Mysterious Madagascar being Mischievous?
The original image is available as free download for wallpaper at:
Maybe it’s Madagascar that Marks Manipulation
OK, the GHCN Apologists have said it’s all just a misunderstanding. GHCN was just ‘made once’ from a whole lot of manual labor looking at paper copies and,well, some countries just don’t make their data available in a nice fast electronic form. So it was not a deliberate act to leave them out of the data set going forward, it is just a matter of the unavailability of ‘real time’ data. And that is why all those places ‘cut off’ in 1990. Just an accident of when the dataset was created.
Yet we have places like Madagascar…
Wunderground has no problem finding Madagascar, and real time too.
That station in the Wunderground link, Majunga, is also listed in the ‘station inventory’ file of GHCN, so they knew about it and used it in the past.
12567009000 DIEGO-SUAREZ -12.35 49.30 105 225S 40HIxxCO 5A 5WATER A 0 12567019000 ANALALAVA -14.63 47.77 57 77R -9HIxxCO 3A-9WARM GRASS/SHRUBA 0 12567025000 ANTALAHA -14.88 50.25 88 48S 18FLxxCO 1A10WATER A 0 12567027000 MAJUNGA -15.67 46.35 18 6U 66FLxxCO 4A 5WATER A 0 12567073000 MAINTIRANO -18.05 44.03 23 10R -9FLxxCO 2A-9WARM GRASS/SHRUBA 0 12567083000 ANTANANARIVO/ -18.80 47.48 1276 1352U 452MVxxno-9A10TROP. SAVANNA B 0 12567095000 TAMATAVE -18.12 49.40 6 25U 77HIxxCO 1A 1WATER C 10 12567143000 MANANJARY -21.20 48.37 6 4S 15HIxxCO 1A 1WARM FOR./FIELD A 0 12567161000 TULEAR -23.38 43.73 8 19S 46HIxxCO 2A 5WARM GRASS/SHRUBA 0 12567197000 FORT-DAUPHIN -25.03 46.95 9 100S 14HIxxCO 2A 2WATER A 0
So I’m left with a mad muse of Madagascar, marveling at the mystery…
For what it’s worth, here is the temperature history of the basic thermometer data from Madagascar in GHCN:
Look at ./Temps/Temps.125.yrs.GAT (Y/N)? y Thermometer Records, Average of Monthly Data and Yearly Average by Year Across Month, with a count of thermometer records in that year -------------------------------------------------------------------------- YEAR JAN FEB MAR APR MAY JUN JULY AUG SEPT OCT NOV DEC YR COUNT -------------------------------------------------------------------------- 1889 21.3 22.7 20.6 20.0 18.3 14.8 14.2 13.8 16.7 18.1 20.7 21.0 18.5 1 1890 21.3 21.3 19.9 18.5 14.6 12.3 12.7 13.7 15.7 17.5 19.0-99.0 17.0 1 1891 19.6 20.8-99.0 18.9 16.5 15.1 13.3 13.7 16.6 18.9 19.9 21.0 17.7 1 1892 20.4 20.1 20.1 19.7 16.4 14.1 13.6 15.8 15.5 19.4 20.6 20.4 18.0 1 1893 20.2 20.3 19.8 18.4 16.6 13.4 13.3 13.4 16.6 18.9 19.8 20.8 17.6 1 1894 19.9 20.6 20.7 18.9 17.1 13.8 14.4 14.8 17.4 20.1 21.2 20.7 18.3 1 1895 20.4 21.0 20.5 19.7 16.4 14.1 14.7 14.7-99.0-99.0-99.0-99.0 17.7 1 1896 21.4 20.9 21.4-99.0-99.0-99.0 13.6 14.1 17.6 19.6 19.1 20.7 18.7 1 1897 20.7 20.2 21.1 18.9 17.2 15.5 15.5 15.2 16.9 19.3 21.4 20.3 18.5 1 1898 20.4 20.8 20.3 19.0 17.6 14.4 14.1 14.6 17.6 20.6 22.2 21.1 18.6 1 1899 21.5 20.2 20.2 19.8 16.2 13.9 13.4 13.4 16.5 19.8 21.6 20.9 18.1 1 1900 21.1 21.2 21.1 19.7 17.2 15.1 14.0 15.6 17.9 20.2 22.2 21.6 18.9 1 1901 22.1 21.2 21.3 19.4 17.3 15.5 14.8 14.4 16.6 19.6 22.8 20.8 18.8 1 1902 22.3 21.4 20.5 20.5 18.0 15.5 15.5 17.1 19.1 20.3 22.0 21.5 19.5 1 1903 21.5 21.1 21.4 20.3 18.5 16.6 14.7 15.6 16.0 19.1 21.2 21.2 18.9 1 1904 20.7 21.0 20.9 19.4 18.1 15.2 14.2 15.6 16.3 20.7 22.6 21.8 18.9 1 1905 21.7 21.2 20.6 18.9 18.3 17.0 14.7 15.9-99.0 20.9 20.9 21.2 19.2 1 1906 22.6 21.6 21.9 20.3 17.8 15.5 14.7 15.0 17.0 19.5 21.1 21.2 19.0 1 1907 21.4 21.3 20.4 19.8 17.8 15.7 13.6 15.4 17.7 21.5 21.6 20.7 18.9 1 1908 21.7 21.0 21.5 20.0 17.0 16.3 14.8 16.1 18.4 21.5 22.2 21.0 19.3 1 1909 21.1 21.2 20.9 19.8 17.1 14.4 15.7 15.4 18.4 19.1 21.8 22.7 19.0 1 1910 21.6 21.4 20.9 20.2 17.2 14.7 14.5 15.6 17.8 20.6 19.9 20.5 18.7 1 1911 21.2 20.5 20.5 18.5 17.0 15.1 15.1 15.4 17.7 20.6 21.2 22.5 18.8 1 1912 22.1 20.9 21.9 20.3 18.4 15.9 14.9 16.0 19.0 20.7 22.7 21.5 19.5 1 1913 20.8 20.9 21.7 18.9 18.7 15.3 14.3 16.4 18.0 21.3 21.3 20.9 19.0 1 1914 21.0 21.9 21.1 20.2 17.9 16.3 14.6 16.0 18.2 20.3 20.7 22.4 19.2 1 1915 21.7 21.6 21.2 20.7 18.1 15.5 15.6 14.8 17.2 21.4 21.6 21.4 19.2 1 1916 21.6 21.0 21.3 20.1 17.4 15.4 13.4 15.7 17.4 20.2 20.2 19.9 18.6 1 1917 20.7 19.9 20.3 19.5 18.5 16.0 15.5 15.5 17.0 18.2 20.9 20.1 18.5 1 1918 21.7 20.7 20.7 19.1 17.3 15.9 14.4 15.0 17.5 20.7 21.4 21.6 18.8 1 1919 21.6 21.5 21.3 18.9 16.4 15.4 13.8 16.7 19.0 20.2 22.4 22.0 19.1 1 1920 21.3 21.0 20.8 18.8 17.8 14.4 13.6 15.7 19.0 21.4 21.0 20.7 18.8 1 1921 21.0 21.0 20.3 18.9 17.4 15.3 14.7 15.5 17.8 20.9 21.2 20.6 18.7 1 1922 20.5 20.9 20.9 18.0 17.2 14.8 13.7 15.4 16.8 19.7 21.5 21.8 18.4 1 1923 22.0 21.5 21.6 20.4 18.4 15.6 15.6 14.9 18.4 20.6 22.2 21.6 19.4 1 1924 21.6 21.2 21.8 19.5 18.2 15.7 15.4 16.0 18.0 20.0 21.3 21.0 19.1 1 1925 21.2 21.2 20.7 20.3 19.9 17.0 16.0 15.2 19.4 20.8 21.3 21.6 19.6 1 1926 22.4 22.0 22.4 21.3 18.6-99.0 15.6 16.4 19.1 21.0 22.7 21.9 20.3 1 1927 21.5 21.6 20.7 20.0 17.9 15.8 14.2 16.1 18.0 20.7 21.5 22.3 19.2 1 1928 22.7 21.7 21.6 21.0 18.7 15.4 14.9 15.8 17.8 20.9 21.9 21.4 19.5 1 1929 22.1 20.1 22.1 20.5 17.8 16.4 15.7 15.9 18.0 21.7 22.1 21.0 19.5 1 1930 20.7 21.8 20.9 19.9 16.6 15.1 14.5 14.7 19.2 20.0 22.0 21.7 18.9 1 1932 21.6 20.9 21.2 19.0 16.4 15.3 14.8 15.5 16.8 19.4 21.9 21.1 18.7 1 1933 21.4 20.6 20.3 19.6 17.4 15.1 14.7 15.4 16.2 20.1 22.0 21.2 18.7 1 1934 21.0 21.0 21.3 19.8 17.7 14.0 14.6 14.1 16.5 19.5 21.1 20.5 18.4 1 1935 21.2 20.9 21.4 19.3 18.2 15.6 13.3 15.8 18.3 20.4 20.7 21.0 18.8 1 1936 21.1 22.2 21.5 19.9 18.0 15.0 14.5 14.9 17.4 21.1 21.6 21.0 19.0 1 1937 21.0 21.4 20.6 19.8 17.2 16.2 14.6 15.2 18.7 19.7 22.8 21.6 19.1 1 1938 21.7 21.4 21.5 20.3 18.2 15.0 15.4 16.0 17.6 19.9 20.3 21.9 19.1 1 1939 21.7 21.7 21.0 19.9 18.1 15.9 15.2 15.0 18.0 20.3 21.6-99.0 18.9 1 1941 24.0 24.9 25.0 24.5 22.6 21.2 20.0 20.4 21.8 23.1 24.3 25.3 23.1 3 1942 25.5 25.8 25.5 24.1 22.2 20.6 19.5 20.3 21.3 21.9 23.9 24.7 22.9 3 1943 24.6 24.4 24.4 24.2 22.3 20.3 19.4 20.1 19.8 22.1 23.6 24.4 22.5 3 1944 24.6 25.3 24.6 23.6 22.2 21.3 19.8 20.2 21.5 22.7 22.9 25.0 22.8 3 1945 25.1 24.5 25.1 24.4 22.3 19.9 19.9 20.1 21.4 23.6 24.4 24.4 22.9 3 1946 24.8 25.4 24.7 23.9 22.4 20.6 20.1 20.4 20.7 23.2 24.2 24.8 22.9 3 1947 25.0 25.1 24.3 24.0 22.2 20.6 20.7 20.1 21.1 22.8 24.2 25.2 22.9 3 1948 25.1 25.0 24.9 24.5 23.3 21.2 20.0 20.1 21.3 22.9 23.9 24.1 23.0 3 1949 25.0 24.4 24.4 23.4 22.2 20.6 19.9 20.4 21.4 23.2 24.0 25.2 22.8 3 1950 25.1 24.4 24.5 24.1 21.5 20.5 19.7 19.9 20.7 22.6 24.0 25.0 22.7 3 1951 25.6 25.5 25.4 25.0 23.3 21.8 21.2 22.1 23.1 24.9 25.7 26.4 24.2 