datadrivenecologicalsynthesis / weirdestspeciescombination Goto Github PK

Requires cleaned gbif species names (#41) and accession numbers (#39) and list of finalized traits (#34).

Uploaded a .csv of potential traits to pull for the trait requests - tried to have a variety of options that should have enough data. Just wanted people's input on it before requesting the data

1. Compare std among the quadrats (so we can say which one is the weirdest i.e. within city comparison)

2. Compare the community composition (species list) among cities and determine how much overlap there is (e.g. are 90% of the species in each city the same?)

3. Figure out the extreme ends of the traits within each city? (like are they all Juniper vs Impatiens?)

4. Make a species richness x quadrat barchart

5. Assign parts to folks & write up the scripts etc.

We got the TRY data for the first request (1000 species) as a .txt file. It came with a readMe that makes it easy to get this data into R and write a .csv, however the .csv is too large for GitHub. It is over 2 mil rows, or 1081GB (limit for github is 25GB). I think we have to make some kind of loop to break it down into small enough .csv (quick math of 1081/25 = ~73 .csvs). Or someone who is better at this than me can find a work around?

To make it accessible for everyone I will share the TRY data email with everyone.

Basically combining data/traitData.csv and MasterSpeciesList_clean.csv

Data folder would be clearer with a short meta file describing what's there and where it's from.

Might also want to put species lists into a folder and then fix paths in scripts.

figure out if there's anything left in matchGbifTry.R that we want, move it somewhere useful. GetTRY_ids.R duplicates some of matchGbifTry.R's functionality and this is confusing.

WINNOW REDUCE THAT LIST

Get it down to 10, 000 species, using an (arbitrary) cut off of occurrences until we're below 10,000 species

Should we include non-urban areas?

We chose (near) urban areas because of sampling density and human interest.

But it might be interesting to include non-urban areas near cities as comparisons.

update meta.md and workflow.md through the the work merged in with #59

also update the roadmap!

Write a brief description of the workflow - order to run scripts and TRY steps.

Place to brainstorm and organize figures

1. Histogram showing SLA distribution among the 6 cities

   • Do we also want to do this by quadrat?

2. List of weird species (5th and 95th quantiles)

   • By city?
   • By quadrats across cities?
   • Quadrats within each city?
   • All species?

3. How to illustrate differences in weirdness within a city

4. How to illustrate differences in weirdness among all cities

Which open refine file cleans what data? Bonus - update workflow.md to reflect new titles

1. Fix the for loop
2. Generate species lists for all 6 cities
3. Generate one Master list that includes all the species names and a column that sums all of their occurrences

We haven't settled on a final list yet, but using these (or some of these) traits suggested by @BenMerSci in issue #34

SLA
Seed dry mass
Plant woodiness
Plant height vegetative
Leaf thickness
Leaf phenols content per leaf dry mass ( proxy for palatable/"toxic"?)
Leaf petiole length
Leaf nitrogen (N) content per leaf dry mass

pull trait data for Toronto plants from TRY. Accession numbers for Toronto plants are now in the data/tryAccession folder.

Minor issue, but I noticed that our LICENSE isn't displaying like the other projects, like we can't see what it includes or not when we click on it.
Feels like it's only a normal text file and not an actual LICENSE.
Can't find what's the problem though....

We've currently limited it to plants because of trait data availability and our own interest.

We may need to limit it further if we want to include a larger area.

Define weird species

• overall (globally weird)
• and for each city (locally weird)

using 5th and 95th percentiles.

Problems:

• There is close to 500 k observations within the Toronto boundary. GBIF allows downloads in batches of 500 with a maximum of 200,000 / day. Too many for DL and takes a long time even smaller batches.
• The Toronto boundary is uneven, based on rounding lat & lon, there is also the lake in the way. Need to create polygon that is comparable to other cities.

Initial tidying in matchGbifDataTry.R

#28 (comment)

Consider

• How many? (Enough to be interesting, not so many that it's overwhelming)
• Different biogeographical regions?

Decide which traits will be included in the analysis.

I think that @BenMerSci 's idea about adding other layers like climate might be useful for answering this question (if we decide to pursue it.)

Group discussion pointed out that city connectivity (e.g. ports) might matter.

I think that Vancouver might win because so many kinds of plants can survive it's mild climate. But that might depend on how many observations got made at the Montreal Botanical Gardens!

example pipeline for matching TRY accession numbers built for Toronto in #37

Extend to all cities.

Associate TRY identifiers with species occurrences from gbif

How exactly are we defining weird?

Thoughts from the group included

• defining a normal and seeing what would violate that
• what makes you say "whoa!"
• trait ordination and seeing which assemblages are the widest range of traits in that space

How are we handling non-native species? Does the dataset we're using include them?