Temperature in Guam: Filling a Gap

Congressman Johnson may be concerned about the increasing population on Guam causing the island to tip over, but has population increase had interesting consequences for one of the island’s weather stations?

Andersen Airforce Base, Guam (original here)

There are three stations for Guam in the GHCN data file:  Andersen AFB; NWSO Agana and WSMO Agana. For such a small island three weather stations, all in similar topography on the northern part of the island, would be expected to show very good correlation, after all they are within 10 miles of each other, but one shows a very large and sudden warming during the1990s.
These are the stations from GISS Station Data:

Nwso Agana, G 13.5N 144.8E (WMO id 91212) Rural area 1921 – 2010
Wsmo Agana, G 13.6N 144.8E (WMO id 91217) Rural area 1956 – 2004
Andersen Afb, 13.6N    144.9E (WMO id 91218) Rural area 1949 – 1981

As these are all classified as rural stations, they are not adjusted. NWSO Agana is located at the busy international airport, seen in the picture on the right just inland from the capital, Hagatna, at night.  [I wonder what the new nightlights analysis being used by GISS will make of this?  Hmm I might check and update here later]. Andersen Airforce Base is in the northeast of the island, but WSMO Agana was hard to find initially due to the very approximate latitudes and longitudes in the GISS station data.

Here is what the three temperature records look like when plotted together:

For Andersen Afb and NWSO Agana there are variations of just over a degree, and there is obviously a discrepancy between them in how the local topography responds to climate.  Also, these are airports so there may be human influences there. However, that is one hell of a rise in WSMO Agana. How can you get that much warming at one of three such close stations in such a short period? Did no-one notice? I thought a lot of the checking that is done of the climate record is supposed to pick up and throw out “outliers”?  OK, perhaps that is just the odd month or the odd year. This has intrigued me for a while. That cannot be normal and it certainly isn’t caused by CO2. Time to do some digging…..

Guam is located in the western North Pacific Ocean, the largest and furthest south island of the chain of volcanic islands that comprise the Mariana archipelago. The island is oriented northeast-southwest, about 20 miles long, 8-12 miles wide and covers an area of 209 square miles. The southern region is mountainous, with eleven peaks rising to 1,000 feet or more that form a discontinuous ridge that extends along the southwestern part. The northern region is a relatively flat limestone plateau, with steep coastal cliffs and narrow coastal plains that dominate this part of the island. The topography of this plateau gently undulates between elevations of 200 to 600 feet. Guam has a tropical climate, with year-round warm weather, a dry and a wet season, and moderate to high humidity. Seasonal temperatures vary approximately 5°F (2.7°C), with an annual average maximum temperature of 86°F (30°C) and an annual average annual minimum temperature of 76°F (24.5°C). The range of temperature between day and night is approximately 15-20°F (1).

OK Nice background, how what about those weather stations? Well Andersen Afb was easy to find and it wasn’t too hard to find out that NWSO Agana was at the International Airport.  The NCDC Station Locator has this to say about WSMO Agana (and provided a more precise latitude/longitude):

“The National Weather Service station on Guam is located on the western side of the northern plateau. The ocean is l l/2 miles to the west, 9 miles to the north and east, and 5 miles to the southeast. The weather instruments at the station are well exposed in the center of an open field that is 40 acres in area. The trade winds reach the station after rising sharply up the 500-foot cliffs on the eastern side of the island and flowing 9 miles on an easy downslope grade across the surface of the northern plateau.”

Sounds good. I also found it through the Western Regional Climate Center. Here it is (pity about the cloud):

Looks rural enough. What could have caused that rise in temperature? So I started to read up about Guam.

I started this a while back and it has been like doing a very big jigsaw – you know one of those several 1000 piece ones you get at Christmas that sits on the dining room table for weeks until you have the patience to finish it; either you eventually get fed up with it taking up space or you invite friends for dinner and it has to go back in the box.  It still amazes me how much information you can find on the internet. I have come to like Guam – I really feel I know the place now and have a real hankering to go there some day, but I digress…

Guam Maps (L to R): Locations (2); Topography; Population Density; Buildings (1)

I looked at topography:
Perhaps there has been vegetation clearance around WSMO Agana, and/or a lot of encroachment of urban areas and buildings. I found that the population of Dededo, for example, increased by 40% from 31,728 to 44,716 between 1990 and 2002 (1 – Table 4.1).

