Quality check coordinates

[AdamRJensen]

The erroneous coordinates in #345 and #343 indicate that we need to be more careful with the station coordinates when adding new stations.

I suggest that we attempt to do some QC of the coordinates. This could be, for example check that the coordinates of the stations are indeed within the specified country of the station.

@IoannisSifnaios, @kandersolar Any other ideas for enhanced quality control of new/existing stations' coordinates?

[kandersolar]

Brainstorming...

• Check for stations in the ocean (may falsely flag island stations)
• Group stations by network/owner and check for stations that are "outliers" in some sense from the rest of the cluster
• Get an intern/grad student to do it manually

[IoannisSifnaios]

There are lists that can give you the average lat/long for each country (e.g., this one). However, you will still need to do some manual edits for some countries that have stations with both positive and negative longitudes (e.g., Spain and USA).

I guess another check would be about duplicate (or very close) coordinates (we ran a manual test some time ago).

[AdamRJensen]

Based on my recent experience, manual visual inspection seems to be the best way. Not fun, but very effective. I've visually inspected some 230 stations used in the IEA PVPS Task 16 benchmark, which has led to a number of changes, e.g., #363, #358