As we’ve found odd “discrepancies” in the Canadian data, between GISS, Wolfram Alpha, and Wunderground; so too are there some “odd things” about Hawaii. These “odd things” look like “selective listening skills” on the part of GHCN. Then again, they could just be differences in the “adjustment history” and / or Wolfram Alpha might have a systemic “bug”. (We find an example of what that “bug” might be down below).
The GHCN stations in Hawaii are basically the largest fastest growing airports. Those which went from “grass field to 10,000 foot concrete” in a few decades and surrounded by hundreds of acres of tarmac taxiways and parking lots on all sides. Some of the other stations in Hawaii are located at old military bases without a history of tourism growth or at smaller less used airports with “puddle jumper” service.
They have very different trends.
I was going to make this posting a pointed questioning of GHCN as being highly biased in their site selection… but with the questions about Wolfram Alpha data “purity”, I’ve now got to leave it as just a large “Dig Here!”. I’d love to know of an alternative way to validate the data from some of these lesser known Hawaii locations outside of GIStemp. (Preferably with a nice built in graphing and trend line function like Wolfram Alpha…)
At any rate, my hope is that the Wolfram Alpha “issue” with Canadian data is only a “Canada thing” and that it does not extend to the USA data. My fear is that I’ve uncovered a fundamental flaw that makes it useless as an A/B comparison with GIStemp.
On The Big Island of Hawaii there is a wonderful town of Hilo. It is one of the places where my spouse and I spent our Honeymoon. In a very nice hotel which faces the bay (from the OCEAN side, on a spit of land…) and gives a spectacular view of the water, cruise ships, and the main volcanic mountains.
A modestly sleepy town, away from the spectacular urbanization of Honolulu and Pearl Harbor. On the other far side of the island, they grow Kona Coffee, and just a ways down the beach the Kona resort area has blossomed. Why there? Well, it’s a dead dry desert region with lots of lava and not much else. Lousy for crops and not the place the Hawaiian Cowboys wanted to pasture cows. (Yes, there is a ranching area on The Big Island, complete with cowboys!) At any rate, truck in some sand to make a beach, pave over the oi oi lava, dump enough top soil to hold a palm tree, and you can make a very attractive “resort” that has 95% of the time under sunny skys as you sit on the desert side of the island. (Hilo gets rain some of the time so is lush and tropical and sometimes cloudy and rainy… I love it!)
As a consequence, the “tourism growth” was split with a large part of it going to Kona, not Hilo.
So what happens when you compare Honolulu International (where GHCN / GIStemp focus) to Hilo?
The GISS Stations
Here are the stations that GISS lists as used in GIStemp at this time. The link will give an easier to read list, but I’ve included a copy here as things with GISS tend to “change unexpectedly” so you must document things every single time you use something from the site… and save copies of every iteration as they will never be the same two months in a row…
GISS Surface Temperature Analysis
Station List Search: (19.4 N,156.5 W)
Stations are ordered by distance from center at (19.4 N,156.5 W). Click the “(*)” next to a station name and the list will be re-sorted by distance from that station.
Distance Station Name Lat Lon
ID Pop. Years
156 km (*) Puunene/Caa Airport 312 20.8 N 156.5
W 425911900010 < 10,000 1949 – 1958
156 km (*) Hilo/Gen. Lym 19.7 N 155.1
W 425912850000 38,000 1905 – 2011
158 km (*) Lanai City Lanai 20.8 N 156.9
W 425911890010 < 10,000 1930 – 1970
158 km (*) Lanai City Lanai 20.8 N 156.9
W 425911890011 < 10,000 1971 – 1980
162 km (*) Hana, Maui Hawaii 20.8 N 156.0
W 425911960010 < 10,000 1907 – 1960
162 km (*) Hana, Maui Hawaii 20.8 N 156.0
W 425911960011 < 10,000 1961 – 1980
164 km (*) Kahului Airpo 20.9 N 156.4
W 425911900000 17,000 1954 – 2011
165 km (*) Lahaina, Maui Hawaii 20.9 N 156.7
W 425911890020 < 10,000 1942 – 1980
201 km (*) Molokai, Molo 21.1 N 157.1
W 425911860000 < 10,000 1951 – 1981
203 km (*) Kualapuu, Molokai 21.2 N 157.0
W 425911860010 < 10,000 1905 – 1954
259 km (*) Honolulu, Oah 21.4 N 157.9
W 425911820000 836,000 1883 – 2011
260 km (*) Kaneohe Bay/Mcas 21.4 N 157.8
W 425911760010 37,000 1945 – 2003
265 km (*) Barbers Point/Nas 21.3 N 158.1
W 425911780020 836,000 1949 – 1999