8 1952 26.4 26.6 26.6 25.7 24.6 22.5 21.9 22.3 23.3 24.3 24.9 26.1 24.6 8 1953 26.7 26.4 26.3 25.7 24.1 22.9 21.5 21.7 22.7 24.2 25.7 26.4 24.5 8 1954 26.4 26.8 26.1 25.5 24.3 22.0 21.7 22.3 23.3 24.5 25.6 26.3 24.6 8 1955 26.8 26.4 25.9 25.1 23.6 22.0 21.8 21.5 22.8 23.8 25.2 26.0 24.2 9 1956 25.9 26.2 26.4 25.2 23.6 22.2 20.9 21.8 22.9 24.0 25.0 25.8 24.2 9 1957 26.6 25.7 26.0 25.3 23.8 22.2 21.5 22.0 23.6 24.8 25.8 26.5 24.5 10 1958 26.6 26.6 26.1 25.9 23.5 22.1 21.6 22.2 23.3 24.1 25.5 26.1 24.5 10 1959 26.2 26.4 25.9 24.7 23.7 21.9 21.5 22.0 22.8 23.8 25.5 26.3 24.2 10 1960 26.0 26.3 25.7 25.1 23.6 22.5 21.1 21.8 23.1 24.3 25.0 26.0 24.2 10 1961 26.3 26.6 26.3 26.0 24.4 22.4 21.9 21.9 22.8 24.5 25.6 25.8 24.5 10 1962 26.2 26.1 25.8 25.1 22.7 21.3 20.9 21.6 22.5 23.7 25.2 25.8 23.9 10 1963 25.9 25.7 25.7 25.1 22.6 21.6 21.1 20.6 22.0 23.7 24.8 25.4 23.7 10 1964 25.9 26.4 25.9 24.9 22.5 21.3 20.2 20.8 21.9 23.3 24.6 25.3 23.6 10 1965 25.1 25.5 24.9 24.4 22.5 20.6 20.8 21.4 22.4 23.4 24.8 25.4 23.4 10 1966 26.1 26.2 25.6 25.0 23.5 22.2 21.5 21.6 23.0 23.4 24.6 25.7 24.0 10 1967 26.0 26.4 26.2 25.4 23.9 22.3 21.8 21.7 22.6 24.1 25.1 25.2 24.2 10 1968 25.9 25.8 25.6 24.4 22.8 21.2 21.2 21.6 22.1 23.6 24.9 25.6 23.7 10 1969 26.6 26.6 27.1 26.3 24.4 21.8 21.9 21.8 22.3 24.8 25.7 26.4 24.6 10 1970 26.1 26.6 25.9 24.7 23.5 22.1 21.7 22.1 22.7 23.8 25.2 25.6 24.2 10 1971 25.9 25.9 25.9 25.6 23.6 21.6 21.7 21.4 22.3 23.8 24.7 25.7 24.0 10 1972 26.0 25.6 25.3 25.4 24.4 21.7 21.9 22.0 22.6 24.2 25.0 26.3 24.2 10 1973 26.3 26.2 26.4 25.3 23.9 22.3 21.5 21.8 22.8 24.4 25.2 25.5 24.3 10 1974 26.0 25.6 25.6 25.1 23.6 21.8 21.3 21.3 22.0 23.5 24.9 25.8 23.9 10 1975 25.9 25.8 25.5 25.3 23.6 22.0 21.4 21.3 22.2 23.3 24.9 25.2 23.9 10 1976 25.5 25.8 25.9 25.3 23.8 22.3 21.5 21.7 22.7 24.0 25.6 26.0 24.2 10 1977 26.4 25.9 25.9 24.8 24.1 22.5 21.8 22.2 23.2 24.6 25.4 26.4 24.4 10 1978 26.7 26.6 26.0 25.2 23.8 22.3 21.3 22.2 23.2 24.5 25.2 25.8 24.4 10 1979 26.0 26.3 26.2 24.9 23.5 21.6 21.2 21.7 23.0 24.3 25.3 26.0 24.2 9 1980 25.9 26.4 26.1 25.4 23.3 21.5 21.5 21.7 22.9 24.3 25.8 26.0 24.2 10 1981 26.9 26.2 25.9 25.1 23.2 21.3 20.8 21.5 22.7 23.8 25.3 25.9 24.1 10 1982 26.1 26.0 25.7 24.8 23.1 22.1 21.4 21.3 22.3 23.5 25.0 25.6 23.9 10 1983 26.0 26.7 26.5 25.3 23.8 22.4 21.5 21.2 22.1 23.6 25.0 25.7 24.2 9 1984 26.0 25.7 25.5 24.6 22.9 21.2 20.5 20.6 21.8 23.5 24.8 25.8 23.6 10 1985 26.3 26.0 25.6 24.7 23.1 21.5 21.2 21.8 22.7 23.6 24.7 25.8 23.9 10 1986 26.1 26.2 25.7 25.1 23.2 20.6 20.2 21.4 21.9 23.8 25.1 25.9 23.8 10 1987 25.7 26.0 26.1 25.1 23.3 20.6 21.3 21.7 22.7 24.0 24.9 26.4 24.0 10 1988 26.8 26.3 26.4 25.8 23.5 22.2 21.7 21.8 22.7 24.6 25.0 26.0 24.4 10 1989 25.8 25.5 25.7 24.7 23.9 21.8 21.3 21.7 22.6 24.0 25.2 25.8 24.0 10 1990 26.0 26.2 26.0 25.4 23.7 22.3 20.7 21.8 22.3 23.9 24.7 26.2 24.1 10 1991 26.7 26.2 26.0 25.0 23.7 21.3 21.1-99.0-99.0-99.0-99.0-99.0 24.3 10 1992 26.2 25.6 26.5 25.6 24.1-99.0-99.0-99.0-99.0-99.0-99.0-99.0 25.6 10 1993-99.0-99.0-99.0-99.0-99.0 22.5-99.0-99.0 21.7-99.0-99.0-99.0 22.1 10 1994-99.0-99.0-99.0-99.0-99.0-99.0 20.5-99.0-99.0 23.8 24.6 25.2 23.5 10 1995 24.0-99.0 25.5 24.6 23.1 20.8 18.7 21.3 22.3-99.0 25.3 25.2 23.1 9 1996 25.7 26.0 25.2 24.6 23.1 21.4 20.4 22.5-99.0 23.7 25.4 25.2 23.9 7 1997 25.4 25.6 25.9 25.1 23.3 21.7 20.2 18.2 23.3 23.9-99.0 26.3 23.5 8 1998-99.0 26.8 26.9 24.3 22.9 21.9 21.2 21.5 21.6 23.0 24.8 25.2 23.6 7 1999 25.7 26.0-99.0 24.7 22.6 20.9-99.0 20.3 22.6 24.2 24.9 27.1 23.9 8 2000 26.1 24.7 25.8 25.4 23.7 22.3 20.5 22.7 22.7 23.3 26.0 23.9 23.9 9 2001 25.8 27.3 26.7 25.5 23.5-99.0-99.0-99.0-99.0 23.9-99.0-99.0 25.4 8 2002-99.0-99.0 25.4-99.0 22.7 20.9 21.5 20.4 21.8 23.1 25.9 25.1 23.0 8 2003 25.9 25.8 26.3 24.8 23.8 21.3 21.1 19.8 21.5 24.4 25.4 25.7 23.8 7 2004 27.3 25.7 25.7 24.6 22.4 19.5 20.7 21.3 22.8 24.1 24.2 27.3 23.8 6 2005 26.1 27.6 25.4 25.8 23.8 21.8 19.8-99.0-99.0-99.0-99.0-99.0 24.3 6 At 25.8 25.8 25.6 24.8 23.2 21.4 20.9 21.2 22.3 23.7 24.9 25.5 23.8 Am 23.8 23.8 23.6 22.6 20.8 18.8 18.1 18.5 20.2 22.1 23.3 23.7 21.6 For Country Code 125 From input file /v/GHCN.Dec09/v2.mean
We again see the ‘typical’ pattern of one early thermometer in a place with a different pattern from the larger area (in this case, colder). It is fairly stable at about 18.5 C to 19 C with some “ripple” over decade time scales. That ought to be this thermometer:
12567083000 ANTANANARIVO/ -18.80 47.48 1276 1352U 452MVxxno-9A10TROP. SAVANNA B 0
Antananarivo at about 1300 m of elevation with savanna as the surrounding landscape, though it is an Urban area. This record also says it is an Airport. Even with that, the basic data show no net warming.
Graph of Temperatures at Antananarivo
Then, as the record is ‘fleshed out’ with more thermometers (one presumes in warmer places) the average moves to about 24 C through to the present. There there is a slight drop toward the end as 3 or 4 thermometers drop from the averages.
Notice also the final data ‘cut off’ in 2005. Not in 1990
So I think this is an ‘existence proof’ of dropping thermometers after the 1990 creation date. As I find more cases like this, I’ll put them up, too, so we can see if there is any interesting ‘pattern’ to the data. Like this data, where the ‘baseline’ period of 1951-1980 used in GIStemp looks to be about 24.5 C while the near term data is about 23.5C (though 2005 jumps to 25.4 C largely on the back of data dropouts [those -99.0 flags] letting a couple of hot months dominate the average.
But I’m sure it’s just an accident of history, or something, that an entire country with no visible warming in the basic data and a recent cooling was dropped in 2005…. Maybe the monkeys hid the data …
What does NASA think of Madagascar?
Well, the GISS anomaly map ( 250 km radius of ‘spread’, December last). And a baseline from 1991 to 2006) has an empty Madagascar just to the right of the Southern end of Africa, but with bright yellow and orange “anomalies” all around it:
Dec 2009 anomaly map compared to 1991-2006 baseline
I think looking at other ‘grey’ areas will be ‘instructive’ too ;-)
FWIW, here is the default view of the world from GISS:
NASA GISS Anomaly map Dec 2009 Default 1200 km 'fill in'
From “No Madagascar” to “Burning Up Madagascar” in one easy step… I also find it interesting how the purple blobs in the USA and Canada shrink too. Then there is central Africa that goes from a substantially “missing in action” to “yellow /orange flood” in one easy in-fill. Must be that baseline cherry pick of GISS in the middle of a cold period.
