realize spatio-temporal metadata for images

[coltongrainger]

• [ ] create metadata for latitude/longitude points (SRID 4326)
• [ ] document MySQL 5.7's limitations for reprojections
  • see https://stackoverflow.com/questions/21877179/mysql-spatial-convert-from-srid-31287-to-4326
• [x] ask Philip for best practices, how to quantify uncertainty?
• [x] decide how to implement "uncertainty buffer" as a boundary polygon
• [x] decide how to implement "region" as a boundary polygon

references

• https://dev.mysql.com/doc/refman/5.7/en/spatial-operator-functions.html#function_st-buffer
• https://itnext.io/playing-with-geometry-spatial-data-type-in-mysql-645b83880331
• http://marineregions.org/gazetteer.php?p=webservices

[coltongrainger]

spatio-temporal metadata fields

[coltongrainger]

9 addresses specifically "decide how to implement region as a boundary polygon"

[coltongrainger]

I'll try to formalize a schema that indicates "fixed" locations for both cases (either explicit coordinates given or zero velocity indicated).

To handle metadata across 24 hours, I'm planning that each entry in the table page will have ~24 (or fewer, depending on the resolution of the logbook) children entries in the table observation, where observation entries will be keyed by the composite of their source page id and their iso-datetime.

Lastly, perhaps later I could ask how one might classify location uncertainty during different technological epochs, e.g., pre/post-chronometer, pre/post-radionavigation.

(I've looked through the WMO's Guide to Meteorological Instruments and Methods of Observation http://library.wmo.int/opac/doc_num.php?explnum_id=3148 (general uncertainty), http://library.wmo.int/opac/doc_num.php?explnum_id=3182 (marine observations), and A review of uncertainty in in situ measurements and data sets of sea surface temperature https://doi.org/10.1002/2013RG000434 for notes on uncertainty in measurement meteorological variables. However, neither reference discusses uncertainty of location. Perhaps there's a better reference here.)

[coltongrainger]

On 2019-06-18, Philip Brohan philip@brohan.org wrote:

Have a look at the log page at

http://s3-eu-west-1.amazonaws.com/textract.samples/Farragut-DD-348-1942-01-0021.jpg

This page has three locations (each lat and lon) - often we have only one, at noon. It also has hourly speed and course measurements. Those speed and course data are the standard fields you mention. See also the older logbook at http://s3-eu-west-1.amazonaws.com/textract.samples/24-118-jeannette-vol003_199.jpg - same fields (though that ship is frozen in the ice so has no speed data recorded).

So if we are willing to transcribe the 24 speed and course measurements (expensive), as well as the three positions, we can estimate a location time-series for the image. The point is that one image is typically 24-hours of data, and the metadata may change over that period.

I think it would be be good to 'associate certain times (i.e., certain image files) between "fixes" with a non-negative radius that would bound the probable location of the observation'. Often our location metadata is approximate - I've used ship examples here, but we have station data where the location metadata is 'London'.