Analyze resilience indices?

[teixeirak]

R1 and R4 both questioned why we don't consider resilience indices. I think we can easily argue that resilience is a different issue, and we wouldn't expect the same set of mechanisms to apply. In fact, I think this would be tricky in that it would require bringing in a whole new set of theory, so I'd lean against this.

On the other hand, it could be interesting. Also, note that R4's comment almost implies that considering resilience would reduce the need too use some other way of quantifying Rt (issue #92), although it doesn't change that in my mind.

R1 says this:

"It is important to better discuss the proper use of Rt, particularly when comparing single drought years, to assess tree resistance to drought. This metric has been much used lately as one of the so-called resilience indices (I wonder why other Rs indices where not included in the manuscript). In the paper, the authors use Rt similarly to most studies in the literature, which is fine. Yet, Rt (and particularly the other resilience indices based on radial growth) does not only depend on the individual response to drought events, but also on stand dynamics around target trees. For this reason, if stands around trees are not comparable or show changes after some disturbance (e.g. the drought events analysed) differently, this may bias comparisons of Resilience indices (although it is true that Rt should be less influenced than the other indices). The problem is that there is not such stand competition data for old droughts. This does not invalidate the study but I think the authors should discuss the issue, particularly for the analysis and comparisons of single droughts. I would suggest including also other Resilience indices, not only Rt, to enrich the discussion."

R4 says this:

"[Rt] has also advantages, particularly when applied in a resilience framework (as in the original Lloret et al. (2011)). However, if you focus specificaly on resistance it seems that directly modelling the growth reduction caused by the drought (e.g., using intervention time series analysis or other methods) would be a more robust approach. At the very least, if you continue using your simpler approach I would strongly suggest including some measure of pre-drought growth in the models to account for differences in growth dynamics among species. It would also nice to clarify why you do not assess recovery and resilience, unlike Lloret et al. (2011)."

[teixeirak]

I just emailed Neil & Alan about this.

[teixeirak]

from Alan:

An approach to respond to the suggestions of Reviewer 1 and 4 to look at other resilience metrics (recovery, resilience, relative resilience), would be to refer to DeSoto et al. 2020, which is one of the papers given in the links provided by Reviewer 3. DeSoto et al. used an extensive dataset of trees sampled after severe droughts and compared the responses to previous, less-severe droughts between trees that survived and trees that were killed by the severe drought. In angiosperms, trees that died from a severe drought had lower resistance to previous droughts, compared to trees that survived the severe drought. In gymnosperms, trees that died during severe droughts had lower recovery following previous droughts, compared to trees that survived the severe drought. These findings would argue that resistance is a more important variable for angiosperms. You can always calculate the other metrics and compare the results, but that would be much too long to include in the manuscript. Maybe it could go in an appendix instead.

[teixeirak]

from me:

Regarding the resilience metrics, I’m curious what we’d get, but the physiological mechanisms would be completely different, so dealing with resilience would require new hypotheses, etc. I’m leaning towards making the argument that that’s beyond the scope of the analysis. The DeSoto paper may be a good citation to further support the focus on resistance (thanks for that!).

[teixeirak]

from Neil:

My gut feeling is that resilience metrics will be too stochastic and unlikely to work well. More fundamentally, we wouldn’t expect the same mechanisms to apply. I think we can just make this argument and skip the other metrics. However, if Ian wants to try and see what happens, I’d be curious to see what we get.

[teixeirak]

from Neil:

Yes, the resilience metrics are really played out and have a few short comings, mostly, the conditions before matter to metric outcomes. It is related to the issues at hand here, like that the conditions before the 60s drought are so different from others.

[teixeirak]

My conclusion: @mcgregorian1, if you have time on your hands (ha!) and want to test the resilience metrics, go for it. Otherwise, we just make the argument that they're not so relevant/ outside scope of this project.

[mcgregorian1]

We technically get all of the resilience metrics when we run the initial functions (since they were developed by Lloret et al, they give all outputs) - but from early on I only focused on extracting / manipulating the resistance. I think it should be somewhat easy to look at each of the others but I say this without yet going through the full code to see what this would entail.

At the very least if we think it would be too much anyways (or it's more involved), all the info is there if you or another intern wants to look into this.

[mcgregorian1]

I can have this be on my to-do list, but I agree in that we can move past this question by referencing the DeSoto paper for a rationale.

[teixeirak]

Let’s just drop this. I feel that we have a very valid argument not to do it.

[teixeirak]

@mcgregorian1, I will write a response to the reviewers about this issue. Consider it resolved, for your part.

[teixeirak]

This is mostly done, but we may want to adjust wording/ add refs in new sentence in this paragraph, particularly relating to item 3. I emailed Alan and Neil to get their expertise on this.

[teixeirak]

From Neil:

Hi Krista,

I hadn’t thought about Rt in that way. I really have no thoughts on it, honestly. I think points 1 & 2 are likely strong enough.

Neil

[teixeirak]

I agree, and updated the text accordingly. This issue is settled.