By Latitude
Not really a lot to see here. We start in the middle and spread out each direction:
Look at ./Lats/Therm.by.lat125.Dec.LAT (Y/N)? y
Year SP -28 -26 -24 -22 -20 -18 -16 -14 -12 -NP
DecPct: 1889 0.0 0.0 0.0 0.0 0.0 100.0 0.0 0.0 0.0 0.0 100.0
DecPct: 1899 0.0 0.0 0.0 0.0 0.0 100.0 0.0 0.0 0.0 0.0 100.0
DecPct: 1909 0.0 0.0 0.0 0.0 0.0 100.0 0.0 0.0 0.0 0.0 100.0
DecPct: 1919 0.0 0.0 0.0 0.0 0.0 100.0 0.0 0.0 0.0 0.0 100.0
DecPct: 1929 0.0 0.0 0.0 0.0 0.0 100.0 0.0 0.0 0.0 0.0 100.0
DecPct: 1949 0.0 0.0 25.0 0.0 0.0 50.0 0.0 0.0 25.0 0.0 100.0
DecPct: 1959 0.0 0.0 13.6 12.9 3.6 32.9 0.0 23.6 13.6 0.0 100.0
DecPct: 1969 0.0 0.0 10.2 10.2 9.1 29.9 0.0 30.5 10.2 0.0 100.0
DecPct: 1979 0.0 0.0 10.6 12.8 6.6 28.2 0.0 29.1 12.8 0.0 100.0
DecPct: 1989 0.0 0.0 8.9 14.0 8.4 29.6 0.0 25.7 13.4 0.0 100.0
DecPct: 1999 0.0 0.0 11.1 11.1 8.3 32.4 0.0 25.9 11.1 0.0 100.0
DecPct: 2005 0.0 0.0 13.6 13.6 0.0 36.4 0.0 22.7 13.6 0.0 100.0
For COUNTRY CODE: 125
Even at the end all we lose is the part a bit south of the middle.
By Altitude
A graphic example (pardon the pun ) of thermometer locations running from the mountains to the beach.
From 100% at altitude to mostly at the beach
Look at ./Alts/Therm.by.Alt125.Dec.ALT (Y/N)? y
Year -MSL 20 50 100 200 300 400 500 1000 2000 Space
DAltPct: 1889 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0100.0 0.0 0.0
DAltPct: 1899 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0100.0 0.0 0.0
DAltPct: 1909 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0100.0 0.0 0.0
DAltPct: 1919 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0100.0 0.0 0.0
DAltPct: 1929 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0100.0 0.0 0.0
DAltPct: 1949 25.0 0.0 0.0 25.0 0.0 0.0 0.0 0.0 50.0 0.0 0.0
DAltPct: 1959 55.7 12.9 10.7 13.6 0.0 0.0 0.0 0.0 7.1 0.0 0.0
DAltPct: 1969 49.7 10.2 20.3 10.2 0.0 0.0 0.0 0.0 9.6 0.0 0.0
DAltPct: 1979 53.3 10.6 18.5 12.8 0.0 0.0 0.0 0.0 4.8 0.0 0.0
DAltPct: 1989 53.6 8.9 16.8 13.4 0.0 0.0 0.0 0.0 7.3 0.0 0.0
DAltPct: 1999 52.8 11.1 14.8 11.1 0.0 0.0 0.0 0.0 10.2 0.0 0.0
DAltPct: 2005 54.5 9.1 9.1 13.6 0.0 0.0 0.0 0.0 13.6 0.0 0.0
For COUNTRY CODE: 125
Interesting. Here, again, we see the importance of High Cold Places. The early cold history comes directly from that one very high altitude thermometer above 1000 m up to 2000 m.
Looking at the annual detail for the last few years we can see that the ongoing thermometer losses in the last two decades were again raising the percentage at altitude.
Year -MSL 20 50 100 200 300 400 500 1000 2000 Space
DAltPct: 1989 53.6 8.9 16.8 13.4 0.0 0.0 0.0 0.0 7.3 0.0 0.0
ALT pct: 1990 51.7 10.3 20.7 10.3 0.0 0.0 0.0 0.0 6.9 0.0 0.0
ALT pct: 1991 50.0 10.0 20.0 10.0 0.0 0.0 0.0 0.0 10.0 0.0 0.0
ALT pct: 1992 50.0 10.0 20.0 10.0 0.0 0.0 0.0 0.0 10.0 0.0 0.0
ALT pct: 1993 50.0 10.0 20.0 10.0 0.0 0.0 0.0 0.0 10.0 0.0 0.0
ALT pct: 1994 50.0 10.0 20.0 10.0 0.0 0.0 0.0 0.0 10.0 0.0 0.0
ALT pct: 1995 55.6 11.1 11.1 11.1 0.0 0.0 0.0 0.0 11.1 0.0 0.0
ALT pct: 1996 57.1 14.3 0.0 14.3 0.0 0.0 0.0 0.0 14.3 0.0 0.0
ALT pct: 1997 62.5 12.5 0.0 12.5 0.0 0.0 0.0 0.0 12.5 0.0 0.0
ALT pct: 1998 57.1 14.3 0.0 14.3 0.0 0.0 0.0 0.0 14.3 0.0 0.0
ALT pct: 1999 50.0 12.5 12.5 12.5 0.0 0.0 0.0 0.0 12.5 0.0 0.0
DAltPct: 1999 52.8 11.1 14.8 11.1 0.0 0.0 0.0 0.0 10.2 0.0 0.0
ALT pct: 2000 44.4 11.1 22.2 11.1 0.0 0.0 0.0 0.0 11.1 0.0 0.0
ALT pct: 2001 50.0 12.5 12.5 12.5 0.0 0.0 0.0 0.0 12.5 0.0 0.0
ALT pct: 2002 50.0 12.5 12.5 12.5 0.0 0.0 0.0 0.0 12.5 0.0 0.0
ALT pct: 2003 57.1 14.3 0.0 14.3 0.0 0.0 0.0 0.0 14.3 0.0 0.0
ALT pct: 2004 66.7 0.0 0.0 16.7 0.0 0.0 0.0 0.0 16.7 0.0 0.0
ALT pct: 2005 66.7 0.0 0.0 16.7 0.0 0.0 0.0 0.0 16.7 0.0 0.0
DAltPct: 2005 54.5 9.1 9.1 13.6 0.0 0.0 0.0 0.0 13.6 0.0 0.0
For COUNTRY CODE: 125
I find 2004 particularly interesting. We drop to what looks like 4 thermometer records at the beach and one each at 100-200m and that lone high altitude location. Even concentrating 2/3 of the thermometers at the beach was just not enough… And with that colder thermometer being the Capitol and largest city: it would be missed if it was left out. Not exactly a remote high cold mountain top… This Yahoo travel guide page even has an annual typical temperature graph on the left hand edge with a current temperature ticker. 79 F and Cloudy as I type. . Guess Yahoo can find the temperatures too…
Wunderground places the thermometer at the airport and says it was 68F at 2:30 AM Friday, their last update as I’m typing. The Yahoo temp matches the forecast highs at Wunderground within a degree or so. I’d guess that Yahoo is giving the daily high (though given their time zone, it might be ‘live’ relative to where I am).
In any case, it does not seem very hard to get the Madagascar temperatures. At least, if you really wanted to get them…
Temperature History of Antananarivo
Just for grins, here is the temperature data from GHCN. Since this is just one station, the ‘monthly averages’ are an average of just the one value for that month. That is, the data itself.
Look at ./Temps/Temps.12567083.yrs.GAT (Y/N)? y Thermometer Records, Average of Monthly Data and Yearly Average by Year Across Month, with a count of thermometer records in that year -------------------------------------------------------------------------- YEAR JAN FEB MAR APR MAY JUN JULY AUG SEPT OCT NOV DEC YR COUNT -------------------------------------------------------------------------- 1889 21.3 22.7 20.6 20.0 18.3 14.8 14.2 13.8 16.7 18.1 20.7 21.0 18.5 1 1890 21.3 21.3 19.9 18.5 14.6 12.3 12.7 13.7 15.7 17.5 19.0-99.0 17.0 1 1891 19.6 20.8-99.0 18.9 16.5 15.1 13.3 13.7 16.6 18.9 19.9 21.0 17.7 1 1892 20.4 20.1 20.1 19.7 16.4 14.1 13.6 15.8 15.5 19.4 20.6 20.4 18.0 1 1893 20.2 20.3 19.8 18.4 16.6 13.4 13.3 13.4 16.6 18.9 19.8 20.8 17.6 1 1894 19.9 20.6 20.7 18.9 17.1 13.8 14.4 14.8 17.4 20.1 21.2 20.7 18.3 1 1895 20.4 21.0 20.5 19.7 16.4 14.1 14.7 14.7-99.0-99.0-99.0-99.0 17.7 1 1896 21.4 20.9 21.4-99.0-99.0-99.0 13.6 14.1 17.6 19.6 19.1 20.7 18.7 1 1897 20.7 20.2 21.1 18.9 17.2 15.5 15.5 15.2 16.9 19.3 21.4 20.3 18.5 1 1898 20.4 20.8 20.3 19.0 17.6 14.4 14.1 14.6 17.6 20.6 22.2 21.1 18.6 1 1899 21.5 20.2 20.2 19.8 16.2 13.9 13.4 13.4 16.5 19.8 21.6 20.9 18.1 1 1900 21.1 21.2 21.1 19.7 17.2 15.1 14.0 15.6 17.9 20.2 22.2 21.6 18.9 1 1901 22.1 21.2 21.3 19.4 17.3 15.5 14.8 14.4 16.6 19.6 22.8 20.8 18.8 1 1902 22.3 21.4 20.5 20.5 18.0 15.5 15.5 17.1 19.1 20.3 22.0 21.5 19.5 1 1903 21.5 21.1 21.4 20.3 18.5 16.6 14.7 15.6 16.0 19.1 21.2 21.2 18.9 1 1904 20.7 21.0 20.9 19.4 18.1 15.2 14.2 15.6 16.3 20.7 22.6 21.8 18.9 1 1905 21.7 21.2 20.6 18.9 18.3 17.0 14.7 15.9-99.0 20.9 20.9 21.2 19.2 1 1906 22.6 21.6 21.9 20.3 17.8 15.5 14.7 15.0 17.0 19.5 21.1 21.2 19.0 1 1907 21.4 21.3 20.4 19.8 17.8 15.7 13.6 15.4 17.7 21.5 21.6 20.7 18.9 1 1908 21.7 21.0 21.5 20.0 17.0 16.3 14.8 16.1 18.4 21.5 22.2 21.0 19.3 1 1909 21.1 21.2 20.9 19.8 17.1 14.4 15.7 15.4 18.4 19.1 21.8 22.7 19.0 1 1910 21.6 21.4 20.9 20.2 17.2 