search

SCBI-ForestGEO / McGregor_climate-sensitivity-variation

repository for linking the climate sensitity of tree growth (derived from cores) to functional traits
0 stars 0 forks source link

test for correlation between traits and microenvironment #43

Closed teixeirak closed 5 years ago

teixeirak commented 5 years ago

@mcgregorian1, you've already made plots showing the relationship between mean traits and microenvironment. Can you now please test whether there are any significant trends?

mcgregorian1 commented 5 years ago

Did you mean a linear model? And between which traits? If I take it as it is now, I'd be doing a linear model for TWI - PLA TWI - TLP TWI - rp

Did you want more than this?

teixeirak commented 5 years ago

Here's what I have currently for the hypotheses/ predictions:

2. How do species' traits--alone and in interaction with tree size--influence drought response?

2.1. Commonly measured traits including wood density (REFS), leaf mass per area [@abrams_adaptations_1990] [@guerfel_impacts_2009], and ring porosity (Elliot et al. 2015, Friedrichs et al. 2009) have been linked to drought responses and likely correlated with drought resistance in this forest. However, we hypothesize that leaf hydraulic traits such as leaf area shrinkage upon dessication and turgor loss point, which are emerging as potentially more informative traits but whose effect on drought resistance has never been tested (CONFIRM), will prove better predictors.

2.2. Traits vary with microenvironment, with more drought-resistant traits associated to taller trees (and drier topographic positions) (issue #43)

 * P2.2a- traits (all 5 above) correlate with tree height 
 * (P2.2b- traits (all 5 above) correlate with TWI)

2.3. Size-dependent drought resistance is driven by--or alternatively, buffered by--functional traits (i.e., larger trees have more drought vulnerable traits).

teixeirak commented 5 years ago

I think it would be good to test all species' traits against both height and TWI.

TLP and PLA are higher priorities because they come out as the strongest predictors.

It could be interesting to expand beyond the 12 species for which we have TLP & PLA with the other traits. WD, LMA, and RP are so widely measured/ known that we could get much better coverage. Of course, that's more work-- not how long it would take you/ if worth it.

mcgregorian1 commented 5 years ago

I agree, I think it would be best to get everything in, especially if we have the data for it.

Regarding TLP and PLA, though, I believe the 12 species are the only species we have data for. That's why originally we took out pist, for example, because we had no species traits data for it.

Given this, I'm also cognizant of how much time I have left before I start having no time (with schoolwork and my research assistantship). I think expanding to get more data and add more species may be too much for me to do with what I have left. In your opinion, do you think this paper would be a lot better getting that extra data in? In that case, I'd be fine doing what I can for we currently have, and then letting someone else take up the next steps.

Otherwise, I think based on the hypothesis predictions you have above, I'm assuming you want to add to this trait table? Is this what you were thinking of? image

mcgregorian1 commented 5 years ago

Making a table like this shouldn't take too much time, it's just a matter of inserting it into the code format I already have

teixeirak commented 5 years ago

I wouldn't change our existing structure for testing the traits. That is, we do still want to include height in the null model.

Testing 2.2 will just be [trait] ~ height or [trait] ~ TWI. That should be a separate table in the SI; I don't expect it will be interesting enough to make it into the main manuscript.

For hypothesis 2.3 (erroneously labeled as 3.1 before), again I wouldn't change our existing analyses. Rather, we the test of this prediction is whether the best model, including traits, still has efects of height & TWI. I've updated the text to clarify this.

mcgregorian1 commented 5 years ago

Ahh I see. So that means 2.2 will be tested with just linear models. I can do that

mcgregorian1 commented 5 years ago

Hi @teixeirak

Here is a chart with the p-values compared to height. According to this, it appears that everything is significant.

I noticed that in the hypotheses tables, you say community mean for the traits, relating to using all tree data for the plot. In the table, though, these results are put under the "Overall" section (though the data that's being tested is not only the overall drought trend). Did we want to update the hypothesis table?

image

mcgregorian1 commented 5 years ago

This is for my understanding:

  1. As you go up in height, WD decreases. This means taller trees are less wood-dense, which means they're more drought-sensitive. (from chen et al 2019)
  2. As you go up in height, LMA increases. This means taller trees have thicker leaves, which means taller trees are more drought-resistant (also supported by chen et al)
  3. As you go up in height, rp becomes more diffuse. This is seen because the categories of rp are ring (1), semi-ring(1.5), and diffuse(2). This is slightly biased because litu are the tallest tree and dominate in terms of biomass in plot, and litu are diffuse.
  4. As you up in height, PLA increases. This means leaf shrinkage is increasing, which means the taller trees are less drought-tolerant.
  5. As you go up in height, TLP increases (becomes less negative). The less negative TLP a tree has, the less resistant to drought it is. This means taller trees are less resistant to drought.

@teixeirak for LMA, I've found some papers confirming that higher LMA means more drought-tolerant. This is the only one of the traits that contrasts the taller trees are more drought-sensitive. Does it make sense to have this contradiction?

teixeirak commented 5 years ago

I don't really have any prior expectation/ reason to believe there should be a trend one way or the other. Obviously there are many traits that contribute to vulnerability or resistance, and it seems they can go opposite directions.

Keep in mind that we're only looking at canopy species, and at species-level traits. What this tells us is that drought sensitivity isn't a strong filter on which (canopy) species dominate canopy positions at this site. That's not surprising, as there's a lot of other things going on.

I do think we'd get some different results if we included the understory species.

teixeirak commented 5 years ago

Closing (complete).