Hypothesis 3: Access to water (elevation) affects drought response

[mcgregorian1]

H3- Rooting volume/depth relative to water sources are critical in drought response. Effects of drought on larger trees are mediated by the fact that large trees have better access to water.

P3a- drought response increases with elevation and/or distance from stream (it would be nice if we knew more about the site hydrology!)

Next step:

• [x] try a model with just elev (+ random effects).
• [x] check/report direction of elev term in model.
• [x] consider whether we may have a better option than elev to characterize access to water (e.g., elevation relative to nearest stream)

P3b- There is a dbh*elev interaction, elevation (relative to stream?) matters less for big trees with larger (and potentially deeper) root system.

Next step:

• [x] try a model with dbh*elev (+ random effects) to see if there is an interaction. Be sure to note direction of interaction, as this will differentiate between H2 and H3 (see below).

Originally posted by @teixeirak's writing in issue #7

[mcgregorian1]

P3a

model with just elev and random effects

EDIT: 4 days later, realized I included year as a random effect when it's a fixed effect. I have changed it.

• the r-squared for the top two models are 0.133.
• there appears to not really be a significant relationship here.
• As for considering a better option with respect to water, I've noticed that the shapefile we have for the streams at SCBI are not accurate anymore. The E-W stream, for example, branches off into 2 and sometimes 3-4 parts in places, which isn't reflected in the shapefile. Ideally, I'd go out in the plot with a Garmin GPS and retake points of the stream line to create a new shapefile that's more accurate. -- Alternatively, I can create a buffer of all trees within a 20m circle of proximity to the stream (as we have it currently) and assign them a "1" and everything else "0" - this would then be a fixed effect. -- OR, I can calculate distance from each tree to the stream and add that in as a separate column / call that a fixed effect.

[mcgregorian1]

P3b

model with dbh_ln*elev plus random effects

• the r-squared for the top model is 0.16.

• definitely more of a significant effect of the interaction compared with the results in P3a.

• Adding in dbh_ln separately to this gives pretty much the same picture. Inherently, dbh_ln*elev_m includes the effects of having dbh_ln and elev_m as separate effects along with the interaction.

[teixeirak]

which direction is the interaction?

[mcgregorian1]

The direction is negative, though the individual effects are both positive

also my bad on the original - I had included year as a random variable even though I don't have enough groups to consider it random

[teixeirak]

Okay, so it indicates that drought response increases with both elevation and dbh, but elevation has less impact on larger trees, supporting P2b.

[teixeirak]

Regarding a metric of distance to the stream, I'd think either of the suggestions above should give us a preliminary idea (and if its borderline or significant it may be worth improving the map). For the second option (distance to stream), you'd want to use ln[distance].

[mcgregorian1]

I've now calculated the distance for each point to water and added it as the ln[distance] (option 2 from above).

Running the model dbh_ln*distance_ln plus random effect, I get:

• r-squared for the modelsis 0.16 for the best.

Here is the interaction summary:

[teixeirak]

Thanks! Could you please try it without the interaction (and give the coefficient)?

[mcgregorian1]

Running with no interaction yields:

• the r-squareds for the top two models are both 0.162 and 0.164, respectively.

These are the coefficients for the top three models:

Running without log-transformed distance gives an even weaker signal:

[mcgregorian1]

For the record, running dbh_ln + elev_m*distance_ln gives me almost the same result as above, with dbh_ln + (1|sp/tree) the best model by 2, but the next best model has the same r-squared (0.16) and is the full model

[mcgregorian1]

@teixeirak both H2 and H3 include a test for dbh*elev interaction, but in the table you listed it as height*elev (just want to make sure this was intentional - I'm assuming this is because height has the larger effect?).

Otherwise, the response we're seeing in the updated table (as of 12.30 today) is what I'm getting for running identical models

[teixeirak]

use height x elev for both. Note that the different hypotheses predict different directions of interaction.

[mcgregorian1]

Ok, sounds good. That's what the script is currently doing so I think it can stand as is

[teixeirak]

I think we're done with this issue.