Like A Man Without A Home
UPDATE (10 Sept 2009):
While chasing down the WSMO (Weather Service Meteorological Observatory) at Guam, I stumbled on this interesting confirmation of the Airport Bias in the weather station record. I bolded a bit From:
Brief History of National Weather Service Offices Past and Present
The National Weather Service (known before 1970 as the Weather Bureau) has had many offices. Many of them evolved as the aviation industry expanded, supporting local airport observations. Many also had varying warning and forecast duties. However, in the 1990’s, the NWS was consolidated into 120+ offices, each with roughly the same duties.
The following is an attempt to list all past and present NWS/WB offices. If you have information on any of these, or ones that were missed, please contact Chris.Geelhart@noaa.gov . Several offices have more detailed histories available, which can be accessed by clicking on the city name.
Get a copy while you can, it may not stay “up” long if folks notice 8-)
While I still have not found the location of the WSMO, I did find this nifty infra red image of the place. You ought to be able to do a direct comparison of the land station and the surrounding “boxes” of water and decide for yourself how valid it is to say they are ‘the same’ as GIStemp does…
The nearest I’ve been able to find is a celebration picture posed on a grass field that claims to be at the WFO (Forecast Office). But notice the airplane tail in the background:
And this is a wonderful link to a site that is also trying to find the thermometer… but has linked in the Google Satellite view of where it ought to be (though notes that the LAT LON do not have enough accuracy to validate that view…) A very interesting site worth a visit:
At 13.4783 N 144.7945 E there is a building with a light blue roof. This looks a lot like the building with a light blue roof in the pdf linked by Ellie down in the comments, so I think it is the right place. Behind it is a path to a white box that looks to me to be about 10 feet on a side. It has potential. There is also a white square in front of it to the right side a little near the road fork and the requisite ‘side yard’ on the left with a collection of whitish things in it and what looks like a fence. Of these, I think I’d go with the side yard just because I’ve seen that a lot and the thing at the end of the path looks a bit large. It sure would be nice to have “eyes on the ground” take a look there. Know anybody in Guam?
The link Ellie gave shows many sites with an odd white “observatory dome”. Perhaps to protect equipment from hurricanes? If so, then that “10 feet on a side white box” might just be one of these domes. That would place the equipment near the airplane parking lot.
http://www.faa.gov/news/conferences_events/pacific_aviation/agenda/media/NWS.pdf is a decent read (and mostly pictures) and shows close ups of several of the Pacific Island sites. As Ellie pointed out, it also states that ALL the sites are going to be at airports. Well, at least now we know where the future warming is going to come from…
Why all the interest in Guam? Ellie, in the comments, found that there were two sites with a very large increase in temperature. These two sites will warm the boxes for 20 degrees of lat and long around them. That is an area about the size of the continental USA, or slightly larger than Europe (excluding Russia, who have always been rather ambiguous about their European nature…). The other one is the Marshall Islands, but the airport there so completely dominates everything that it is pretty clear what it is doing. For Guam, there are more places to hide the Stevenson Screen ;-)
Basically, we have two stellar examples of growth of aviation making a site warm, then using those sites to “warm” a 20 x 20 degree chunk of the globe. It would be very interesting to compare these two sites to a couple of islands “nearby” with no airports one sunny day…
So, it looks to me, like we have a very simple explanation for the “warming of the planet”. It is simply measuring the growth of Aviation and the expansion of Airports, with the Airport Heat Island effect, around the world; to islands and “underdeveloped” nations post WWII, and with growth and development in the industrialized states.
Original Islands In The Sun
Well, I decided to do a bit of a “peek ahead” into the last land step – STEP3.
This was partly because I’d gotten some flack from Warmistas that “the anomaly would save them” or that “Grids and Boxes were immune to thermometer count changes!”. They have had to slowly retreat as each STEP was shown to “have issues” that lead to warming based on the code and process, not on the data. They were now standing on this last island of code and tossing rocks. OK, then that is where I go next.
