This Perspective manuscript presents a framework for quantifying spatial insurance of functional traits., working at the community level. This involves identifying sources and sinks of particular functions, defined as areas of trait space. The sources represent communities that disproportionately insure their neighbouring communities, while sinks represent communities that disproportionately depend on their neighbouring communities. There seem to be many potential applications of this, several of these mentioned by the authors. The framework is explored using simulated data, and the manuscript and supplementary material also include examples and demonstrations of use of the framework with real data at different scales. Overall, in my opinion it is potentially a very good addition to the literature.

I do have some suggestions for improvement, set out below. But first, I think I should note that there is a lot going on in this manuscript, with a considerable amount of empirical results as well as the framework – resulting in a long manuscript even before the pages of supplementary material are considered. The empirical examples do help to illustrate the use of the framework, but this raises an issue in that there are some potentially important results reported in these examples, which get rather lost and buried, and do not even feature in the abstract. I am just noting this here, rather than asking for particular changes. It’s more for the authors to reflect on – for example whether they are potentially pre-empting an interesting empirical paper that they might write, based on the empirical demonstrations within this manuscript. The manuscript is a Perspective one, and rightly so because its main business is to present the framework. As a result, the empirical results are secondary, which is a bit of an issue if they in themselves are important.

The most interesting empirical result, for me, is reported in L414-417, for the global-extent bird analysis:
“We found that source cells were exposed to higher human footprint than intermediate and sink cells (Fig. 6a, Supporting information S3), the latter being found in areas with a higher proportion of protected areas (Fig. 6b, Supporting information S3).”
This seems like a potentially important finding, which is rather buried in the manuscript and not mentioned in the abstract. On the other hand, from a quick read of section S3 (all I could make time for!), this result is perhaps rather nuanced. Perhaps that is why the authors have not made much of it?

I found the Introduction to be good, mostly well written and setting the rest of the paper up nicely.

The methods are mostly clear and good, except that I found the functional radius to be quite opaque. The first mention of it is in the caption of Fig 2 (L199), where it is not defined, so it is quite mystifying there. Its importance is immediately clear, as demonstrated in that same caption: “Increasing the functional insurance radius can switch a sink community to a source” (L201). So it matters how it is determined, and that seems to be a matter left rather hanging by the manuscript. There are some suggestions for how to determine it, but none of them with convincing reasoning for why it is appropriate. The authors do recognise this problem in L482-4, and speculate about a way forward in the following sentences:
“Here, we used a threshold distance in the functional trait space (i.e., functional insurance radius), but this requires choosing it, somehow arbitrarily. An alternative could be to take it from a continuous point of view, for instance by incrementally varying the radius and recalculating functional insurance to obtain a distribution of insurance value, instead of a single value for a given radius. A kernel decreasing function of both spatial and functional distances, reflecting the declining potential for insurance with increasing dissimilarity and dispersal cost could also be explored (Mammola & Cardoso 2020).”
These do seem like appropriate suggestions, and one could argue that they should be brought into this manuscript, but here I refer back to how much is already going on in the manuscript. I think on balance it is fine to do what the authors have done and keep the simple but arbitrary approach for the purposes of demonstrating the framework, while indicating improvements that can come later. I am therefore not asking for changes, but do invite the authors to reflect a bit more on this.

In a similar vein, sources and sinks are categorised in the manuscript, in an arbitrary way, and this could be improved by a more ‘continuous’ approach. Again I invite the authors to reflect on this, potentially making changes, but I would not want to require this.

And similarly, functionally distinct species are categorised arbitrarily (L234-6: “We considered species with local functional distinctiveness greater than 90% of local species distinctiveness values to be functionally distinct; other species were considered as functionally common.”) Have the authors considered adapting the method to use all the information in the continuum, rather than this binning approach? And/or, if a binning approach is retained, consider cutting out some values in between common and distinct (e.g. >90% = distinct; <50% = common), to get a clearer distinction between them. This approach is used elsewhere in the manuscript, and could improve Fig 3.

The Figures are mostly well conceived, but the execution must improve. The main problem is the colour schemes, and I have to say that in 2025 I am quite shocked to see such poorly chosen colour schemes in a paper submitted to a good journal. It is almost impossible for me, with my red-green colour-blindness, to interpret several of the figures properly. And even when I showed the figures to my non-colour-blind wife, she struggled with some of it and commented that the colour schemes are not acceptable.

In particular, Fig 4 is awful with respect to the colours. When I first looked at Fig 4a, I saw lots of dots, all of the same colour, interpreting it simply as a map of the vegetation plots. Reading the caption then alerted me to the fact that there are both red and green dots. Zooming in, I can start to see some distinction, but struggle with the pixellation, and cannot interpret this map in the way that is intended by the authors. I also think I see at least one dot that is not red or green, which looks like the intermediate colour in the colour legend to the right of the map, but the caption says the intermediates are not shown on the map.

Fig 4d is also very hard for me in terms of colour scheme. First, I cannot tell the light blue from the grey - and my wife struggled with that, too, when I asked her to look. I see the thick red circles round 4 points, but she tells me that there are also apparently thin red circles round lots of other dots. I cannot pick these out, even when I zoom in. At normal zoom, I think I see lots of red-outlined dots towards the middle of the graph, but my wife tells me the outlines of these are not red – thus I get the opposite impression from this graph of what is actually being shown. This issue is related to what seem to be unexplained borders around some dots but not others. The overall impression is that there seem to be dots of different sizes, some with dark borders and some without, none of which I can interpret properly.

Fig 4b has coloured dots on a graph, but no legend on the Figure. The caption explains it, but there is no scale.

In Fig 5, the large map (5a) is horrible. When I first looked at it, I saw most of the terrestrial areas of the world in yellow and lots of patches of a darker colour. Only on reading the caption did I realise that it is actually two darker colours, red and green. I can hardly distinguish between these, even when zoomed in (except for the large patches), and overall cannot see any pattern in the red vs green on this map. My wife tells me that, for example, the deserts are mostly red, so my understanding is that I should be able to see pattern.

A FEW MINOR SPECIFICS

L73-5, sentence starting ‘Originally’ lacks citations, but these are surely necessary, and need adding.

L123-5, sentence starting ‘Indeed’: I found it hard to make sense of this sentence, and suggest that it is rewritten.

Fig 2 – the red circles are not easy for me to distinguish from the dots they surround, especially when those dots are green. However, in this figure it doesn’t matter much because the highlighted dots are considerably larger than all the others, so can be distinguished by size. But if the colour scheme is to be changed in later figures (see above), it would make sense to keep it consistent, so should also be changed in Fig 2.

L524: “overall, both source and sink habitats deserve particular conservation attention.” This needs unpacking. The authors can probably point to the paragraph starting L534, which does some of this. But the least intuitive part, for me, is why 'sink habitats deserve particular attention', and I did not find this explained. The authors also need to be careful not to end up concluding that all the habitats need particular attention!

Fig S2.1: While not the focus here, it would help interpretability if there were some indication of what the two axes represent.

Section S2 has no line numbers but 3rd sentence of ‘Computing functional insurance’ is “To measure insurance we used a functional insurance radius proportional to the median value of the global functional distance matrix (median = 1.43; radius = 0.46).” This needs explaining more! Proportional how? Why does a median of 1.43 translate into a radius of 0.46? And also the authors should explain what the units are here.

Section S3: Why is there no PCA, equivalent to the one in S2, here?