14.7 14.5 15.6 17.8 20.6 19.9 20.5 18.7 1 1911 21.2 20.5 20.5 18.5 17.0 15.1 15.1 15.4 17.7 20.6 21.2 22.5 18.8 1 1912 22.1 20.9 21.9 20.3 18.4 15.9 14.9 16.0 19.0 20.7 22.7 21.5 19.5 1 1913 20.8 20.9 21.7 18.9 18.7 15.3 14.3 16.4 18.0 21.3 21.3 20.9 19.0 1 1914 21.0 21.9 21.1 20.2 17.9 16.3 14.6 16.0 18.2 20.3 20.7 22.4 19.2 1 1915 21.7 21.6 21.2 20.7 18.1 15.5 15.6 14.8 17.2 21.4 21.6 21.4 19.2 1 1916 21.6 21.0 21.3 20.1 17.4 15.4 13.4 15.7 17.4 20.2 20.2 19.9 18.6 1 1917 20.7 19.9 20.3 19.5 18.5 16.0 15.5 15.5 17.0 18.2 20.9 20.1 18.5 1 1918 21.7 20.7 20.7 19.1 17.3 15.9 14.4 15.0 17.5 20.7 21.4 21.6 18.8 1 1919 21.6 21.5 21.3 18.9 16.4 15.4 13.8 16.7 19.0 20.2 22.4 22.0 19.1 1 1920 21.3 21.0 20.8 18.8 17.8 14.4 13.6 15.7 19.0 21.4 21.0 20.7 18.8 1 1921 21.0 21.0 20.3 18.9 17.4 15.3 14.7 15.5 17.8 20.9 21.2 20.6 18.7 1 1922 20.5 20.9 20.9 18.0 17.2 14.8 13.7 15.4 16.8 19.7 21.5 21.8 18.4 1 1923 22.0 21.5 21.6 20.4 18.4 15.6 15.6 14.9 18.4 20.6 22.2 21.6 19.4 1 1924 21.6 21.2 21.8 19.5 18.2 15.7 15.4 16.0 18.0 20.0 21.3 21.0 19.1 1 1925 21.2 21.2 20.7 20.3 19.9 17.0 16.0 15.2 19.4 20.8 21.3 21.6 19.6 1 1926 22.4 22.0 22.4 21.3 18.6-99.0 15.6 16.4 19.1 21.0 22.7 21.9 20.3 1 1927 21.5 21.6 20.7 20.0 17.9 15.8 14.2 16.1 18.0 20.7 21.5 22.3 19.2 1 1928 22.7 21.7 21.6 21.0 18.7 15.4 14.9 15.8 17.8 20.9 21.9 21.4 19.5 1 1929 22.1 20.1 22.1 20.5 17.8 16.4 15.7 15.9 18.0 21.7 22.1 21.0 19.5 1 1930 20.7 21.8 20.9 19.9 16.6 15.1 14.5 14.7 19.2 20.0 22.0 21.7 18.9 1 1932 21.6 20.9 21.2 19.0 16.4 15.3 14.8 15.5 16.8 19.4 21.9 21.1 18.7 1 1933 21.4 20.6 20.3 19.6 17.4 15.1 14.7 15.4 16.2 20.1 22.0 21.2 18.7 1 1934 21.0 21.0 21.3 19.8 17.7 14.0 14.6 14.1 16.5 19.5 21.1 20.5 18.4 1 1935 21.2 20.9 21.4 19.3 18.2 15.6 13.3 15.8 18.3 20.4 20.7 21.0 18.8 1 1936 21.1 22.2 21.5 19.9 18.0 15.0 14.5 14.9 17.4 21.1 21.6 21.0 19.0 1 1937 21.0 21.4 20.6 19.8 17.2 16.2 14.6 15.2 18.7 19.7 22.8 21.6 19.1 1 1938 21.7 21.4 21.5 20.3 18.2 15.0 15.4 16.0 17.6 19.9 20.3 21.9 19.1 1 1939 21.7 21.7 21.0 19.9 18.1 15.9 15.2 15.0 18.0 20.3 21.6-99.0 18.9 1 1941 20.1 21.4 21.5 20.2 17.5 16.0 14.6 15.6 18.1 20.0 20.9 21.7 19.0 1 1942 21.6 21.8 21.7 19.9 17.1 15.3 14.3 15.6 17.9 18.4 21.0 21.3 18.8 1 1943 20.9 20.6 20.6 20.1 17.2 15.0 14.0 15.4 15.3 18.9 20.8 21.2 18.3 1 1944 21.0 21.5 20.6 19.2 17.5 17.0 14.3 15.4 17.8 18.9 21.5 21.2 18.8 1 1945 22.2 20.9 21.6 20.7 17.3 15.0 14.3 15.3 17.6 21.4 20.9 21.1 19.0 1 1946 21.5 22.0 20.9 19.7 17.7 15.3 15.2 15.7 16.0 20.2 20.7 21.6 18.9 1 1947 22.0 22.1 20.5 20.3 17.8 15.5 16.0 14.7 17.3 19.5 20.7 21.9 19.0 1 1948 21.6 21.6 21.3 20.8 19.1 16.0 15.2 15.4 17.2 19.5 21.1 20.7 19.1 1 1949 22.0 21.8 21.2 19.5 17.7 15.5 14.9 16.2 18.1 20.6 21.6 22.0 19.3 1 1950 22.0 20.7 20.6 19.6 16.4 15.1 13.9 15.3 16.0 19.3 21.1 21.2 18.4 1 1951 19.8 19.3 20.0 19.2 17.3 14.8 15.2 16.3 16.6 19.0 19.7 20.4 18.1 1 1952 20.4 21.0 21.0 19.4 17.7 14.4 14.4 15.0 16.6 18.0 18.8 20.1 18.1 1 1953 20.4 20.4 20.7 19.4 17.2 16.9 15.1 14.5 16.2 17.9 20.8 21.0 18.4 1 1954 20.5 21.3 20.4 19.4 17.6 15.6 15.0 15.2 16.2 19.6 19.9 20.7 18.4 1 1955 21.1 21.1 20.4 18.5 17.0 14.7 14.9 15.6 16.8 19.4 20.5 20.6 18.4 1 1956 20.1 19.7 20.3 18.9 16.0 14.6 14.4 15.0 16.8 19.0 21.0 21.0 18.1 1 1957 21.5 20.3 20.3 18.9 16.8 15.6 14.4 15.3 17.1 19.0 21.2 21.0 18.5 1 1958 21.1 21.3 21.3 20.1 16.5 14.9 14.8 15.6 17.6 18.3 20.4 20.7 18.6 1 1959 20.9 21.0 20.4 18.7 17.2 15.4 13.9 14.9 16.7 18.0 20.8 21.4 18.3 1 1960 21.3 20.7 19.6 18.9 17.6 15.7 14.4 14.6 15.6 17.4 18.8 19.8 17.9 1 1961 21.1 21.8 21.0 20.6 19.1 16.4 15.4 15.2 16.1 19.6 21.0 20.9 19.0 1 1962 21.6 21.2 20.6 19.3 16.4 14.9 14.1 14.8 16.3 17.5 20.2 21.0 18.2 1 1963 20.8 20.9 20.4 20.1 16.5 15.3 14.8 14.2 16.1 19.7 19.4 20.6 18.2 1 1964 21.0 21.8 20.6 19.2 16.3 15.3 13.9 14.3 15.9 17.6 20.4 20.0 18.0 1 1965 19.9 20.6 19.4 18.8 16.1 14.9 14.4 15.1 16.5 17.9 19.5 20.4 17.8 1 1966 20.8 20.8 19.7 20.0 18.5 15.4 14.3 14.9 17.6 18.0 20.2 20.9 18.4 1 1967 21.2 21.3 21.3 19.9 17.8 15.7 15.6 15.0 17.0 19.2 19.8 20.4 18.7 1 1968 20.3 20.6 20.4 19.2 17.2 15.2 14.1 14.7 16.4 18.5 19.8 20.3 18.1 1 1969 21.0 21.8 21.5 21.3 18.6 15.1 15.1 14.4 17.7 20.4 21.3 21.9 19.2 1 1970 21.3 22.6 21.5 19.6 17.6 15.5 15.2 14.7 16.2 19.3 20.9 20.5 18.7 1 1971 20.1 20.9 20.1 19.4 17.8 14.2 14.2 13.7 16.0 18.8 19.4 19.9 17.9 1 1972 20.0 19.7 20.0 19.2 17.6 14.3 14.9 15.8 16.5 18.5 20.1 20.5 18.1 1 1973 20.1 20.4 20.3 18.6 17.1 15.1 14.1 14.1 16.9 18.5 20.3 19.3 17.9 1 1974 19.6 19.4 18.9 19.0 17.4 14.3 13.7 14.3 15.7 17.8 19.8 19.9 17.5 1 1975 20.3 19.8 19.6 19.3 16.8 14.5 14.1 13.4 15.1 17.6 18.9 19.5 17.4 1 1976 18.8 19.9 19.5 19.2 17.2 14.7 14.7 15.5 16.5 18.4 19.7 20.0 17.8 1 1977 20.4 19.9 20.2 18.9 17.7 15.3 14.1 15.7 16.1 19.5 19.7 20.6 18.2 1 1978 21.4 21.2 19.3 19.0 16.7 14.3 14.3 14.2 17.8 19.1 19.3 20.0 18.1 1 1979 20.4 20.0 20.1 18.5 16.5 14.0 13.8 14.8 17.3 19.1 19.7 20.4 17.9 1 1980 20.5 20.7 20.5 19.7 16.8 15.1 14.1 13.9 16.2 18.8 20.0 20.0 18.0 1 1981 20.8 20.7 20.4 18.8 16.5 14.9 13.6 15.5 16.4 19.0 19.9 20.4 18.1 1 1982 20.1 20.6 20.2 19.4 16.9 15.5 15.5 15.1 16.4 18.2 19.8 19.9 18.1 1 1983 20.4 21.5 20.9 19.3 17.5 15.5 14.6 14.7 16.1 18.7 20.6 20.0 18.3 1 1984 20.4 20.4 19.8 18.5 16.7 14.2 12.9 14.0 16.8 18.0 19.5 20.4 17.6 1 1985 20.7 20.4 19.9 18.5 16.5 14.6 14.3 14.4 15.9 16.8 19.2 20.1 17.6 1 1986 20.5 20.6 19.7 19.4 16.6 13.0 13.2 14.5 16.1 18.1 19.4 20.2 17.6 1 1987 20.5 21.0 20.9 19.5 16.9 14.2 15.4 14.8 16.7 19.2 19.4 21.6 18.3 1 1988 21.0 20.4 20.3 19.6 16.0 15.5 14.4 14.8 15.8 18.7 19.4 20.0 18.0 1 1989 20.3 20.0 20.2 18.6 16.5 15.0 14.5 14.2 16.4 18.7 19.5 20.2 17.8 1 1990 20.9 21.2 20.8 20.1 17.2 15.2 13.5 15.4 17.3 19.4 19.6 21.1 18.5 1 1991 21.5 21.0 20.6 19.4 17.9 15.1 13.6-99.0-99.0-99.0-99.0-99.0 18.4 1 1992 21.1-99.0 21.9 19.7 17.7-99.0-99.0-99.0-99.0-99.0-99.0-99.0 20.1 1 1993-99.0-99.0-99.0-99.0-99.0 16.0-99.0-99.0 16.0-99.0-99.0-99.0 16.0 1 1994-99.0-99.0-99.0-99.0-99.0-99.0 13.0-99.0-99.0 19.2 20.4 21.1 18.4 1 1995 20.7-99.0 21.1 19.9 17.5 15.5 14.4 15.1 17.4-99.0-99.0 20.9 18.1 1 1996 20.7 20.4 20.9 20.1 17.5 15.3 14.3-99.0-99.0 18.0 20.9 20.5 18.9 1 1997 20.3 20.6 21.0 20.0 17.6 15.3 14.1 15.4-99.0 18.0-99.0 20.9 18.3 1 1998-99.0 21.6 21.5 19.6 17.6 15.4 13.7 15.1 16.6 18.6 19.9 20.3 18.2 1 1999 20.3 21.1-99.0 18.8 17.5 14.9-99.0 14.6 16.8 18.2 19.1-99.0 17.9 1 2000 20.8 19.9 19.6 19.6 17.7 14.7 14.3-99.0 15.9 18.8-99.0 20.5 18.2 1 2001 20.9-99.0 22.0 19.7 17.5-99.0-99.0-99.0-99.0 18.5-99.0-99.0 19.7 1 2002-99.0-99.0 20.7-99.0 17.3 14.5 14.7 16.1 16.8 18.5-99.0 20.6 17.4 1 2003 20.4 20.7 20.8 19.7 18.9 14.9-99.0 14.6 16.5 19.4 20.3 20.3 18.8 1 2004-99.0 20.6 20.0 19.3 16.4 14.1 14.0 15.1 18.6-99.0 19.5-99.0 17.5 1 2005 21.3-99.0 21.2-99.0-99.0 15.6 14.0-99.0-99.0-99.0-99.0-99.0 18.0 1 At 20.9 20.9 20.7 19.5 17.3 15.1 14.4 15.0 17.0 19.3 20.6 20.8 18.5 Am 21.0 21.0 20.7 19.5 17.4 15.2 14.5 15.1 17.1 19.4 20.7 20.9 18.5 For Country Code 12567083 From input file /v/GHCN.Dec09/v2.mean
Remarkably devoid of warming.