It was also partly because in the Pisa thread Peter O’Neill had gotten ahead of me so I thought I ought to do a “look ahead” just to stay competitive, if nothing else.
Peter is doing a great job and will be producing important results (and more formally packaged than mine) so watch for his work to be published. But I’m not ready to abandon the field, so I decided to “plough the field ahead” a bit and see if I could turn up something new (even if Peter may run ahead of me next week ;-)
Finally, there was also the issue of “STEP2 begins the anomaly steps” (it uses anomalies as part of the UHI “correction” ) and I wanted to see how much commonality there was to the code (a fair amount was ‘re-hacked’ from earlier steps and shows it) and partly because it also does a swap of data format from fixed format CHARACTERS to a more free form FORTRAN specific binary. That makes some of my “quick” benchmarking not-so-quick in that I need to write programs to read the binary format files and, frankly, I’d become a bit saturated on that front and needed a diversion to “recharge”.
So I did a “peek ahead”.
Up until now, we’ve watched the data warm as The March Of The Thermometers heads south. We’ve seen airports used as pristine “rural stations” for UHI adjustment when they are not. We’ve seen GIStemp add about 1/2 C of warming to the data in the first two STEPs. We have also seen that the older and more stable records do not show warming, while new younger records carry all the signal. And we’ve seen that the 6 zones they use in the zonal step of STEP2 are just way too few to get any decent protection from The March Of The Thermometers. Then Peter turned up the heat by showing that the UHI “Adjustment” was sensitive to exactly which stations make it into the UHI “Adjustment” batch to the tune of 0.4 C for Pisa. (That is, take out mountain stations over 900 M and the bogus “cooling” of the past for Pisa changes from -1.4C to -1C based on that alone.)
And no, the “anomaly” processing in STEP2 did not save the AGW thesis nor did it save GIStemp from the observation that the 1/10C place is sensitive to changes in exactly what thermometer records are used for UHI “adjustment” to the tune of a bit shy of 1/2C. (And those records are basically randomly chosen. The selection criteria are not as rational as one would like and distance it not sufficiently controlled. Hot sub-urban airports from 1000 km away can be “nearby rural”.) So the 1/10 C position of the GIStemp data are basically randomly chosen since they are derived from this process.
But “Grids And Boxes And Anomalies – Oh My” will save it all! Was the chant from the Warmistas. So I took a look.
And the Warmistas are looking more and more like “A Man Without A Home”…
Floating On A Log
My first step is just to examine the code. Then I examine the log files. The “code review” was uneventful. More of the same. Semi-cryptic stuff built as a “hand tool” with lots of parameters for cherry picking and tuning. Clear evidence of making the program fit the data. (Things like using large zones so the program doesn’t dump everything into the “not enough data to fudge together some in-fill” bucket. They use an “80 Region for the whole globe” pass THEN fill in the smaller 100 “sub-boxes” in the Region. Dig Here!)
Then I look in the log files.
It didn’t take long. On the 3rd ASCII log file I starting hitting “issues”. That file is:
To decode that name a bit, we can see that it takes the temperature data (smeared over 6 equal 30 degree latitude bands in STEP2, with a variety of infilling, stretching, and homogenizing already done) and tries to pin it into a SBBX “sub-box” on a grid; the data have a cut off in the cold period of 1880 and is based on the GHCN set (with amendments and modifications) and the STEP2 PA anomaly process was applied; while in this step the “Reference Station Method” look around was stretched out to 1200 km in the hopes of finding a station, any station, that could be used to fill in missing boxes.
That the file name encodes the “tweaking” done to parameters is an interesting testimony to the hand tuning that was done. They needed to keep straight which run produced what output. It tells you the parameters that were cherry picked, in a way. (Another “Dig Here”).
So how well did they do?
Well, pretty good, I guess… The code looks like it uses an 80 x 100 matrix for boxes. That’s about 8000 of them. (If I read the code right, it’s a set of 80 “regions” from N to S pole each with 100 “sub boxes” in in. A gross size to enable more “spread” if needed, then smaller boxes for detail, if available.)
