Allow filtering by sea/ocean/marine region

[MattBlissett]

A while ago I prepared two layers for the geocoder cache: IHO and SeaVoX. Using them is waiting for a couple of decisions:

1) Which layer

The IHO areas are described here: https://www.marineregions.org/sources.php#iho (quick view: https://raw.githubusercontent.com/gbif/geocode/master/geocode-ws/src/main/resources/org/gbif/geocode/ws/layers/iho.png )

And SeaVoX here: https://www.marineregions.org/sources.php#seavox (quick view: https://raw.githubusercontent.com/gbif/geocode/master/geocode-ws/src/main/resources/org/gbif/geocode/ws/layers/seavox.png )

The IHO areas seem to fit exactly with the EEZ areas, so would be a better fit for complementing a continent layer. But a continent layer could also be calculated as the complement of the SeaVoX areas.

Green dots = political, magenta horizontal / brown diagonal = EEZ/IHO (same polygon), dark brown diagonal = SeaVoX (different polygon). http://ws.gbif.org:9012/ but it's extremely slow (hence not public).

This is a decision for @ahahn-gbif.

2) Which terms

Taking into account the discussion in https://github.com/gbif/parsers/issues/26 , I suggest we populate dwc:continent for all terrestrial occurrences, and set dwc:waterBody to a sea/ocean for marine occurrences. I think this leaves dwc:waterBody as-is for rivers, lakes, ponds etc.

I'll specify the interpretation process in detail later, but it would be something like

• geocode to IHO/SeaVoX; if it's null
  • if dwc:waterBody parses to a sea/ocean, keep it, add an issue
  • otherwise keep dwc:waterBody as it is
• if the IHO/SeaVoX interpretation returns a sea/ocean
  • if dwc:waterBody parses to the same, excellent, replace with the normalized one
  • if dwc:waterBody parses to a different sea/ocean, replace, add an issue
  • if dwc:waterBody is not null and not a sea/ocean: todo MattBlissett to investigate
  • if dwc:waterBody is null, set it

A similar process can be used for dwc:continent, but the shapefiles for that aren't ready yet.

Andrea, is using the dwc:waterBody term like this appropriate? I think it's similar to what we do with dwc:country/dwc:countryCode.

The other option is to make a new non-DWC term, and probably just populate it without reference to verbatim fields, as we do for GADM.

[jhnwllr]

Here is how dwc:waterBody is used in practice. I personally feel like replacing someone's dwc:waterBody term with one from a sea polygon is not so nice, so I would personally lean in the direction of using a custom term like GADM.

waterbody	count
null	1.88E+09
Gulf of Mexico	7565698
North Pacific Ocean	1693380
Pacific Ocean	1638520
north; west; offshore; European; Atlantic; Iris...	1122886
North Atlantic Ocean	1054138
South Pacific Ocean	921469
Northwest Atlantic	841704
Southern Ocean	797383
Northeastern Pacific Ocean	696623
northeast Pacific	471986
Sleipner area	468729
Atlantic Ocean	387188
Statfjord	384972
Trondelag area	332001
Oseberg area	281674
Ebro Basin	266402
Pacific	249816
Ekofisk area	224578
not applicable	221456
North Atlantic,New York Bight	214476
Northwest Atlantic,Atlantic Canada,Maritimes (C...	209039
Oir	206247
Mobile Bay	195453
Mar Caribe	191757
Finnmark	182944
North Atlantic Ocean, Gulf of Mexico	180812
Eastern North Pacific Ocean	163273
Scorff	162867
Southern Pacific	156350
Indian Ocean	144883
Southern Eastern Pacific	143568
Molene Archipelago,Iroise Sea,English Channel,I...	138877
North Atlantic,North Carolina,South Carolina	132081
Global	129557
Barents Sea South	119859
Patos Lagoon estuary and adjacent coast	102333
Alaska EEZ	101360
Caribbean Sea	99852
Atlantic	97143
Roche	91222
Corpus Christi Bay	90908
North Atlantic Ocean, Caribbean Sea	90882
Biscayne Bay	89678
Lake Superior	89676
Florida Bay	88501
Mississippi Sound	84402
Aransas Bay	76720
France,Northeast Atlantic	75639
Arctic Ocean	72757
Zeeschelde	72741
San Antonio Bay	71543
Lower Laguna Madre	70968
Matagorda Bay	64597
Little Lagoon	61758
Lake Ontario	61200
Baie de Somme	61049
Perdido Bay	58925

[MattBlissett]

I will run a test interpretation and see what would happen.

I'm not proposing to change "Lake Ontario", but it would normalize "Caribbean Sea" and "North Atlantic Ocean, Caribbean Sea". The loss would be if "Perdido Bay" just becomes "Gulf of Mexico".

[MattBlissett]

We'll also need to take care of what happens around some islands, like Kwajalein Atoll, Marshall Islands:

[ahahn-gbif]

Concerning (1) above (layers for marine areas): personally I would weigh the function in searching for "everything marine" over the complementarity with EEZs, but this may be a question best to consult with OBIS on (@albenson-usgs, would you like to weigh in?). Is it more important to define marine areas politically as "international waters" (excl EEZs), or habitat-based?

That leaves the question whether we would have to aim for having a seamless fit between continents and marine areas, or whether we would tolerate an overlap (continent as a sum of countries, incl. EEZ, + marine areas covering everything right up to the coast).

[albenson-usgs]

This is going to be tricky. @pieterprovoost is going to provide the best advice here. OBIS uses Marine Regions but as far as I know doesn't change any of the data and doesn't add any information to datasets. There's a separate way to filter the data by areas: https://api.obis.org/#/Area.

[MattBlissett]

Since continents are landmasses, I was intending to use polygons that included only land, without coastal waters or EEZ etc.

This is 99% ready to be the continents, although there are some performance issues around the current polygons I have. (The black country borders can be ignored; it's because the continent polygons have been assembled by joining and splitting country polygons.)

This is IHO,

and this is SeaVoX, though I think we'd exclude the North American / European / Asian "Mainland" polygons they have.

Either IHO or SeaVoX can answer "is_marine" within our usual tolerances for determining country.

[pieterprovoost]

The OBIS areas are a combination of full EEZs and intersections between EEZs and IHO (a user may want to filter on US North Pacific for example). We also determine shore distance and hence if records are located on land or not using the OpenStreetMap land polygons as these seem to be more detailed than the typical EEZ / IHO layers.

[timrobertson100]

While we are in here, should we also validate and nullify the altitude field for marine data?

The motivation for this is described here such as this example record.

[dagendresen]

Possible to add Marineregions same way as GADM is? To use the georeferences to search against the oceans?

See also https://github.com/gbif/pipelines/issues/524