Geek Corner
Oh, and as an amusing “math geek” point: We get to see how doing even simple math in a computer program can make the ‘final digit’ a dodgy digit.
I keep two “grand total averages”. The At is an Average of all the Temps. The Am is an Average of the Monthly averages. That is, total all the temperatures and divide or total all the averages and divide. In this case, the two ought to be the same, yet the 1/10 place has ‘jitter’ in it. There is a 1/10 C ‘warming bias’ in the “average of averages”, probably due to a rounding artifact at some point (that added ‘divide’ step for each value as, for example, 20.6 is divided by 1 to give 20.6). As things get turned from base 10, to binary, binary math done, then back to base ten and sometimes with type conversions (REAL to DOUBLE to INTEGER to…) the ‘low order bits’ can drift. And over a very large number of additions or divisions, move your final digit by one or two…
While I might go track it down, the point is fairly simple. You get a 1/10 C change in your average temperatures some times, just due to the way you choose to do the math. Even if you think you are being very very careful… And that, boys and girls, is why I don’t trust the rats nest of code that is GIStemp to say ANYTHING of value in the 1/10 C position. I’ve already found one location inside GIStemp where it has a 1/10 C jitter from “order of math” issues. Are you willing to bet the world economy that there are not a couple of more?
OT but I don’t know where else to put it yet
I’d completely missed the little story covered here:
http://boballab.wordpress.com/2010/02/01/the-goat-ate-the-data/
It would seem that New Zealand had their adjustment ‘data’ lost too, and by a CRU crew alumni… The story also has a description of folks who also looked at the ‘raw’ N.Z. data and found no warming. I’ll take that as a confirmation of my findings that N.Z. was flat if you didn’t have: Campbell Island in the baseline but missing in the present. (A cold island much closer to the pole. Biases the baseline down, making the present, where it is missing, warmer in comparison.) Or maybe I’m confirming their findings. Whatever. New Zealand is Just Fine, thank you very much, as long as it isn’t adjusted, homogenized, and baseline diddled…
I think that the story in that link ought to ‘have legs’. So take a look.
Musings from the Chiefio 
One could be forgiven for thinking that the NCDC believe that everyone is stupid and won’t check things like this.
I’m looking forward to seeing what patterns emerge..
Sombody made a few comments on Ireland below. Ireland has no high points but it sure has some cool ones
http://www.boards.ie/vbulletin/showpost.php?p=64141209&postcount=7
man,
Put all those tables in links or charts or someting. For some reason I just cant read through all this stuff and I just throw up my hands in desparation and miss the whole point.
@Moshier:
Hey, I’ve got more graphical content than when I started!
And I’m working on getting the ability to turn the tables into graphs. But there is only so much I can do so fast. I’ve spent the last week “building up” the box that will be used to do it (now a “dual boot” Windows and Linux box) and all that remains is to put a recent version of “Office” or something related on it.
So if you don’t like the tables, just skip them, they are largely there just so folks can see I’m not just making up the words I type. I also do think it is important to actually show the data. Others can grab it and make many interesting charts, and some folks can see patterns in the data that charts hide.
Or, put more bluntly: You want graphs? Buy me a copy of MS Office. Otherwise you get to wait a few weeks until someone else makes a graph or I get to the store and buy the needed “goods” and install ’em.
Oh, and “the point”: Pretty simple, really. AGW advocates are saying “GHCN was a one shot in 1990. We don’t drop thermometers any more.” This shows they dropped Madagascar. It also shows changes of locations reported after 1990 and before the 2005 drop date and that those follow the ‘usual pattern’ of heading for the beach.
i’m lost after the first table. which station is that.. then this
We again see the ‘typical’ pattern of one early thermometer in a place with a different pattern from the larger area (in this case, colder). It is fairly stable at about 18.5 C to 19 C with some “ripple” over decade time scales. That ought to be this thermometer:
12567083000 ANTANANARIVO/ -18.80 47.48 1276 1352U 452MVxxno-9A10TROP. SAVANNA B 0
Antananarivo at about 1300 m with savanna.
REPLY: [ The station name is ANTANANARIVO and it is the capitol of Madagascar. (The station ID is 1256708300, with a 0 modification flag). It is located at 1276 m altitude or 1352 m altitude from map estimates. It is at LAT 18.80 S and LON 47.48 East in a land type called ‘Savanna” even though it is the capitol city. If you look at the ‘next to right most’ column of the temperature table with the title “YR” you will see that the annual average temperature reading does not change much, and what change there is slowly rises and falls by small amounts in a periodic fashion. ( I could even add ‘and it is an airport’, based on the “A” at -9A10TROP but that would be a bit much…) -E.M.SMITH ]
hmm.
http://data.giss.nasa.gov/cgi-bin/gistemp/findstation.py?lat=-18.8&lon=47.48&datatype=gistemp&data_set=1
Pingback: More climate shenanigans « Jim’s Blog
EM instead of M$ Office have you thought of using Open Office? Open Office can use the same file system as Office but it has the added benefits of being open source and free (Open Office is what I use and seems to work fine).
http://www.openoffice.org/
@boballab:
Last time I tried it was ‘a while ago’ and it was a bit buggy.
But if you assure me that it works, and I can suck in a couple of lines of numbers and make a line graph, I’d rather us OO then M$ O any day.
My whole purpose is just to be able to turn that set of “annual temperature by year” next to the last column on the right into a line graph (and things of similar lack of complexity…).
I just don’t want to spend a month trying to get it to work only to find out that it crashes when I click on “make graph” ;-)
Well believe it or not I’m a relative newbie to spreadsheet programs. Never used one until this past November and I had a choice between the one that came with the “box” install of an old MS Works program (This is an old computer, with an even older processor [AMD Semperon in a Compaq Presario Walmart Special :) ]). I decided to give the Open Office program a try and so far so good, haven’t pushed the bounderies of the spreadsheet program yet as in how many columns it can take, but if you want to see what the graphs turn out like, head over to the blog I started just for my own ramblings, I got some posted up there.
Also from the the OO site they are beta testing the newest release canidate iteration now, 3.2 I believe. The other advantage is that you can download a ton of differenet packages for the programs from them.
Chiefio,
You keep talking about warmer-colder places when all climate data is reporting anomalies which means it should not matter whether the place is warmer or colder as long as the change in average is the same.
I realize that it may not be reasonable to assume that different places experience the same climate change but isn’t it possible that colder places experience a larger shift than warmer places? That could mean that removing the thermometers in colder places would introduce a cooling bias.
Is the something in the algorithms that systematically increases the anomalies if you remove thermometers in colder places? If so what is the mechanism?
Is this the answer to my question above?
Raven there is an easy way to show how when you interpolate and extrapolate the data from areas that have data to areas that don’t have data and that is to visit the GISS map maker and change the settings from Anomaly to trends and then compare 1200km infill to 250km infill
Here is a link to the 1200km infill gridded anomaly trend map with the dates 1881-2009 and SST data added in on the 51-80 baseline.
http://tiny.cc/QVMnG
Now up in the upper right hand corner you get a trend of .73 C
Now I’ll turn down the infill to 250km.
http://tiny.cc/i28FC
Now look at the trend number it dropped to .67 C. Keep in mind that there is still infilling at this level and if we totally removed that infill the trend number would change again.
Re boballab, and this is with rather adjusted data is it not?
Yep GISS does all of its adjustments prior to making the anomaly’s.
Chief,
graph of temperature data at Anantanarivo, that I already had for another reason..
@boballab: Your graphs look fine. Just the kind of thing I’m looking to do. Those were on on O.O.? Ok, OO is FINE for what I want to do. Were you on MS or Linux? I can do either. I’ve always assumed OO was mostly a Linux thing, but if it works on both, I’ll put it on both. (And it gets me past some ‘issues’ I have with walking past the “Don’t feed the animals” signs ;-)
@Raven: Substantially, yes. There “are other issues” with the thesis that “The Anomaly Will Save Us!!” and I explore a few of them in this posting:
Basically, the “TAWSU” folks keep trotting out a “Hypothetical Cow” and I keep saying “but were is the beef in this real program?”. It’s nice that a Theoretical Anomaly Process would work. But what is done is a different animal …
@vjones: Thanks!
Yep done with OO on a Win XP system. The thing has some nice features including a HTML web/writer and PDF exporter.
Oops forgot to mention the OO spreadsheet defaults to Column=data, so with an anomaly graph for 1880-2009 Column A would have the years in it with 1880 in cell A2, all the data would be Column B starting in cell B2 and B1 would be your header. You can change from Column to Row in the Data Series setting in the graph pop up menu when you do a graph. You can also set Row 1 to work like Column A.
Another useful graphic for you:
Just realised I mis-typed the adjusted trend for Anantanarivo – it should be -1.57 not -1.97. I’ve tried another version but trashed the numbers labelling the stations.
I’ll post a new link when I’ve fixed it.
Nailed it this time:
Vjones,
Was your source ghcn raw or after combining stations?
also, Madagascar makes an interesting test case.
It would be instructive to have a graphic of all 10 stations
temperature on one spagetti graph.
probably best after combining stations.
Steven Mosher,
I agree this is a good one. So is Bolivia/Peru due to relatively little data.
This data has come from a download of GISS combined/unadjusted and combined/homogenised data as explained here: http://diggingintheclay.blogspot.com/2010/01/climate-database-development.html
These were then mapped as described here: http://diggingintheclay.blogspot.com/2010/01/mapping-global-warming.html
The map shown was taken from is: http://82.42.138.62/GISSMaps/stationtrendsraw.asp (these take time to load beacuse of the amount of data – keep hitting ‘NO’ when asked if you wish to abort). The graphs for the individual stations are available from clicking on each dot.
