I think I’ll learn more from you than you did from me 😅
Congratulations, Kerry!

with @stefanomammola.bsky.social, @gracoroza.bsky.social, Gianluigi Ottaviani and Francesco de Bello

Too many metrics, I agree! But accumulation curves make sense only for metrics that have set monotonicity/concavity. Divergence or evenness do not meet those properties, so the z slope loses meaning. We still need to match each metric to the facet we want to study.

2. I think they are fundamentally different aspects; a single number wont suffice. For example, some aspects are related to different aspects of the raw distribution of traits (richness, divergence, evennes), but other aspects require species identities (redundancy). But good luck with the search!

1. Yes, papers should report different indices, carefully chosen so that they are suitable for your question.

I would say that the field nowadays treats those criteria as facet-specific (richness, divergence, evenness) rather than universal.

My main concern is that the paper frames the 14 requirements as must-haves for any reliable FD metric, then counts how many each index meets. That creates the impression that indices are flawed.

So, if functional structure (or functional diversity if you prefer) is a distribution, you cannot simply characterize it with a single number, like you cant characterize a probabilistic distribution only with the mean, or the variance, or the kurtosis

The two aspects of extinction are relevant, so you need to estimate both things simultaneously (among other things) to understand well what is going on with functional structure (the distribution of species in the trait space)

But if what you are losing are redundant species, then divergence indices will increase, which matches their expected behaviour (you end up with communities with fewer but more unique species).

In the scenarios related to extinctions (7-8), you expect functional diversity to decrease. This expectation matches the behaviour of richness indices, which work well.

In your ecosystem scenarios, you test different things. For example in scenarios 2, 3 & 4 you are essentially increasing divergence, and this is well detected by divergence indices, but not by richness indices.

And so on. At the end, you can see from your tests that thare are sets of indices that tend to respond in similar ways (like the two sets I mentioned above). And these are the ways one would expect them to respond according to their original conception.

Indices of functional richness (FaD, mFaD, FD) should be insensitive to addition of redundant species and to changes in abundances, an only change when something modifies the boundaries of the distribution of species in the trait space.

Hi Adji! It was just an example. What I mean is that not all criteria should be fulfilled at once for an index to be useful. For example, dispersion indices (Rao, FDiv, FDis) by definition should not fulfil set monotonicity, because they reflect the average dissimilarity among components (species).

You would not judge a hammer by how well it drives a screw; likewise, do not blame a spread metric for failing a richness test. Functional diversity is multifaceted and needs a toolbox, not a universal index.

👏👏👏