GHCN Airports: Leaving on a Jet Plane …
Airports are getting increasing attention from those looking at surface-records as they have become an increasing fraction of the currently reporting weather/climate stations.
Airport locations. Airports are, of course, clearly marked on ONC charts. If a station is located at an airport, this information along with the distance from its associated city or small town (if present) are included as part of GHCN metadata.
Digital Aeronautical Flight Information File
DAFIF or the Digital Aeronautical Flight Information File is a complete and comprehensive database of up-to-date aeronautical data, including information on airports, airways, airspaces, navigation data and other facts relevant to flying in the entire world, managed by the National Geospatial-Intelligence Agency (NGA).
Once publicly available, the US government withdrew public distribution in 2006 based on potential intellectual property claims. One open source for the redistribution of the older data is the Pacific Disaster Center (airports_dafif.zip). This data is in ARC vector format (shape files) which I do not yet know how to handle. Fortunately, spatial-analyst.net has a GeoTiff file (airports.zip) prepared from the shape file format.
Resolution is 1/20th degree (0.05 deg). Airports are presented as one of four classes:
Cat 1 : Active civil airports controlled and operated by civil authority primarily for use by civil aircraft
Cat 2 : Airports jointly controlled, used and/or operated by both civil and military agencies
Cat 3 : Active military airports controlled and operated by military authorities primarily
for use by military aircraft
Cat 4 : Active airports having permanent type surface runways with less than the minimum facilities required for A, B, or C airports above
Running the v2 station inventory through the DataReader gives a fair match count between the DAFIF and the GHCN airport tags.
DAFIF GHCN GHCN AIRPORTS 1049 2390 44% GHCN NO AIRPT 4620 4890 95% ----------------------------- 5669 7280 78%
Reading the failed matches in the ‘ghcn airport’ file, the Dublin Airport jumped out at me. To verify the ‘fail’, I manually ran the GHCN coordinates into the DataReader. It confirmed the ‘no match.’ Then I ran the wiki coordinates; now I had a match.
Dublin Airport 53.43 -6.25 (ghcn) -> 0 53.42 -6.27 (wiki) -> 1
Adding a neighborhood search loop to look at adjacent cells if an airport fails to be located in the initial lookup increases the number of airport hits substantially, but it also increases the number of ‘false hits’ from the non-airport category.
Neighborhood Search (+/- 0.05) DAFIF GHCN GHCN AIRPORTS 1780 2390 74% GHCN NO AIRPT 3771 4890 77% ----------------------------- 5551 7280 76%
We broaden the search once more, to use the original lat,lon pair +/- 0.10 deg:
Neighborhood Search (+/- 0.10) DAFIF GHCN GHCN AIRPORTS 1872 2390 78% GHCN NO AIRPT 3182 4890 65% ----------------------------- 5054 7280 69%
If I had reason to believe that the DAFIF file was authoritatively complete, I could use this method to estimate a rough standard error for the lat,lon of the stations. But comments on the web indicate that the DAFIF file does not contain all ‘airports’ – more likely just those that were of interest to the DOD – which in general would mean ‘bigger.’ It is also likely that while DAFIF contains all major airports, that many of the airport-flagged stations that are not found in it are likely to be at either defunct locations or smaller, local airports.
For instance, GHCN has Denver/Lowry AFB located at 39.72, -104.90 but the DAFIF derived airports.tif file lists this location as ‘no airport.’ Both are right. GHCN has the correct location. And Lowry AFB was closed in 1996 and the whole thing sold off and redeveloped in later years. So no airport in the DAFIF file.
For this particular run, I split the GHCN v2.temperature.inv into two files, one with airports and one without. I then used the +/- 0.10 deg resolution neighborhood to broaden the airport lookup region. If I found an airport outside the given lat,lon, I then adjusted the lat,lon to reflect the ‘new’ coordinates. You can find these location adjusted records by looking for an ‘8’ in 83rd character of the record. In addition, the 84th character gives the DAFIF classification (1-4) instead of the distance to the nearest city. The ‘no airports’ file did not use a neighborhood search and did not include any coordinate changes. The two output files were then merged and sorted.
Neighborhood Search (+/- 0.10) for AIRPORTS only DAFIF GHCN GHCN AIRPORTS 1872 2390 78% GHCN NO AIRPT 4620 4890 95% ----------------------------- 6492 7280 89%
In the inventory, there are the following airport categorizations:
Cat 1: 958
Cat 2: 251
Cat 3: 252
Cat 4: 681
Columns in the new inventories …
82: A for airport or x for none
83: – for none, 8 for coordinate adjusted, or blank
84: 9 for none, 1-4 for airport classification
The v2.airports.inv file: v2.airports.inv.txt
The stacked comparison file: v2.airports.compare.inv.txt
The DAFIF file provides one source for airport lookups. But it’s lack of historical data may make it insufficient for the purpose of wx station classification. This may be one case where automated processing might not be available.
The increasing hit count at wider resolutions may offer a clue to the lat,long deviation – especially when used to calibrate against well known locations such as the Dublin Airport.
I looked at some other information sources:
http://openflights.org/data.html I found very few matches at a .01 degree resolution, around 100. Decreasing the resolution to .1 degree only brought the hit count up to about 500.
http://www2.navigraph.com/www/fmsdata.asp An intriguing source could be in data files prepared for flight simulators. But I have not followed up on this yet.
Another alternative might be as simple as locating an ICAO dataset that includes coordinates.