266 km (*) Honolulu Obs Oahu 21.3 N 158.1
W 425911780010 836,000 1906 – 1980
409 km (*) Lihue, Kauai, 22.0 N 159.3
W 425911650000 < 10,000 1905 – 2011
414 km (*) Lihue Kauai 22.0 N 159.4
W 425911650010 < 10,000 1950 – 1963
446 km (*) Barking Sands 22.0 N 159.8
W 425911620010 < 10,000 1968 – 2003
1123 km (*) French Frigate Shoals Detache 23.9 N 166.3
W 425911550010 < 10,000 1952 – 1970
1407 km (*) Johnston Isla 16.7 N 169.5
W 530912750000 < 10,000 1949 – 1984
1761 km (*) Fanning Islan 3.9 N 159.4
W 504914870000 < 10,000 1922 – 1980
1944 km (*) Christmas Isl 2.0 N 157.5
W 504914900000 < 10,000 1969 – 1980
2039 km (*) Ship N 30.0 N 140.0
W 800999110010 < 10,000 1947 – 1974
2332 km (*) Midway Island 28.2 N 177.4
W 425910660000 < 10,000 1921 – 1991
2618 km (*) Malden Island 4.0 S 155.0
W 504919020000 < 10,000 1891 – 1919
2977 km (*) Kanton Island 2.8 S 171.7
W 504917010000 < 10,000 1937 – 1967
First thing to notice is that some of the stations you get when you click on Hawaii are up to 2977 km away. Out there even beyond Midway. GIStemp will reach up to 1200 km to get data to “fill in” missing bits of temperatures at a station, then it will use that modified ‘in-filled’ data to fill in a “grid / box” up to another 1200 km way. So I guess it’s appropriate that they show temperature locations 2400+ km away when you click on Hawaii….
Second thing to notice is that very few of these stations actually extend into 2011. I get 4 of them.
That last one is one of the OTHER places where we spent our Honeymoon. At a resort called something like the “Coco Palms” where we had a private bungalow with an outdoor “lava rock” tub / shower (with a high vine covered privacy wall) in a large area of coco palm trees… “Kings Cottage”? Something like that… the different rooms had different names.
Lihue, and Kauai in general, had a giant growth spurt from sleepy nowhere (as it was when “Puff The Magic Dragon” was written about the far side of the island…) to major tourist attraction. The airport has gone from “grass field” to major international airport in fairly short order. It’ has continued to grow in the last 1/4 century (after our “visit”).
Sidebar: The “average mean” for 2010 per Wunderground was 74 F or 23 C. Basically “middle of the pack” of this data / graph. Not much has changed even with the GISS machinations since 1920 or so when it was also 23 C in Hilo. There may have been a warm bump, but it’s gone now.
For Honolulu, Wunderground gives the 2010 “average mean” as 78 F or about 25.5 C. Oddly, it gives the Jan 1940 to Jan 1941 “average mean” as being 77 F or not very much different at all from today… Call it 25 C. Roughly a full degree C below the GIStemp graph for that time.
The Wunderground history page for Jan 2010 to Jan 2011 puts the “average mean” at 76 F ( about 24.4 C ) that by my eye is just about dead center of this data. Meaning that not much has really changed over the years. Any “slope” is most likely from the PDO cycle that it neatly samples from “low to high” and we’ve now turned back toward “low” and are leaving that 1997-98 top behind us now.
Overall, these graphs look mighty hot! And we see echos of a “step function” around the 1970 – 1976 era… except it looks like it happened on different islands at different times… Yet lately several of the stations have returned to near their longer term midpoints. To me, it mostly looks like the very large airports grew large Heat Islands and the smaller airports have short records of a partial “cold to hot” PDO cycle that is now turning cooler again.
Oddly, the Honolulu Observatory doesn’t have nearly as much “rise” in that time period. But then it gets dropped in the ’80s. I’m sure it’s just a coincidence…
Here are some Hawaii graphs from Wolfram Alpha. It has “another view”…
While I could not find this station in GIStemp, I did find it in Wunderground. They could not give me a 2010 average mean, but I did get the 1980 average mean as 58 F
Max Avg Min
Max Temperature 77 °F 68 °F 55 °F
Mean Temperature 72 °F 58 °F 43 °F
Min Temperature 71 °F 49 °F 29 °F
and while ’61, ’62, ’70, ’71, and ’72’ all claimed to have data (in that they were in the “drop down” menu selection) all declined to display any. But Jan ’73 to Jan ’74 DID give an average and a graph:
Max Avg Min
Max Temperature 82 °F 67 °F 28 °F
Mean Temperature 79 °F 62 °F 32 °F
Min Temperature 79 °F 60 °F 1 °F
As soon as you see a graph like that, with jumpy blocky shapes to the curve, you just have to think “data dropouts”.
