improve vertical profiles figure

[teixeirak]

@mcgregorian1,

As I think more about this, I've thought of a few things that could make this figure more meaningful:

• [x] adjust NEON instrument heights slightly -- see .pdf from John Schnebelen. (also listed in issue #48 )
• [x] rather than displaying the mean (±SD) of daily minima and maxima, what if we display the 5th percentile of the min and 95th percentile of the max. This basically gives us the range of conditions experienced at each height, and therefore more information about the extremes that may be reached on a hot/dry day, which may represent more average conditions during a drought. [I suspect that if the biological temperature readings are giving meaningful info about leaf temperature, we may see a profile. See issue #57.]
• [x] can we add [CO_2] to the height profile? There's likely a gradient, with higher concentrations at lower heights, especially during the day when photosynthesis is drawing down concentrations, or on nights with little mixing when soil respiration makes high concentrations close to the ground (for example, see this pub). This is physiologically meaningful in that understory trees can have higher water use efficiency during drought-- a mechanism that we didn't discuss in Bennett et al.!

[mcgregorian1]

• [ ] adjust NEON instrument heights slightly -- see .pdf from John Schnebelen. (also listed in issue #48 )

I'm not sure about doing this - the vertical position heights that I've been using are the ones that are coming straight from the NEON data, so they're the ones who label the measurements as being from 10,20,30,40,50,60m. Did John say we should specifically use these heights? I'm wondering if changing them would make it unnecessarily complex.

We could instead refer people to the pdf? I'm struggling to understand why, if these are the correct heights, NEON would simplify them to give us, essentially, incorrect heights in the actual data.

data height	pdf drawing height
10	5.8
20	19.2
30	26
40	32.9
50	38
60	51.8

[teixeirak]

Yeah... but I'm pretty sure most of the documentation is an oversimplification and these are the actual heights. It's probably a good idea to check with John to verify. Those are quite significant differences, so not good if the pdf drawing is accurate!

[mcgregorian1]

Exactly! I completely agree. Fixing them will be simple since it doesn't involve any computation, so I'll focus on doing the major changes for now

[mcgregorian1]

• [x] rather than displaying the mean (±SD) of daily minima and maxima, what if we display the 5th percentile of the min and 95th percentile of the max. This basically gives us the range of conditions experienced at each height, and therefore more information about the extremes that may be reached on a hot/dry day, which may represent more average conditions during a drought. [I suspect that if the biological temperature readings are giving meaningful info about leaf temperature, we may see a profile. See issue #57.]

To clarify, there are 2 ways I can do this:

1. I take the 5th and 95th percentile of the max and min values themselves, before I calculate the mean value of each that's currently plotted.

2. I take the 5th and 95th percentile of the raw min and max data, then when I get the monthly average for min and max, I also take the average of the 5th and 95th percentiles.

I'm not sure which makes the most scientific sense

[mcgregorian1]

• [x] can we add [CO_2] to the height profile? There's likely a gradient, with higher concentrations at lower heights, especially during the day when photosynthesis is drawing down concentrations, or on nights with little mixing when soil respiration makes high concentrations close to the ground (for example, see this pub). This is physiologically meaningful in that understory trees can have higher water use efficiency during drought-- a mechanism that we didn't discuss in Bennett et al.!

I agree this would be good! However I just checked the NEON site and it seems the data isn't available yet and no word on when it will be...

[teixeirak]

1. I take the 5th and 95th percentile of the max and min values themselves, before I calculate the mean value of each that's currently plotted.

This is what you want to do.

[teixeirak]

*

I agree this would be good! However I just checked the NEON site and it seems the data isn't available yet and no word on when it will be...

Bummer!

[mcgregorian1]

1. I take the 5th and 95th percentile of the max and min values themselves, before I calculate the mean value of each that's currently plotted.

This is what you want to do.

I've made a new graph called Figure2_test - it's not perfect in terms of label spacing because I'm holding off in case we decide to not use this.

From what I can tell, using the 5th and 95th percentiles do give us a different angle, and it's interesting to see how the error bars are almost negligible for some values (I'm using the same standard error that was used for the mean error bars).

Sorry that the linetypes are switched from the original. I'm not sure how that happened.

New

Original

[teixeirak]

Thanks! That does do a better job characterizing the range of conditions experienced, but in some ways becomes less interesting/ informative (e.g., 100% humidity as 95th percentile for all months/heights). I think I'd actually prefer to stick with the original.

Because nothing interesting comes out of T_biological, I'd just drop it.

[mcgregorian1]

Ok, sounds good; I've pushed the result

[teixeirak]

Looks good; thanks! We can close this once we resolve the NEON profile heights.

[teixeirak]

I assume you saw John's email that the heights on the diagram are correct, so we will need to make that change. I will close this issue, as it's covered in issue #48.

[mcgregorian1]

@teixeirak I've fixed the heights. Currently they're plotting fine (Figure2.png), but up to you if you want the actual heights represented on the labels of the graphs. What do you think?

[teixeirak]

It's good as is.