Centroids of squares can fall outside country borders: considered not georeferenced by GBIF

[peterdesmet]

For some datasets, such as the butterflies, we use the centroids of the 5x5 km UTM squares as decimal coordinates. Some of those centroids however fall outside the country borders.

GBIF flags those occurrences with COUNTRY_COORDINATE_MISMATCH. This is why only 705,941 of the 761,660 are listed as georeferenced, a loss of 55,719 occurrences (7%). Luckily, when you download the occurrences (as simple csv), all coordinates are included.

Solutions:

• GBIF should take into account the coordinateUncertaintyInMeters when flagging records. Since we use a radius that encompasses the whole 5x5km square, it should intersect with the country. That way the flag COUNTRY_COORDINATE_MISMATCH would become much more useful and only flag real mismatch issues.
• We should use a more advanced way of translating squares to coordinates, i.e. forcing the centroid to be part of the shapefile of the country. This would a) be more complicated to calculate, b) be more difficult to explain as a method in georeferenceRemarks and c) on a map, look like records that are not part of a grid.

[peterdesmet]

I have submitted this issue at http://dev.gbif.org/issues/browse/POR-3027