Autogenerated ODMAP for SDM component

[simonrolph]

Essentially ODMAP (Overview, Data, Model, Assessment, Prediction) protocol provides a format of supplementary material that explains everything about a SDM. Dylan will look at how we can make out SDMs generate this programmatically alongside the model runs.

It is introduced here: https://onlinelibrary.wiley.com/doi/full/10.1111/ecog.05700 and https://onlinelibrary.wiley.com/doi/full/10.1111/ecog.04960

2 outputs, one as data format and another as a rendered pdf?

ODMAP has functionality to import from RMMS meta data standard. I think we should generate this first then pass it to ODMAP.

Tasks

• [x] Read ODMAP paper
• [x] Read RMMS paper
• [x] Work through RMMS vignettes (linked from here https://cran.r-project.org/web/packages/rangeModelMetadata/index.html)
• [ ] Produce a standalone script for testing out RMMS functionality
• [ ] Integrate RMMS into workflow Rmarkdown file
• [ ] Convert RMMS into ODMAP as part of workflow running on LUMI

[simonrolph]

Range Modeling Metadata Standards (RMMS) A similarly aligned piece of work: https://onlinelibrary.wiley.com/doi/10.1111/geb.12993 CRAN: https://cran.r-project.org/web/packages/rangeModelMetadata/index.html

[juliancervos]

As we discussed in our meeting of last Friday (with @simonrolph and @DylanCarbone), I've uploaded an updated version of the RO-Crate metadata including some fields from the Overview section of the ODMAP protocol (see fe0170d and 89c68a1).

It's just the first rough version, and the metadata fields will change or be assigned different terms, but it's helpful for getting us started in collecting that info. Some attributes were already there (author, codeRepository...), some others I added but are linked to existing Schema.org or Darwin Core terms (contentLocation, taxonID...), and some others are placeholders I made up for now (modelObjective, predictorsType...).

I took the values from the table with the ODMAP fields that Dylan uploaded, and also added a new column to the table specifying the name of the metadata attribute(s) that correspond to that ODMAP field to make clear where things go.

There are a few of those fields that still need to be fixed and could be filled in from the code. Maybe a good place to start are the "Location" and "Scale of analysis" ODMAP subsections.

• "Location" is mapped to the contentLocation metadata attribute, which is assigned an URI to GeoNames for the Cairngorms National Park.
• In the metadata, that URI has its own entry in the graph with properties to specify the name, description, geocoordinates, etc (to which other ODMAP fields are mapped to, like "Spatial extent", "Spatial resolution", and "Boundary").
  • [ ] Something that is now missing is the "Spatial extent", which we could start giving as a pair of coordinates (or a circle or a polygon) enclosing the area taken into account for the model. Are those coordinates somewhere in the code?
  • [ ] The same could be done for the temporalCoverage and temporalResolution attributes. What temporal range is the model applicable to? Is that something we can get from the model parameters?

After we solve this, we can tackle the taxon IDs, which is another attribute we could automate from the model runs.

[juliancervos]

Range Modeling Metadata Standards (RMMS) A similarly aligned piece of work: https://onlinelibrary.wiley.com/doi/10.1111/geb.12993 CRAN: https://cran.r-project.org/web/packages/rangeModelMetadata/index.html

Also, I discussed the RMMS standard in a WP5 meeting we had this week, since it could be related to a WP5 task about creating a mapping tool for converting between metadata standards. The short story is that if we want to use RMMS, we should upgrade it to have a better definition of terms (instead of just this csv file).

But then, we can just skip that and create our own... it boils down to whether we want to adopt and extend RMMS because we think it'll make our lives easier or the model more understandable. I think ODMAP is great for structuring the metadata and info we want to collect for the model, but with RMMS I don't see the immediate benefit. Maybe we can leave it as a nice to have for the future.