I’ll see what I can do about a spaghetti graph.
Steve Mosher (Moshpit),
Please visit the ‘interactive map’ link that Vjones has linked to. Please also have a look at the other maps as well which can all be accessed from the following links
NOAA interactive maps”>
GISS interactive maps”>
These maps show (on separate maps) raw and adjusted temperature data individual station trends with a ‘colour coded dot’ for each individual station. There is NO anomalising and NO contour plot gridding going on just straightforward ‘single point’ data.
In particular please look at the maps that show the relative trends (warming/cooling) over the 1910 to 1939 and 1970 to 2010 periods and let me know what you think? Then look at the 1940 to 1969 maps and tell me whether or not you think the 1970 to 2010 period trends are ‘unprecedented’ within the last 100 years let alone 1000 years as Michael Mann would have us all believe? When comparing the 1910 to 1939 to the 1970 to 2010 period please make due allowance for the relative sparsity of data during the 1910 to 1939 period relative to the 1970 to 2010 period and whther or not you think that NOAA or GISS have properly allowed for the UHI effect in the latter 1970 to 2010 period.
Then tell me if you think Global Warming is actually ‘global’ or not? Tell me if you see any evidence for significant multi-decadal natural climate variability or not in these maps? And remember these are NOT anomaly ‘contour’ maps just simple straightforward (i.e. no fancy statistics just simple linear regression trends) station trend maps.
At this site you can plot the individual stations as well as the 5×5 degree HadCRU grid data:
http://www.appinsys.com/GlobalWarming/climap.aspx?area=africa
Alan I get this when clicking the link you provided:
Boballab:
Oops – can’t go directly to a world area.
Need to start at least here: http://www.appinsys.com/GlobalWarming/climapview.aspx
Then select an area of the world and click View Map / Graph
(The whole graphing application starts here: http://www.appinsys.com/GlobalWarming/climate.aspx)
Alan,
The climate data web application/database provided by AppinSys is excellent. Have you been involved in its development at all? Do you know Margaret Wilkinson?
It has a comprehensive user interface (which is much better than GISS’s site) and supplies NOAA and HadCRU (should be CRUTem I think arther than HadCRU?) data. It’s a pity that they haven’t also fitted trends to the data as otherwise its pretty much got everything you’d want. I also have the raw and adjusted dataset from GISS, so perhaps I should send them an email and suggest we collaborate?
REPLY: [ Yes, nice stuff. BTW, in the GIStemp code they talk about using HadCRUT data and one key file has a name SBBX.HadR2 so in some way or another even the ‘insiders’ think Hadley is somehow connected to CRU and the CRU-Temp series… If they are willing to do it, I see no reason why we ought not. After all, they are “Peer Reviewed” ;-) -E.M.Smith ]
Kevin:
Yes – I developed the graphing application.
(I write all the stuff at http://www.appinsys.com/GlobalWarming/)
My email address is on my main global warming page – email me.
The graphing from the map has linear trends. I haven’t added it yet to the graphing from the station list.
I have also added a link to your Digging blog (as well as a link to Chiefio)
@Alan
the appinsys site is awesome. It caused me to abandon my meagre mapping efforts, and try to find something actually useful.
A small request if you are in the mood;
The US thermometers are so densely packed that I am having trouble clicking on the ones I am seeking.
Is there an easy way to allow MoreZoom ?
Or maybe the way to do it is to offer more pre-chosen ‘sectors’ of smaller span, and then have the existing zoom thing zoom acheive higher zoom-in factor?
Thanks In Advance;
RR
@RuhRoh:
You can zoom in using the zoom-in icon in the navigator panel in the upper-left corner of the map.
You can also zoom in to a location by having the mouse pointer over the location and rolling the mouse wheel.
You can pan by using the navigator control in the upper left or by clicking and dragging on the map area with the mouse.
I have many ideas for enhancements but not enough time.
(One would be to select stations in the station list and create a map view containing only those stations.)
Alan
@KevinUK & E.M.Smith:
I have a link to the original data sources on the main graphing page. For CRU it is http://www.cru.uea.ac.uk/cru/data/temperature/
On their page, they refer to the data sets as:
CRUTEM3, CRUTEM3v, HadCRUT3, HadCRUT3v
They state: “The datasets have been developed in conjunction with Hadley Centre of the UK Met Office. These datasets will be updated at roughly monthly intervals into the future.”
I don’t know what is taking them so long to update the data – it currently has data to the end of Sep 2009. Once they update through Dec 2009 I will update the data in my graphing application.
boballab: Oops forgot to mention the OO spreadsheet defaults to Column=data, so with an anomaly graph for 1880-2009 Column A would have the years in it with 1880 in cell A2, all the data would be Column B starting in cell B2 and B1 would be your header.
OK, I’ve got O.O. installed on Windoz (even though I’d prefer it on Linux, my Linux install complained about something missing… so it’s for later) and got a spread sheet entered and even made a nice little graph.
I added trend lines, error bars, etc. All very nice.
Now how the heck do I turn it into a GIF or JPG or????
Best I could find was one at a time picking up the elements of the graph (is there a ‘glue elements together into a single object’ option somewhere? or do I get to re-fit the trend line by eye when I paste it elsewhere?…) and pasting them into a “picture” then saving it as a PDF.
Seemed really wrong…
So what “magic sauce” did I miss and where is it hidden?
(It’s just so humbling to be a ‘newby’ again at something… )
But at least I’m now able to make the graphs I’ve been wanting to make. Just need to figure out how to get them out of OO in some format WordPress will like. – THEN I’ll get to spend a few weeks going back through the 200 postings I have made and adding graphs to them 8-{
@EM Smith,
You mention your source as: input file /v/GHCN.Dec09/v2.mean
Can you confirm the source and download date? – we have a discrepancy a la what Jim Masterson mentions here:
@vjones: The name of the file says the source and download date. It is the GHCN data downloaded on Dec09 (the 28th of the month IIRC. I can check if needed).
What Jim is talking about is after GIStemp has ‘done things’. Exactly which things depends on which data he grabbed. The “raw” is in fact “USHCN.v2 +GHCN” and it not raw. The Homogenized is after STEP1 homgenizing (and I think is what he’s talking about) while the final choice adds in the STEP2 “toss short records” and UHI “correction”. ALL of those will be different from the straigh GHCN “unadjusted” that I used.
Further, that ‘move to Badwater’ might contribute to some of the homogenizing ( I’m not sure…) but ought to be tossed out in STEP2 (less than 20 years) only to suddenly be added back in whenever the thermometer ages to 20… so… if there are any OTHER records near by that moved from 19 to 20 years old between Jim’s runs, the results will change.
That is one of the more pernicious aspects of thermometer change. It ‘loads the pipeline’ with changes that may slowly dribble in over the next 20 years… 1990 => 2010 think about it…
So you drop a cold trend themometer in 1990 and add a warm one. It waits, maybe adding a bit via homogenizing it’s neighbors (1000 km away…) but otherwise “not used” per the official anomaly logs. Then, 20 years later, it bursts into the anomaly maps fully formed with 20 years of added heat.
It’s that kind of thing that makes GIStemp different from the “Anomaly Hypothetical Cow”…
Hmm… I wonder if THAT is the pipeline the “warmers” keep talking about where AGW is ‘hiding in the pipeline’ :-)
Thanks, I think we’ve sorted it. There are some 0.01C changes at monthly level but it is mostly GISS using data that TEKtemp thows out due to lots of missing months. Will follow up with email and may have detailed (potential) post for you.
EM give this a try:
Ok first on the Linux problem: The Linux OO is a different build and I don;t know if you saw the link for it but here it is:
http://download.openoffice.org/other.html#en-US
Scroll down to see all the English options for OO 3.1 including Linux 32 and 64 bit.
As to getting the charts up on WordPress you basically got to copy your graph into the HTML document writer then save the Graph as either a JPEG, PNG or another form of image file.
Step 1 click on File and then place cursor over new and watch for the next menu to open up. From that menu select HTML document. This will open a blank HTML page.
Step 2 Right click on your graph and select copy from the menu.
Step 3 Paste the graph into the HTML document.
Step 4 While the document is style not embedded in the HTML document (You should see the Anchor Icon and the anchor points) right click on the graph. This brings up another menu and one of the options is save graphics. Select Save Graphics.
Step 5 You should have a save document dialog box open in front of you. Select the Folder you want to save in, enter the name of the file you want and select the format. Make sure its in a format that WordPress recognizes such as PNG or Jpeg. Save.
Step 6 when you want to use the graphic in a post select upload image from the WordPress system and navigate to the Folder and file and then upload.
OK, Cheif,
Just tested this out on a MacG3 running OOO2.2.?
After you make the chart;
select it,
copy it
File=> Open New Presentation
do the wizard for blank presentation;
Paste the chart into the single page presentation.
Export (as GIF)
Enjoy your day.
Hope this works at your place.
RR
Dang, I check, I write, I post,
and blammo,
redundancy again.
Yikes…
RR
REPLY: [ It’s not redundancy, it’s confirmation ;-) -E.M.Smith ]
Here’s the clunky way I made animated gif;
Use PHotoshop.
Copy/paste second image into a new layer of first image.
Align as needed.
Jump to ImageMaker (funny thing at bottom of photoshop tool palette.)
Find Animation tool somewhere (I already forgot where)
Select the option to make layers into successive frames.
Select time per frame.
Export as GIF. (Save as GIF?)
Anyway, I can’t recommend this, but am happy to do it for any sequence of images that deserves a blinker.
So far no requests.
RR
REPLY: [ Hey, I’m still working on just getting static graphics. It will take me a while to work up to blinkers… -E.M.Smith ]
@EM Smith
Could I ask you a quick but important question?
When GIStemp outputs its results from STEP2, it apears to output it’s ‘adjusted data’ to six separate files that all start/end in Ts…..PA.n where n is 1 to 6. These are so called ‘peri-urban’ adjustment files that are in a binary format.
Could you confirm that it is these files that form the primary (only?) input into STEP 3 of GISTemp where it proceeds to subsequently ‘anomalise’ and ‘grid’ the adjusted station data contained within these files?
If this is the case then it appears that GISS is rejecting all the raw data that it did not see fit to adjust in STEP 1/2. Is this correct? If so on what basis can they justify doing this as looking at much of this raw data for many stations it looks fine to me? What happens to the ‘scary red all over the Arctic/Antarctic anomaly colour contour maps’ if the ‘unadjusted’ data is also passed through to STEP 3?
I’ve been thinking of doing some further ‘trend’ maps that just show the slope trends for the 1990 to 2010 period. It would be interesting to contrast the maps that show the ‘raw’ slope trend relative to the ‘adjusted’ slope trend for each station fo rthis time period as I’d expect the ‘adjusted’ slope trend map for 1990 to 2010 to be much more sparse than the ‘raw’ slope trend map.