It manages to fill in a lot of them, even though we know the ‘data’ filled in are a complete fiction since the early steps have less than 8000 thermometers for the whole planet; and the lions share of them are concentrated in Europe and North America (as we saw in looking at The March Of The Thermometers; the Southern Hemisphere has a lot of “Big Empty”…). So at the end of this, how many boxes are still “left blank”? “Only” 1/8 of them:
$ grep "NO STATIONS" to.SBBXgrid.1880.GHCN.CL.PA.1200.log | wc -l 1026 $
Though there are a bit over 300 that have no data in the baseline period but do have some data now. (Dig Here! One could also do a general “quality metric” for grid boxes based on some sort of “stations used” vs grid box and “station-months” vs grid box. It would also be interesting to look into those boxes with more stations to see what impact comes from adding stations over time… )
Not too bad, I guess! So we probably have about 1/4 of the planet decently instrumented. About 1/4 “so-so”. That only leaves about 1/4 for “poor” and 1/4 for “OMG” bad (with only half of them still being completely hopeless) after all the torturing of the data done so far.
I took a look into the log file and noticing that the further down the list I got into Southern Hemisphere boxes, the more I ran into “NO STATION” flags. So it looks like there are plenty of boxes that can get a new thermometer in the Southern Hemisphere and continue to contribute to warming the “Northern Hemisphere winter” on a global average basis.
FWIW, the exact log entries look like this:
REGION 56 134849 182 STATIONS USED LAT,LON,STN-MNTHS,STNS,IDS -4325-17250 2143 5 939870000 934360010 932920000 935460000 933730000 LAT,LON,STN-MNTHS,STNS,IDS -4325-16950 779 2 939870000 932920000 LAT,LON,STN-MNTHS,STNS,IDS -4325-16650 237 1 939870000 LAT,LON,STN-MNTHS,STNS,IDS -4325-16350 237 1 939870000 NO STATIONS FOR CENTER 57 7 NO STATIONS FOR CENTER 57 8 NO STATIONS FOR CENTER 57 9 NO STATIONS FOR CENTER 57 10 LAT,LON,STN-MNTHS,STNS,IDS -4093-17250 2824 6 939870000 933090000 934360010 932920000 935460000 933730000 LAT,LON,STN-MNTHS,STNS,IDS -4093-16950 1135 3 939870000 932920000 933730000 LAT,LON,STN-MNTHS,STNS,IDS -4093-16650 237 1 939870000 LAT,LON,STN-MNTHS,STNS,IDS -4093-16350 237 1 939870000 NO STATIONS FOR CENTER 57 17 NO STATIONS FOR CENTER 57 18 NO STATIONS FOR CENTER 57 19 NO STATIONS FOR CENTER 57 20 LAT,LON,STN-MNTHS,STNS,IDS -3869-17250 3290 7 939870000 933090000 934360010 939940001 932920000 933730000 931120000 LAT,LON,STN-MNTHS,STNS,IDS -3869-16950 1135 3 939870000 932920000 933730000 LAT,LON,STN-MNTHS,STNS,IDS -3869-16650 237 1 939870000 NO STATIONS FOR CENTER 57 26 NO STATIONS FOR CENTER 57 27 NO STATIONS FOR CENTER 57 28 NO STATIONS FOR CENTER 57 29 NO STATIONS FOR CENTER 57 30 LAT,LON,STN-MNTHS,STNS,IDS -3652-17250 1971 5 939870000 939940001 932920000 933730000 931120000 LAT,LON,STN-MNTHS,STNS,IDS -3652-16950 1375 3 939870000 939940001 932920000 LAT,LON,STN-MNTHS,STNS,IDS -3652-16650 237 1 939870000 NO STATIONS FOR CENTER 57 36 NO STATIONS FOR CENTER 57 37 NO STATIONS FOR CENTER 57 38 NO STATIONS FOR CENTER 57 39 NO STATIONS FOR CENTER 57 40 LAT,LON,STN-MNTHS,STNS,IDS -3441-17550 3811 8 939870000 933090000 934360010 939940001 932920000 930120000 933730000 931120000 LAT,LON,STN-MNTHS,STNS,IDS -3441-17250 1971 5 939870000 939940001 932920000 933730000 931120000 LAT,LON,STN-MNTHS,STNS,IDS -3441-16950 892 1 939940001 NO STATIONS FOR CENTER 57 45 NO STATIONS FOR CENTER 57 46 NO STATIONS FOR CENTER 57 47 NO STATIONS FOR CENTER 57 48 NO STATIONS FOR CENTER 57 49
I’ve left this log file ‘truncated right’ rather than letting it wrap and strip out blanks. The exact station ID’s off the right edge are not so important just yet.
