Saturday, July 30, 2011

Missing Lat/Lons in CRUTEM3

I found the latitude/longitude info in CRUTEM3 to be generally better than in some earlier versions of GHCN. There seemed to be fewer cases of stations turning up in the sea - the KML files that I did make this rather obvious.

However, there were some down as missing data. TempLS needs some lat/lon info - perfect accuracy isn't necessary, but for the weighting it needs to know within a degree or so. So I tracked some of them down. None of the stations are major data sources, so they could have been omitted. Many were in a group of Argentine stations from the early Smithsonian data set. But I tracked the lat/lon to an accuracy that was adequate for my needs - below the jump.

The table of missing lat/lon data is here:
85997377254-999SERRO DO PILARPORTUGAL19011930101901
999201-346583-999ANGEL GALLARDOARGENTINA19321960101932
999207-391676-999CORONEL J.J. GOMEZARGENTINA19311960101931
999211-346595-999FORTIN MERCEDESARGENTINA19311958101931
999213-391-1672-999GEN. GODOYARGENTINA19321959101932
999216-226623-999 LAS DELICIASARGENTINA19311960101931
999236-390681-999VILLA NEUGUESARGENTINA19311960101931

In the table, I've replaced the missing lat/lon (2nd and 3rd cols) by what I think is the coords of the town referred to. Remember that CRUTEM uses negative longitudes. Sometimes there is some ambiguity - for TOSARI, for example, I punted for the mountain town in Java (as suggested by altitude and WMO number).

I couldn't locate LONDON, Australia (it's hard to search for), but there were only two years of readings, so I dropped it. VILLA NEUGUES I assumed was a typo for something involving Neuquen. Tjibodas I tt\ook to be the botanic gargens near Jakarta, and Angel Gallardo the district in BA.

I corrected an error with Clare Post Office. It was entered twice - once with the major set of data, and once with a patchy set, mostly missing , and with latitude of wrong sign. I corrected the latitude, but probably the whole set should be omitted.

There is a lot of altitude data missing (-999), but I don't need it.


  1. Nick Perhaps submitting a note to CRU would help a bit. I'm looking into some web services to correct the data, but without a good name, it looks like one would have to do things by hand. better that CRU fix things on their end

  2. Anon,
    Yes, I'll do that. I plan to wait a couple of days in case anyone else has found things (or I find more). But there's nothing now that obviously gets in the way of reconstruction.