@KevinUK
Well, checking my notes which are on line here:
were we find a quote from the GISS docs saying:
STEP2 – FORTRAN zonalizing
First up, what does gistemp.txt say:
And what does the gistemp.txt file have to say about STEP2?
Step 2 : Splitting into zonal sections and homogeneization (do_comb_step2.sh)
———————————————————-
To speed up processing, Ts.txt is converted to a binary file and split
into 6 files, each covering a latitudinal zone of a width of 30 degrees.
The goal of the homogeneization effort is to avoid any impact (warming
or cooling) of the changing environment that some stations experienced
by changing the long term trend of any non-rural station to match the
long term trend of their rural neighbors, while retaining the short term
monthly and annual variations. If no such neighbors exist, the station is
completely dropped, if the rural records are shorter, part of the
non-rural record is dropped.
Result: Ts.GHCN.CL.1-6 – before peri-urban adjustment
Ts.GHCN.CL.PA.1-6 – after peri-urban adjustment
It looks like there are 12 files, 6 unadjusted and 6 that got adjusted, that are all ‘sent forward’… But wait, there’s more:
The STEP3 comments are here:
and the code shows the controlling script only using the .PA. files:
</pre
label=’GHCN.CL.PA’ ; rad=1200
if [[ $# -gt 0 ]] ; then rad=$1 ; fi
i=”to_next_step/Ts.${label}”
if [[ ! -s $i.1 ]] ; then echo “input files ${i}.1-6 missing”; exit; fi
## Input files:
n=1
while [[ $n -le 6 ]]
do
ln ${i}.$n fort.3$n
(( n=$n + 1 ))
done
One is left to wonder what happens to all the “before peri-urban adjustment’ files…
But doing an “ls -l” on the files they look to be of very similar sizes. I would expect that the ‘adjusted’ files also contain those records where no adjustment was done. That is typically what GIStemp does. STUFF into PROCESS changed to output, unchanged passed to output too.
As soon as I get some time I’ll try chasing that down, but I suspect they are ‘sent along’ for things other than GIStemp to use, or GIStemp may use them in the web pages but not the STEP3 code? or ???
It’s a strange and wondrous product…
Of course, in this rats nest I might have missed where they are used in some other bit… but I don’t think so. After dinner I’ll spend a little quality “grep” time with the source code assuring were ever the name goes is identified…
From inside PApars.f as listed here:
first off, the comments:
So as we have a more “moth eaten” thermometer record from station “dropouts” less of the ‘record’ will be supported by 3 “rural neighbors” and more of the record will be simply tossed out. Heaven forbid that a stable thermometer be kept if the “rural” airports within 1000 km drop below 3…
Then it opens it’s files:
C**** Output text file
open(78,file=’fort.78′,form=’formatted’) ! table of adj parameters
C**** Diagnostic output files in addition to log on standard output
open(66,file=’fort.66′,form=’formatted’) ! combination info
open(77,file=’fort.77′,form=’formatted’) ! station usage stats
open(79,file=’fort.79′,form=’formatted’) ! isolated urban stations
so 78 is the default output of adjusted records while 79 is where unadjustable urban stations go to die…
Then it starts a loop:
C**** Combine time series for rural stations around each urban station
DO 200 NURB=1,NSTAu
skipping down… (79 i s log file for urban stations that are dropped):
WRITE(79,'(a3,i9.9,a13,i5,a15,i5,a50,a5)') CC(NURB),IDU(NURB), * ' good years:',N3,' total years:',N3L-N3F+1, * ' too little rural-neighbors-overlap - drop station',' 9999' GO TO 200So we see that ‘unsupported urban stations’ send you to 200
and just above 200 we have the only place where ‘adjusted’ data are written to the output file:
C**** Write out a table entry for the table of adjustment parameters write(78,'(a3,i9.9,2f9.3,i5,5f9.3,I5,a1,I4,i5,a1,i4)') CC(nurb), * idu(nurb),(fpar(i),i=1,2),nint(fpar(3)+X0),(fpar(i),i=4,6), * (rmsp(i),i=1,2),N3F+IYOFF,'-',N3L+IYOFF,N1X,'-',N2X END IF 200 CONTINUESo it does look like it just drops on the floor any “urban” station with less than 3 “rural” neighbors.
Golly.
I’d like to confirm that in some way (even if it is just showing that a station is in the input file and not in the output). But that’s pretty icky. Especially given that the “rural” data show more of a ‘hockey stick’ shape to their ‘warming’ that I think is due to all the airports classed as ‘rural’…
It would be really annoying to find out that “Global Warming” all came down to dropping more stable places called “urban” and adjusting the rest of them to match nearby airports…
To have it all come down to screwed up metadata would be an incredible travesty…
EM Funny you should ask about:
How about this little piece of information straight from NASA GISS. I got a reply when I asked GISS what happened to the data for Auckland NZ. Seems they had the ‘raw’ data but no output and it was on their list of stations used:
REPLY [ Wow. So out of about 1500 stations in GHCN, 284 are dropped on the floor in this step. And that’s AFTER all the ones dropped for being under 20 years in length… I’ll have to figure out just how many stations actually survive to be IN the anomaly map at all. (Unless someone else wants to do it first. Hint. Hint. -E.M.Smith ]
Well, thanks guys for all the help with OO.
If you look in the posting, you will now see a graph of the migration of thermometers ‘by altitude’ from 100% at elevation down to the majority below 20 M elevation.
@ruhRoh: No worries, eh! It’s nice to have more than one way to do something and it’s also nice to have confirmation it can be done!
@boballab: There are a couple of subtleties you left out (and a couple I managed to do; but still don’t know exactly what I did ;-)
It seems that OO is very “modal”. There is an “edit the chart” mode and a “not editing” mode that give you different features (that is the key to having everything stay together or not), for example. Depending on what you’ve clicked on, not clicked on, or right clicked on: you get different things to do. (On one occasion I managed to change the range of a chart, but I’ve not figured out the magic incantation again, so this one is going to stay at 120% max range ;-)
So I’m exploring and making a ‘modal crib sheet’ to remind what must be done in what order to get “that option I need”…
@ All: Dang it! Now I don’t have the excuses of:
a) No software.
b) No money for hardware or software.
c) Don’t know how to make graphs.
d) Don’t have time to figure it out.
etc.
Guess I’ll have to fall back on just “Don’t have enough time at all” ;-)
So from here on out, I’ll need to allow a bit more time to ‘polish skill with software’ and ‘make graphs prior to posting’. Oh well, it’s only time …
At least I can now say I know how to make graphs, I just didn’t do it ;-)
EM
What you did was right click on the Y Axis on the graph while in “edit” mode which brought up a menu. From that menu select “Object Properties”. This will bring up a new menu that will let you change the properties of the Y axis including the scale (You can move the minimun and maximum as well as the interval of the lines. Same process for the X axis. There is a built in help file for editing the graphs (which myself am still working through since I self taught myself how to use a spreadsheet just this last Nov).
REPLY: [ Oh, is that what I did ;-) Guess I didn’t notice that I’d had the cursor ON the axis… BTW, I have the right build for linux. It complained about some library gone walkabout. Probably a 5 minute fix, but en-queue for another day… Meanwhile I’ll practice learning my ‘stand on right toes while left clicking with right mouse on Tuesday when Scorpio is rising’ mantra on Windoz… -E.M.Smith ]
So, If I’m tracking correctly,
The bottom line is to zero in on the ‘rural neighbors’ to check them for micro-siting (and/or macro-siting issues).
At one point I dredged up a big list of Airports of the World.
I Guess one ‘quickie’ approach is to just Geo-plot all of the ‘rural’ reference stations, and then on another layer, plot the Airports of the World underneath them, and then see which ones have not joined the Aviation Age.
Another approach would be to compare the coordinates of each GISS ‘rural’ station to the big list and find the nearest airport., and then put the coordinates of that airport on the same row as the rural station.
So, then what? I suppose the next step is to see if there is a significant population of non-airport, high-quality rural stations that could be separated from the other group, and look at anomalized maps over a suitably long period.
Probably the only place with a sufficient population of long-lived records not at airports is the US.
So, probably not possible to get to a global average temperature, but at least the data quality issues could be examined for a nontrivial data set.
This has undoubtedly been explored hither and thither in Cheifio’s voluminous postings. But maybe this will give some smart guys a chance to say what is correct.
RR
REPLY: [ I’ve not done anything near that fancy. As far as I got was to make a couple of reports of percentage of airports by various things. Look in the “AGW and GIStemp issues” category on the right margin and scroll though them looking for “airport” in the title. I think there are one or two under the “GHCN / NCDC issues” category as well. Just a few clicks is all. But doing the kind of comparison you thought up? I’m nowhere near that sophisticate here. Heck, I’m using FORTRAN after all… The v2.inv file has an “A” for airstation flag and it has both a USR (Urban Suburban Rural) flag along with some population info. You could make a report based on those (i.e. I could make a report…) grouping stations in each bucket. But that metadata is pretty iffy. Your approach might be better. -E.M.Smith ]
@All,
What can you tell me about compatibility of the various versions of MS Excel and OO versions?
EM,
If you have time, can you check what rural stations are used to adjust in Madagascar? there are quite a few:
http://data.giss.nasa.gov/cgi-bin/gistemp/findstation.py?lat=-18.8&lon=47.48&datatype=gistemp&data_set=2
From initial graphs it looks as if Maintirano is the only one (on the island) that is not adjusted and it has a strong warming trend – the other stations’ trends are altered accordingly.
@Vjones: Are you asking me to look at the GISS page, or to make a report from my copy of GIStemp about what stations are used to fill in those grid cells?
I looked at the page, yeah, it’s a list (but remember it will only reach out to 1200 km. “only”… that would be Mauritius, Seychelles, Mozambique.
I’ll run temp charts on them tonight (i.e. now) and see what the look like.
I can look at the GISS page. I was hoping you would be able to tell us exactly what stations are used by GIStemp in the homogenisation of urban stations in Madagascar. I’m just not sure how big a task that is.
@EMSmith,
Thanks for the further details about what happens in the GISTemp STEP3 step. If I’ve read correctly you are confirming that the Ts…PA.n zonal files are the primary input to STEP3. As you may already know (courtesy of a little utility given to me by vjones) I’ve managed to ‘unscrample’ these six binary files into a single text file that is of exactly the same format as the NOAA GHCN v2.mean_adj file i.e. into a file that contains all ‘adjusted data’.
I then use this file within TEKTemp to derive slope trends for all stations that have ‘adjusted’ data. It therefore looks very much to me that the Ts…PA.n files DO NOT have any data for ‘unadjusted’ stations but rather only for stations where adjustments have been applied.
It therefore looks to me as though the ‘anomalising’ and ‘gridding’ on to the wonderfully scary colour contour anomaly maps only uses adjusted station data and DOES NOT use ANY station data that has been unadjusted (i.e. that is still raw). Do you concur?
@vjones
From painful past experience in using OO, the only real way to check for compatibility between MS Excel and OO spreadsheet aplication is to inter-change files between them and see what happens when you make changes in them and swap the files back and forth between them (are the chnages retained? Do the charts look exactly the same in both applications etc).
