

Except that it was not the whole truth. We explored this a bit further. With the PaNDiv experiment, we had the perfect setup to create variation in traits, both within and between species. We had gradients in species richness and functional composition, but also a nitrogen enrichment and fungicide spraying treatment, in a full-factorial design. After filling the office’s white board with squares of different sizes, formulae, and several structural equation models, we found that intraspecific variation in SLA and in LDMC was, in fact, large. Larger than expected. Within a large initial trait gradient, it explained as much variation in community weighted mean traits as the shift in traits due to species relative abundance. It means that to characterise a community functional composition, and understand how it changes in different environmental conditions, we need to take intraspecific variation into account. However, this large shift in species traits did not translate into ecosystem functioning, suggesting that effect and response traits can be decoupled at certain scales. This is why including intraspecific variation to predict biomass was worse than using monoculture traits.