The following year the graph looks a little better, but inspection of the data by month shows why things are so screwy. LOTS of missing data. If you click this link, you can scroll down to the data blocks and see the gaps in it.
So is the “issue” just that Wolfram Alpha is willing to make a graph of whatever it has, and is very uncritical of what “holes” there are in the data? I think, perhaps, yes…
I find Henderson Field interesting as it shows a “jump” when it looks like some work was done to the tarmac. I also find Bikini Atol interesting as a ‘bit further away’ station. So, how do you get such divergent “trends” from monotonically increasing CO2? How do you reconcile them with the GIStemp results? How is it that the stations rising most are the ones kept in GHCN? THE largest and fastest growing airports with the most tarmac…
I was hoping to pick up a “state trend” from the NOAA NCDC data set here:
But they seem to think that Alaska and Hawaii aren’t REAL states…
Then we have the “issue” that Wunderground will present you a “custom” history screen where you can put in long durations of time, but now they will “helpfully” limit the display to a single year after the start date you select.
It sure would be nice to have an alternative place to pick up “historical” charts…
But at least I can “spot check” a couple of years.
From this site:
We have annual means along with monthly and some other interesting data. No pretty graphs and no trends over time, but still useful. For some locations there are multiple sites / stations given. Right off the bat for Hilo I notice that “station to station” variation can easily be 1 F or more. How anyone thinks they can get 1/100 C out of data with 1 F error band for the same place is, er, “interesting”…
Mean-Average Max-Average Min-Average Max-Yr-Average Max-Yr Min-Yr-Average Min-Yr
72.9 80.0 65.8 73.7 57 72.5 53
HILO WSO AP 87
73.7 81.2 66.2 75.6 95 71.4 55
75.5 83.7 67.4 77.8 80 73.1 55
75.9 84.6 67.1 76.9 88 74.6 83
HONOLULU WB AIRPORT
77.0 84.0 70.1 79.3 95 74.4 55
HONOLULU WB CITY
75.3 79.8 70.7 76.1 61 73.9 55
73.6 80.8 66.4 74.3 59 72.5 55
LIHUE WSO AP 1020.1
75.4 81.1 69.7 77.3 84 73.6 65
No data found
Kaneoeh Bay Marine Corp Field
KANEOHE MAUKA 781
74.3 79.8 68.8 75.8 82 72.3 55
77.4 82.9 71.9 *****************
KANEOHE RANCH 838
72.1 78.1 66.1 *****************
Midway Henderson Field
No data found
OK, That’s a bag of worms
Looking at how thermometers that are more or less in “the same place” have a very different average year, and how the “coldest” or “warmest” average year bounces around quite a bit, makes it very clear that a different period of time or a different instrument, or even a different microclimate, can really change your “average”…
Wunderground Spot Checks
OK, I decided to try a “spot check” on a couple of stations. For Hilo, the Wuderground graph / page for 1995 (that the Hawaii.edu site says is the hottest average year at 75.6) Wunderground has it at 75 F “average mean” for the year from Jan 95 to Jan 96. Is 0.6 F “close enough”?
For the Jan 69 to Jan 70 year, it has “average mean” at 73 F.
Inspection of the Wolfram Alpha graph puts it at about 76 F to 77 F then dropping to about 74 F in ’95-6.
So which was valid for 1969? Was it 76 F or was it 73 F ; and did it drop to 74 F or rise to 75 F
Will the REAL Blue Hawaii please stand up?
I’m not sure what to make of this. Has it shown that the data in the USHCN / GHCN has been “preened, cleaned, and adjusted” to change history? Or has it shown that Wolfram Alpha has buggered data and will graph any old thing, holes and all? Or some of each?
I think we can clearly pick out that Massive Jet Age Airport Growth matters, rather a lot. It’s also pretty clear that all the data sets are pretty crappy.
I am a man with two watches wondering where I can get them calibrated so I know what time it might be..
Can anyone think of any reasonable way to “sort this out”? Are there any military temperature records for those military fields that might clarify what the actual recorded temperatures were? Is there anyplace ELSE to get a valid temperature history for comparison?
This is just a mess…