I looked in more detail into the weather:
Wind speed and direction varies with Guam’s two seasons. The wet season lasts typically from June to December and is characterized by a high relative humidity averaging 86% and weak southeasterly or southerly winds. The dry season generally lasts from December to June and is characterized by a low average relative humidity of 71% and “trade winds” from the northeast. So the station seems to be downwind of an increasing urban area in both seasons – could that be it?

And then I found another reason…..

Look at that -17 stations in total! But what I failed to notice at first on the NCDC site – the end date for WSMO Agana: 2nd April 1998. NCDC concurs. Quick – scroll back up to that graph again. Yup – data for WSMO Agana up to 2004.  How can this be?  FILLED in DATA?

Now the WWRC also had data for these three stations, although it said “last updated 2006”. There was a good fit with what I had pulled off the GISS site (at the end of November) – the data originating in the NCDC GHCN data file.  I just plotted what I found and there are minor discrepancies granted, but it is the gaps I was interested in:

In the WRCC data, Andersen Afb continues reporting until 2002, with rising temperatures; WMSO Agana stops reporting in 1997 (as it should); there is no gap for NWSO Agana, but that fits reasonably closely with the post-1997 WMSO Agana data. So in the GHCN/GISS data NWSO Agana stops reporting for a while, but the data seems to be spliced into another station.  Interesting. So what? Well the implications only become clear when you realise what this does to the temperature anomalies. The following graph shows the effect:

Look at that hockey stick – wow! This gives a sustained 1.5 degree anomaly for Guam for nearly a decade. And that is what drew my attention to it in the first place – a hot spot on the GISS anomaly map.

The thing is, an anomaly like this does not just affect a discrete location, it affects at least a grid square on the global map and is extrapolated into other grid squares too.  Here’s how GIStemp sees this area:

GISS grid boxes for the Guam area: superimposed on satellite view (L) and detail of GIStemp grids (R)

Here is what the anomaly data looks like for the stations in this area (note that Koror (Palau) and Truk (Chuuk) are over a 1000km from Guam). WSMO Agana still sticks up like a sore thumb:

Here is how it looks in terms of the data plotted for the Guam grid cell after it has been through GIStemp (this is Step 3, with which I have very little familiarity as yet – try here at E.M.Smith’s blog for further information):

Gridcell output for Guam: with and without WSMO Agana (L); difference (R)

The stations contributing to this gridcell are: NWSO Agana; WSMO Agana; Andersen Afb; Truk; Falalop Island; Saipan. Likewise the three stations on Guam contribute to surrounding gridcells, which depending on the smoothing radius, can be up to 1200km away. There are 55 such gridcells. Taking WSMO Agana data out of the gridcell anomaly calculation makes a large difference to many of the cells, principally in 1998 and 1999 when the data for Yap and Truk shows a low anomaly value. Here is a table of data for the eight grid squares surrounding Guam, for which the maximum anomaly increase due to the WSMO Agana data has been calculated:

No wonder there is a nice hot spot around Guam on anomaly maps. Actually I should say not “is” but “was”, because I found a few weeks ago that all three Guam stations are no longer available from the “after Homogeneity Adjustment” version of the GISS station data web page. This should mean they are no longer part of the GISS adjusted data set and should therefore not contribute to the output of GIStemp – anomaly maps and global mean surface temperature etc. I did consider abandoning the work I had done for this post in the light of this, however I decided to post it anyway in the hope it contributes to others’ understanding, as it has to mine.

References
1. Guam Hazard Mitigation Plan (2005)  Section 4.0 Island Description
2. Guam Hazard Mitigation Plan (2005)  Section 5.0 Risk Assessment (Figure 5-12. Airport and Ports Map)

Acknowledgement
Thanks to Peter O’Neill for the GIStemp files, plots and general useful discussions.