The first line is the summary for the prior “region” number 56 showing the totals used in it. The rest of the entry is for the next “region” of 57 as we fill in the sub-boxes.
For now, just notice that there are a LOT of boxes with no stations…
This is as you step through the LAT / LON for box locations.
Now if you look at one WITH data, you will see an embedded “side header” that says the record consists of the latitude and longitude (without decimal or degree marks), the “station months” of data that go into that box value, the number of thermometers that contribute to that box, then the list of thermometers (Station IDs) used to make that box value.
OK, so what?
Well, first I noticed that several of the boxes have a “1” for thermometer count. A SINGLE thermometer for the whole BOX. So much for multiple thermometers with gridding, averaging, and boxing smoothing out any anomalies from the particular locations! (“Dig Here!” It would be very interesting to rank the boxes by that thermometer count and look for patterns. How many in each rank. Geographic asymmetry. Percent airports. Etc.)
Then I noticed that a lot of these started with the same STATION ID.
My but that gets around a lot, I think. Wonder who it is?
A brief “grep” in v2.inv file pulls up the record:
50793987000 CHATHAM ISLAN -43.95 -176.57 49 0R -9HIxxCO 1A-9WATER A 0
and it is flagged as an airport (the “A’ at 1A-9WATER)
Hmmm. Add an airport on an island and your Airport Heat Island effect can warm 1200 x 1200 x PI square kilometers around. That is 4,523,893.344 square km or 1,746,684.268 square miles. Take just a moment to look at those numbers again. 1.7 Million square miles or 4.5 Million square km all controlled by one little box near the runway… and there are thousands of islands to choose from.
Do this where the box was not an airport in prior years, you can warm a great number of boxes. All those boxes warmed with both The March Of The Thermometers to warmer climates and The Airport UHI Correction “issues”. Now look at the ones with more than a single Station ID. The second station is often the same. Even when there are multiple records, they may simply be using 2 Airports to make a box rather than one… The “box and grid” does not get rid of the Airport bias.
After all of about 10 seconds pause, I pondered Diego Garcia. A tropical paradise that was part of The British Empire from time to time but has recently had a bunch of big airplanes added. Wonder what’s happening there?
A “grep” Is A Terrible Thing To Waste – So I Didn’t
(The Unix / Linux command to find text in a file is the “grep” command, that stands for “Globally search with a Regular Expression and Print”)
16161967000 DIEGO GARCIA -7.30 72.40 3 0R -9FLxxCO 1A-9WATER C 17
So are there any records in the log file for STATION ID 61967000?
