Script to calc site level mean SOC and standard dev

[piersond]

How do we do this with the variable experimental structures? missing data? outliers? etc.

[wwieder]

One helpful resource I'd still like to see is a table from SoDaH that replicates Table C1 from this paper with data from all the sites (or at least NEON, LTER, and NutNet sites). https://www.geosci-model-dev.net/8/1789/2015/gmd-8-1789-2015.pdf

Below are some suggestions that may make this easier:

• I'd focus on the top mineral horizon (0-x cm), or maybe the top several so we can at least cover 0-10 cm.
• We can report the mean & std dev for each variable out of control plots.
• Additional fields to consider would be lat, lon, layer_bot, vegetation cover / ecosystem type, soil C:N.
• Litter chemistry and ANPP may be a pain to track down, but should be doable for NEON. _NOTE, litter chemistry is only available from forest sites, but foliar chemistry is more widely measured, see more on this repo link_
• We'll have to think about how to handle gradients, but at this point I'm included to just include them

From there we'd have to think about where we want to go, but I can imagine this kind of summary table will be useful for a variety of questions & applications.

[wwieder]

@piersond is this something you think you're able to think about? @EmilyKykerSnowman would this be helpful for your paper? @srearl, is this something Nathan can do?

[brunj7]

@wwieder @srearl Feel free to reach out to me if you need help on a specific task and we can further discuss. We have more support capacity over the summer.

[piersond]

@wwieder I have some time this week to start outlining a script for pulling the site means.

[piersond]

I've uploaded a start to the script on GitHub, here.

For now, it's a straightforward group and summarize, and the script only dives into the process of pulling together the soil C data. When we get the soil C summary to work right, I expect we can find points along the way to revise and pull MAP, MAT, etc.

Script pulls the SoDaH database from the EDI script in the parent folder and all looks good so far.

Questions: 1) @wwieder The example table reports C stocks, so I'm assuming we're confined to using lyr_soc_stock_calc? Is lyr_soc relevant to the table without bulk density? So much more data in lyr_soc...thoughts?

2) @srearl Any ideas on a good way to sum soil C over depth increments? The curve ball is that sometimes the depth increments have gaps. So, I'm thinking we'll have to pass the data one sample location at a time through a function that can both find and extrapolate over the depth gaps, then sum the analyte (e.g. soil C) by depth increment to a specified target depth. I'm glad to get started on such a function, but I'm also wondering if Nathan would be interested in giving it a shot? Might also be worthwhile to talk it through on Zoom to either outline the function or to see if we can find a better approach.

[wwieder]

let's go ahead and lyr_soc to get more data for now. Ultimately, we may need to use the stock data for particular applications, but working through some of these issues with a bigger dataset seem useful.

Re. vertically resolved data:

• I think we'll want to sum over depth for individual profiles and then try to average across profiles (as opposed to averaging across depths and then summing the mean).
• I'd also expect we want to set an arbitrary maximum depth (e.g. 10 or 20 cm) and mask out data below that?
• Another alternative would be to just take the top measurement in each profile, and record the bottom depth that's being recorded?
• In short, we may want to evaluate several options about how to do this...

On Mon, Jul 13, 2020 at 7:27 PM Derek Pierson notifications@github.com wrote:

I've uploaded a start to the script on GitHub, here https://github.com/lter/lterwg-som/tree/main/data-aggregation/Summary%20tbl .

For now, it's a straightforward group and summarize, and the script only dives into the process of pulling together the soil C data. When we get the soil C summary to work right, I expect we can find points along the way to revise and pull MAP, MAT, etc.

Script pulls the SoDaH database from the EDI script in the parent folder and all looks good so far.

Questions:

1.

@wwieder https://github.com/wwieder The example table reports C stocks, so I'm assuming we're confined to using lyr_soc_stock_calc? Is lyr_soc relevant to the table without bulk density? So much more data in lyr_soc...thoughts? 2.

@srearl https://github.com/srearl Any ideas on a good way to sum soil C over depth increments? The curve ball is that sometimes the depth increments have gaps. So, I'm thinking we'll have to pass the data one sample location at a time through a function that can both find and extrapolate over the depth gaps, then sum the analyte (e.g. soil C) by depth increment to a specified target depth. I'm glad to get started on such a function, but I'm also wondering if Nathan would be interested in giving it a shot? Might also be worthwhile to talk it through on Zoom to either outline the function or to see if we can find a better approach.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/lter/lterwg-som/issues/72#issuecomment-657913029, or unsubscribe https://github.com/notifications/unsubscribe-auth/AB5IWJGOYQKV77F5I6L5S7LR3OYBDANCNFSM4JDUMFNA .

-- Will Wieder Project Scientist CGD, NCAR 303-497-1352

[srearl]

Thanks for getting this started @piersond. Could you hold my hand a little bit, and point me to an example of where the gaps come into play? Also bringing @NateHaywood into this loop.

[piersond]

@srearl Sorry about the delayed response. I outlined a functional approach to analyte stock calcs by lat-long location last night. I'll touch base when I have the script working and we can all discuss.

[srearl]

cool and no worries - @NateHaywood is already starting to tune on this so let us keep him in the loop

[piersond]

The calc_profile_stocks.R script is now functional, though it still needs a lot of proofing, an option to calc stocks using bulk denisty (right now it's doing a profile avg for lyr_soc), and further summary functions to combine near locations.

You can see the output for lyr_soc in this table. The 1's and 2's in the table rows are error codes for locations that have only one layer, or have many layers that overlap.

@wwieder if you have a few minutes next week, let's discuss the more exact details for the tbl values. i.e. how to combine locations? depth limits? Shouldn't take long to work up those final lyr_soc numbers from where the script stands now.