Offer occurrence cubes as data product

[gbif-portal]

Offer occurrence cubes as data product

We just published a preprint (Oldoni et al. 2020) describing a method

... for aggregating species occurrence data into what we coined “occurrence cubes”. The aggregated data can be perceived as a cube with three dimensions - taxonomic, temporal and geographic - and takes into account the spatial uncertainty of each occurrence. The aggregation level of each of the three dimensions can be adapted to the scope. Built on Open Science principles, the method is easily automated and reproducible, and can be used for species trend indicators, maps and distribution models. We are using the method to aggregate species occurrence data for Europe per taxon, year and 1km2 European reference grid, to feed indicators and risk mapping/modelling for the Tracking Invasive Alien Species (TrIAS) project.

We are currently producing these cubes ourselves from GBIF downloads (code is open and referenced in the links below) and publishing the data on Zenodo:

• Occurrence cubes for non-native taxa in Belgium and Europe
• Occurrence cubes at species level for European countries

But it would be great if these could be directly offered as a data product by GBIF (with some limited aggregation options, such as by taxonKey or speciesKey, which reference grid to use, etc.) and then also be referenced by their DOI, just like regular GBIF downloads.

@timrobertson100 suggested to add this idea here in this portal-feedback repo once our preprint is out. We are planning to submit a paper for peer review later this year, but I wanted to get the conversation starting already. Also pinging co-authors @damianooldoni and @qgroom

---

Github user: @peterdesmet User: See in registry System: Chrome 80.0.3987 / Mac OS X 10.14.6 Referer: https://www.gbif.org/user/profile Window size: width 1363 - height 717 API log&_a=(columns:!(_source),index:'prod-varnish-',interval:auto,query:(query_string:(analyze_wildcard:!t,query:'response:%3E499')),sort:!('@timestamp',desc))) Site log&_a=(columns:!(_source),index:'prod-portal-',interval:auto,query:(query_string:(analyze_wildcard:!t,query:'response:%3E499')),sort:!('@timestamp',desc))) System health at time of feedback: INFO

[peterdesmet]

Links that were removed:

• Oldoni et al. 2020: https://doi.org/10.1101/2020.03.23.983601
• Occurrence cubes for non-native taxa in Belgium and Europe: https://doi.org/10.5281/zenodo.3635510
• Occurrence cubes at species level for European countries: https://doi.org/10.5281/zenodo.3637911

[MortenHofft]

Links that were removed

Sorry about that. We had a ton of spam at some point. Perhaps we should allow richer markdown again, we do not get any spam these days.

[dschigel]

This is a very nice idea.

A few opportunities that might help implementation:

• Screen EBV literature from GLOBIS-B make a list of GBIF-dependent EBVs
• Check which of those remain as air castles / concepts, which go into production?
• Where is this production, GEO BON, Senckenberg, other places?
• Work in close contact with people who make GBIF-dependent EBVs, invite them as early testers
• In other words, ensure that outputs is suitable for needs of the target community
• Check for possible the overlap with the CBD indicators (new contract), maybe the same effort on cubes can have additional benefits beyond EBVs
• Communicate well the launch of the cubes output, encourage uptake and stimulate general use (e.g. in ecology).
• Update GBIF documentation and training materials to reflect the new output.

[jhnwllr]

I think outside of Europe polygon-based global grids could be a good choice. https://www.discreteglobalgrids.org/

I have already been using them with some success on aggregating GBIF occurrence data. https://github.com/jhnwllr/gbif_shapefile_geocoder

[timrobertson100]

I had similar thoughts to @jhnwllr that it would be good to allow the user to choose the gridding scheme (e.g. Google S2, UTM-based etc)

[qgroom]

True, but I suspect we need both. The vast majority of people are still working with squares, and all the modelling software is expecting squares. However, for global modelling we do need to start moving towards polygons, so we should start to support them.

[peterdesmet]

Nice! Can you screenshot an example of such a DGG? I can’t see one on the website.

[jhnwllr]

Not on gbif.org. I meant like here https://data-blog.gbif.org/post/exploring-es50-for-gbif/

[peterdesmet]

Aha, so DGGs are ways to divide the earth in equal-sized areas. And those gridding schemes can be referenced. Yes, that would be a good option to offer, in addition to widely used traditional grids like EEA's.