LAT,LON,STN-MNTHS,STNS,IDS -1507 7537 291 1 619670001 LAT,LON,STN-MNTHS,STNS,IDS -1507 7762 291 1 619670001 LAT,LON,STN-MNTHS,STNS,IDS -1507 7987 291 1 619670001 NO STATIONS FOR CENTER 52 37 NO STATIONS FOR CENTER 52 38 LAT,LON,STN-MNTHS,STNS,IDS -1507 8662 288 1 969960000 LAT,LON,STN-MNTHS,STNS,IDS -1507 8887 288 1 969960000 LAT,LON,STN-MNTHS,STNS,IDS -1271 6862 1618 3 619860003 619880003 619670001 LAT,LON,STN-MNTHS,STNS,IDS -1271 7087 887 2 619880003 619670001 LAT,LON,STN-MNTHS,STNS,IDS -1271 7312 291 1 619670001 LAT,LON,STN-MNTHS,STNS,IDS -1271 7537 291 1 619670001 LAT,LON,STN-MNTHS,STNS,IDS -1271 7762 291 1 619670001 LAT,LON,STN-MNTHS,STNS,IDS -1271 7987 291 1 619670001 NO STATIONS FOR CENTER 52 47 NO STATIONS FOR CENTER 52 48 LAT,LON,STN-MNTHS,STNS,IDS -1271 8662 288 1 969960000 LAT,LON,STN-MNTHS,STNS,IDS -1271 8887 288 1 969960000 LAT,LON,STN-MNTHS,STNS,IDS -1037 6862 1618 3 619860003 619880003 619670001 LAT,LON,STN-MNTHS,STNS,IDS -1037 7087 291 1 619670001 LAT,LON,STN-MNTHS,STNS,IDS -1037 7312 291 1 619670001 LAT,LON,STN-MNTHS,STNS,IDS -1037 7537 291 1 619670001 LAT,LON,STN-MNTHS,STNS,IDS -1037 7762 291 1 619670001 LAT,LON,STN-MNTHS,STNS,IDS -1037 7987 291 1 619670001 LAT,LON,STN-MNTHS,STNS,IDS -1037 8212 291 1 619670001 NO STATIONS FOR CENTER 52 58 LAT,LON,STN-MNTHS,STNS,IDS -1037 8662 288 1 969960000 LAT,LON,STN-MNTHS,STNS,IDS -1037 8887 288 1 969960000 LAT,LON,STN-MNTHS,STNS,IDS -805 6862 291 1 619670001 LAT,LON,STN-MNTHS,STNS,IDS -805 7087 291 1 619670001 LAT,LON,STN-MNTHS,STNS,IDS -805 7312 291 1 619670001 LAT,LON,STN-MNTHS,STNS,IDS -805 7537 291 1 619670001 LAT,LON,STN-MNTHS,STNS,IDS -805 7762 291 1 619670001 LAT,LON,STN-MNTHS,STNS,IDS -805 7987 291 1 619670001 LAT,LON,STN-MNTHS,STNS,IDS -805 8212 291 1 619670001 NO STATIONS FOR CENTER 52 68 NO STATIONS FOR CENTER 52 69 LAT,LON,STN-MNTHS,STNS,IDS -805 8887 288 1 969960000 LAT,LON,STN-MNTHS,STNS,IDS -574 6862 291 1 619670001
As a SMALL sample. And notice how many of THEM have a 1 for station count…
So we search for that Station Id in the log file, then count how many lines that is with the unix “wc -l” command (“word count, but count whole lines only”).
$ grep 619670001 to.SBBXgrid.1880.GHCN.CL.PA.12 | wc -l 64 $
So this says 64 distinct lines have that STATION ID number in them. That is a LOT of boxes… Now multiply THAT behaviour by all the Tropical Pacific Islands…
Nothing Wrong with A Pristine Tropical Island Paradise
Nope, not at all. If only these were such. Now Chatham Island is more a temperate Island Paradise near New Zealand. And it is not too bad; away from the airport it comes close, except it is a commercial airport.
And it has grown over time. From:
The grass landing-field at Hapupu, at the northern end of the Island, proved a limiting factor, as few aircraft apart from the Bristol Freighter had both the range to fly to the islands and the ruggedness to land on the grass airstrip. Although other aircraft did use the landing field occasionally, they would often require repairs to fix damage resulting from the rough landing. Hapupu is also the site of the JM Barker (Hapupu) National Historic Reserve (one of only two in New Zealand) where there are momori rakau (Moriori tree carvings).
