Missing microclimate data in SW USA

[mpinsky]

We seem to be missing a lot of microclimate (temperature SD) data in the SW USA.

See map produced here

Is this a bug that we can fix?

[lauraantao]

This seems to be related to these cells being coastal, and with the central lat-long being "too far inland" to be able to extract information from the marine raster (so all the extracted values are NA; there were only two terrestrial cells). One solution would be to use a larger buffer for these locations, but this still did not fix it for all the cells...

I re-did the extraction using a 80 km buffer (only for the marine cells with Temp_sd20km==NA), and this still had ~810 records without sd values. The new file with this new extraction is in the output folder (here) - let me know if this makes sense, and I can then also update "assemble_microclimates.R" with this "extra step".

[AmandaEBates]

but shouldn't we use a finer scale temperature data set?

[mpinsky]

Amanda, Laura's using WorldClim on land and BioOracle in the ocean. If you know anything finer scale, let us know.

Laura, I think something is wrong with the lat-lon or the REALM classification for those studies. Do you mind digging in to figure it out? I made a plot, and they are in the middle of California or even Arizona, very far from ocean. The plot zoomed it:

[AmandaEBates]

okay, it is the climatology - got it.

On Wed, May 27, 2020 at 6:43 AM Malin Pinsky notifications@github.com wrote:

Amanda, Laura's using WorldClim on land and BioOracle in the ocean. If you know anything finer scale, let us know.

Laura, I think something is wrong with the lat-lon or the REALM classification for those studies. Do you mind digging in to figure it out? I made a plot, and they are in the middle of California or even Arizona, very far from ocean. The plot zoomed it:

[image: image] https://user-images.githubusercontent.com/6370532/83000686-fa704400-a00a-11ea-931d-545b7c2b5f77.png

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/pinskylab/climate_community_crossrealm/issues/6#issuecomment-634534397, or unsubscribe https://github.com/notifications/unsubscribe-auth/AD2B4CK7RPKBLDDBSKUW7NLRTTKTVANCNFSM4NGUULLA .

-- I am working afternoons and evenings, so apologies for delays in my email responses. Keep safe!

Amanda E. Bird, Mum & Amanda E. Bates, Associate Professor Canada Research Chair in Marine Physiological Ecology

Memorial University of Newfoundland 0 Marine Lab Road, Department of Ocean Sciences Logy Bay, NL, Canada Phone: 709-864-3241 Office Number: OS-3014A

abates@mun.ca abates@mun.ca

https://www.physiologicaldiversitylab.com https://www.physiologicaldiversitylab.com

[lauraantao]

Thanks Amanda! Yes, this is to calculate an estimate of microclimates for each grid cell, and we are using the same data for getting "climatology" as in our previous analysis.

About the coordinates in relation to Realm, there is no issue per se with the studies classification, as the variable "REALM" is based on the original studies information, which was "transferred" to the individual grid cells, depending on whether they contained samples for each study ID or not. Then issues might arise related to the location of the grid cells and where the central lat-long falls.

Digging down a bit, almost all of those cells refer to studyID=169 ("CalCOFI and NMFS Seabird and Marine Mammal Observation Data. 1987-2006 (SEAMAP)") - so it could be there was some issue with the original coordinates from this study...

[mpinsky]

Hmm. That looks like a big issue with the coordinates for that study! That's very much a marine study, not inland. Maybe an error in the data Shane sent us? Or the problem could trace back to the original BioTime.

[AmandaEBates]

I would love to know. a

On Wed, May 27, 2020 at 5:25 PM Malin Pinsky notifications@github.com wrote:

Hmm. That looks like a big issue with the coordinates for that study! That's very much a marine study, not inland. Maybe an error in the data Shane sent us? Or the problem could trace back to the original BioTime.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/pinskylab/climate_community_crossrealm/issues/6#issuecomment-634906185, or unsubscribe https://github.com/notifications/unsubscribe-auth/AD2B4CM5CLDSDZJAPM42LFDRTVV4DANCNFSM4NGUULLA .

-- I am working afternoons and evenings, so apologies for delays in my email responses. Keep safe!

Amanda E. Bird, Mum & Amanda E. Bates, Associate Professor Canada Research Chair in Marine Physiological Ecology

Memorial University of Newfoundland 0 Marine Lab Road, Department of Ocean Sciences Logy Bay, NL, Canada Phone: 709-864-3241 Office Number: OS-3014A

abates@mun.ca abates@mun.ca

https://www.physiologicaldiversitylab.com https://www.physiologicaldiversitylab.com

[mpinsky]

Tagging @sablowes (I'm still learning how to use github...)

[lauraantao]

It looks like the issue originates in the original records... This is the map when plotting the individual sample locations (data as in BioTIME query in 10-2018):

And this is what the distribution map looks like in the OBIS page for this study:

And they note these data issues:

This study was added in 2012, so I will let the data manager know there are issues to be checked :( ... I guess one (very unsatisfying) solution for us now would be to exclude this study from the analysis... :/

[sablowes]

Laura's plot from the BioTIME query suggests that the gridding process would have landed these samples on land. So I don't think I need to chase down a mistake in my stuff, do I?

An alternate to dropping the whole study would be to identify the cells that have the samples with terrestrial coords - can we do that?

[mpinsky]

@sablowes , I agree, seems to be a problem in the original BioTime data based on @lauraantao 's sleuthing.

Dropping just the coords on land is attractive and might be a good solution. The coords in the ocean look about correct as compared to the OBIS plot.