In 1991, after many years of requests by locals and the imminent demise of the aging Bristol Freighter aircraft, the construction of a sealed runway at Karewa, Tuuta Airport, allowed more modern aircraft to land safely. The Chathams’ own airline, Air Chathams, now operates services to Auckland on Thursdays, Wellington on Mondays, Wednesdays and Fridays and Christchurch on Tuesdays. The timetable varies seasonally, but generally planes depart the Chathams around 10.30 am (Chathams Time) and arrive in the mainland around noon. Then they refuel and reload, depart again at around 1 pm back to the Chathams. Air Chathams operates twin turboprop Convair 580 aircraft in combi (freight and passenger) configurations and Fairchild Metroliners.
It looks to me like the temperature series is more or less dead flat through about 1980, though it becomes a bit more volatile in the ’60s through ’80s with higher highs and more low lows, then the bottom end gets pulled up after the airport move. Maybe warm tarmac keeping nighttime lows up a 1/2 degree or so or warming the day by a degree so the minimum average is higher? It would not take much to move the low range of that “average of averages” up by that 1/2 C.
More dramatic is Diego Garcia that has had rapid and major expansion into a world wide crossroads for Military Heavy Lift aircraft.
Here is one fellow’s memory book of his time there:
And a bit more “formal” look at the place:
And I’m sure you have all noticed when on vacation to tropical “paradise” that the tarmac gets much hotter under the tropical sun than it does “up north”. Direct angle of illumination. 12 hours of it. Little seasonal relief. etc. Just as I’m sure you’ve also noticed how much cooler it is under the palm trees near the jungle or near the ocean.
From these two semi-randomly selected samples we have a temperate airport warming the cold southern band open water 1200 km south; and we have a tropical military base airport warming open ocean all around it. Kind of makes you wonder what Guam and Palau are doing… (“Dig Here!”)
ISLAND Heat Island Effect?
So, do we need IHI too? The ability to plant a jet airport on some tiny rock in the middle of the ocean and warm dozens of surrounding grid cell boxes of open water? Sure looks like it to me…
And I’m pretty sure that all those remote tropical island getaways spots that have sprouted airports since WWII did not have them in 1900 …
Now take a look at a globe. Notice that the Tropical Pacific has a lot more islands than anywhere else? Nature has a bias here. There are not very many islands in the “circumpolar cold band” near Antarctica. There are not very many islands north of Hawaii and south of Alaska. There are not very many islands in the Cold Atlantic. Etc. So nature has a bias toward lots more islands in tropical places; and STEP3 will have taken that bias and run with it. And a big “Dig Here” would be verifying that most of those island records come from Airports. (My sampling says so, but an exhaustive chart and graph would be authoritative.)
So my first “dip of a toe” into the last “land” step, STEP3, finds a continuation of prior themes. Data spread too thin and too far to where there are none. Missing data “filled in” by using lots of Airports; “rural” means near the tarmac and the jet exhaust. Southern Hemisphere still slowly being “filled in” as The March Of The Thermometers continues…
And no, Virginia, Grids, Zones, and Boxes have most decidedly NOT prevented a single hot thermometer record from warming lots of surrounding “turf” even if it is open water… Though I’m left to wonder how much “near the tarmac Diego Garcia Military / Airport” is representative of “over the water 1200 km away”…
The only step after this is STEP4_5, where “4” just lets you get an updated version of the Hadley SST anomaly map and “5” just merges it with the output of STEP3. So the buggering of the “land” data is complete at this point, even when “land” looks to be boxes of nearby sea…
I have not bothered to look into how Hadley “re-imagine” their SST data, but given the history of lost original data, obfuscation, resistance to FOIA requests, non-publishing of methods, etc. I suspect that there is not much of merit to find there.
This “first look” shows “islands” are surprisingly important to STEP3 (just as cold mountains play a big role in Pisa in STEP2).
It sure looks to me like we need a “Station Survey” of islands, looking to check the integrity of those stations and just how many of them are at airports and too near the tarmac, hangers, and terminal buildings. (Of course, this also means one must do “due diligence” and find a proper alternate site to put the thermometer… One near the palm trees and not too far from the beach. Maybe near a thatched cabana… It would be very important to monitor the location for at least 24 hours; and to prevent “warming bias” from your physical presence, many cold beverages will need to be applied ;-) Any volunteers? 8-)