I haven’t tried the latest version of OO on Windows but more often than not I’ve ended up reverting back to using MS Office and insisting that if clients want to send me data they have created using OO that they only send me data with no embedded charts and definitely no macros etc. In other words they might as well just send me a CSV file.
@KevinUK
I try to reserve the word “adjusted” for the changes done by NCDC to “raw” data to turn it into “value added” data. For what GIStemp does, I’ve generally used UHI “corrected”, “homogenized” and “in-fill” or “Filled in”. Generically, I’ll call it “GIStemp processed”. But now we have a new need. A word to keep “GIStemp ‘value added’ data” clearly distinct from “GIStemp dropped on the floor”… Oh what a tangled web…
So what I would say is this. ALL GIStemp data has been processed by GIStemp through STEP0 (glue and drop pre 1800) and the survivors go to STEP1 that does “homogenizing” and then sent on to STEP2 that does UHI (and drops anything shorter than 20 years length). That code ALSO drops any URBAN station that it can not “UHI Correct”.
So the stations that make it out of STEP2, and go into the “grid box” STEP3 are all “adjusted” by NCDC and “glue / truncated” by STEP0, and “homogenized” by STEP1, and then a whole slew of them are dropped in STEP2, leaving only those that are:
1) Flagged as rural or
2) Flagged as urban but have been “UHI Corrected” (and perhaps segments truncated if not overlapping with long enough rural”
(Though I can’t find in the code where #1 stations are written out… see below)
Theoretically, I suppose, a station could have nothing for NCDC to adjust, be newer than 1880, not have a UHSCN record to merge, have no need for “infill” or segment splicing “homogenizing” and be marked as rural so no UHI needed and longer than 20 years so not dropped. And thus pass relatively unscathed though GIStemp. Theoretically ;-)
But all station data has been through those steps, and any station with any missing months has been dealt with…
But there is just one thing bothering me. I can’t find where PApars.f writes out any rural station data. The only write statement for ’78’ writes out the urban adjusted stations… I can’t find where ‘case 1’ records are written out.
It’s pushing 2:30 AM for me right now, so maybe I’m just getting a bit muddle headed… But it sure looks like only Urban-after-adjustment stations make it through.
I can’t believe they would do that. I must have missed something…
vjones
I can look at the GISS page. I was hoping you would be able to tell us exactly what stations are used by GIStemp in the homogenisation of urban stations in Madagascar. I’m just not sure how big a task that is.
Well, there are these log files for each step…
FWIW in looking at the logs for “STEP3” to see where it does the “Grid / Box fill in” I found an interesting thing. It looks like Madagascar stations that have been reporting no data since 2005 are being used to create boxes…
[chiefio@Hummer GHCN-Dec09]$ grep 67083000 v2.inv
12567083000 ANTANANARIVO/ -18.80 47.48 1276 1352U 452MVxxno-9A10TROP. SAVANNA B 0
so only one station pops with that ID (no duplicates in other countries).
The following has a set of ‘infixed headers’ then the Latitude and Longitude of the grid/box being filled in, the number of ‘station months’ that go into that grid/box, the number of stations total, then a list of the stationIDs (without leading 3 digit country code, but with ‘mod flag’)
Or, in ‘ragged right’ so we can see all the data, if messy:
[chiefio@Hummer work_files]$ grep 67083000 to.SBBXgrid.1880.GHCN.CL.PA.1200.log
LAT,LON,STN-MNTHS,STNS,IDS -2234 5737 4599 8 670830002 619900003 619860003 619880003 619950010 619800000 619760000 670250000
LAT,LON,STN-MNTHS,STNS,IDS -2836 4350 4165 8 673230004 670830002 671610003 673410000 670730000 619720000 684960000 684980010
LAT,LON,STN-MNTHS,STNS,IDS -2836 4650 2298 4 670830002 671610003 670730000 619720000
LAT,LON,STN-MNTHS,STNS,IDS -2836 4950 2296 4 670830002 671610003 619800000 619720000
LAT,LON,STN-MNTHS,STNS,IDS -2836 5250 2965 5 670830002 619900003 619950010 671610003 619800000
LAT,LON,STN-MNTHS,STNS,IDS -2643 4650 4215 8 673230004 670830002 671610003 670730000 619800000 670270000 619720000 619700000
LAT,LON,STN-MNTHS,STNS,IDS -2643 4950 4043 7 670830002 619900003 619950010 671610003 670730000 619800000 619720000
LAT,LON,STN-MNTHS,STNS,IDS -2643 5250 3503 6 670830002 619900003 619950010 671610003 619800000 619760000
LAT,LON,STN-MNTHS,STNS,IDS -2643 5550 4303 7 670830002 619900003 619860003 619880003 619950010 619800000 619760000
LAT,LON,STN-MNTHS,STNS,IDS -2452 5550 4303 7 670830002 619900003 619860003 619880003 619950010 619800000 619760000
[chiefio@Hummer work_files]$
Yet the data for that station end in 2005:
1256708300022000 208 199 196 196 177 147 143-9999 159 188-9999 205
1256708300022001 209-9999 220 197 175-9999-9999-9999-9999 185-9999-9999
1256708300022002-9999-9999 207-9999 173 145 147 161 168 185-9999 206
1256708300022003 204 207 208 197 189 149-9999 146 165 194 203 203
1256708300022004-9999 206 200 193 164 141 140 151 186-9999 195-9999
1256708300022005 213-9999 212-9999-9999 156 140-9999-9999-9999-9999-9999
[chiefio@Hummer GHCN-Dec09]$
I wonder where it is getting the ‘data’….
Which, I think, is your question. Where is that value made up (or ‘homogenized in’)?
And I don’t know. I’ll have to go looking. (Though as you can see, it’s not very hard to fish out what GRID/Boxes a station helps to make.).
So I’m assuming when you say which stations homogenize Madagascar you are asking for what provides the ‘fill in’ values and not for what does the UHI? That is, the STEP1 product, not the STEP2 product?
Back to the testing station…
Well, from: PApars.GHCN.CL.1000.20.log
in STEP2/work_files
we have the semi cryptic entries:
*** urb stnID: 670830002 # rur: 2 ranges: 1889 2004 Radius: 500.
LONGEST rur range: 1951-2002 43 670730000
add stn 2 range: 1957-1990 28 670190000
data added: 28 overlap: 28 years
trying full radius 1000.
*** urb stnID: 670830002 # rur: 9 ranges: 1889 2004 Radius: 1000.
LONGEST rur range: 1955-2007 44 619760000
add stn 2 range: 1951-1998 44 619720000
data added: 44 overlap: 40 years
add stn 3 range: 1948-1990 43 672150000
data added: 43 overlap: 40 years
add stn 4 range: 1951-2002 43 670730000
data added: 43 overlap: 40 years
add stn 5 range: 1951-2004 42 670050000
data added: 42 overlap: 42 years
add stn 6 range: 1960-1998 35 619680000
data added: 35 overlap: 35 years
add stn 7 range: 1957-1990 28 670190000
data added: 28 overlap: 28 years
add stn 8 range: 1974-1998 21 619700000
data added: 21 overlap: 21 years
add stn 9 range: 1966-1982 16 670040000
data added: 0 overlap: 16 years
possible range increase 18 44 48
So I think that is where it got ‘extended’. The stationIDs are without country code, though. I can likely figure them out if you need me to.
The STEP1 ‘splice and combine’ process logs report 3 records for 67083000 were used / combined to make one:
12567083000 125670830002 1889 2005 -- MCDW 125670830000 1889 1990 0.017330725051 125670830001 1961 1970 -0.0133174230972So is any of this useful for you?
It looks like those ‘9 stations’ are:
So that’s about how much work it takes to find out this stuff about one of the stations. If you let me know what part is helpful (if any) and which stations you want ‘more’ about, I can dig it out…
@EM Smith,
Enjoy your well earned sleep.
The reason why I’m interesting in knowing what makes it to the ‘anomlising’ and ‘gridding’ stage by way of stations is that I believe that part of what NOAA/GISS/CRU are doing is attempting to obfuscate (i.e. hide’ what is a great deal of natural climatic variability within the raw (unadjusted) data. By fitting trends to different time periods (1880 to 1909, 1910 to 1939, 1940 to 1969, 1970 to 2010) I think I’ve shown significant evidence of multi-decadal cyclic warming/cooling trends in the raw (unadjusted) data. This cyclic trends are on an approximate 60 year cycle i.e 30 years of warming followed by 30 years of cooling followed by 30 years of warming etc and appear to correspond with the known postive/negative phases of the AMO/PDO/SO.
I therefore want a proper explanation and justification as to why raw data for many stations appears to ‘not require any adjustments’ and is nonetheless dropped from the anomlising and gridding stages.
Over the next few days I’ll try to show using the ‘intercative maps’ what effects the ‘not including unadjusted data’ process has on the trends between 1990 to 2010. I’m also going to be looking at the ‘missing months’ problem as vjones has shown that GISS (and NOAA) appear to derive seasonal means and so year means for some years for stations where some of the monthly data for that seasonal are missing i.e. are set to -9999. For example if the value for January is missing for a given year e.g. 2001, they appear to still be able to calculate a mean for the DJF seasonal mean and then go on to use this value to derive the year mean i..e
year mean = (DJFmean + MAMmean + JJAmean + SONmean)/4
If they have been unable to derive a seosanal mean for say DJF then they do
year mean = (MAMmean + JJAmean + SONmean)/3
etc.
I think this is flawed and in TEKTemp I reject the whole years data if any one month is missing. As vjones has shown this leads to differences in the slope trends derived over a given period (e.g. 1880 to 2010) between TEKTemp and GISS for example. Visually I’ve noticed that there are LOTS of missing months in the post 1990 data, particularly the most recent data i.e. 2000 onwards, so straight off I’m going to look at the temporal ‘distribution’ for this ‘missing months’ data.
@KevinUK
“Do you know Margaret Wilkinson?”
Yes – do you?
Wow, this reminds me of that story of the shoemaker who would leave the leather out and come back in the morning to find the finished shoes.
The synergy here is a beautiful thing to behold.
Big thanks to the Euro guys for fleshing out the
‘Elvis in the Night’
part of the picture…
From my perspective, it is noteworthy that they do so much work to adjust the urban data by the ‘rural’ data, but then apparently don’t retain the benchmark (albeit glued, mohoginized, whatever) ‘rural’ data as a distinct class.
If that wee-hours observation by El Cheifio is verified,
that would seem to be a huge development in this saga.
Wow, curiouser and curiouser…
RR
Cheif,
Kevin and I have sent you that “Station life by Years” file. I hope you can open it on OO.
OK, I’m back. After 3 hours sleep (spouse decided I needed to get up earlier… “thanks” dear…) and a couple of hours of looking all all the STEP2 and STEP3 code, I still having figured out how the “rural” and “suburban” stations make it onto the anomaly grid, but I’m closer.
I think I figured out at least a way to check that they are getting there. In STEP3/work_files there is a log file:
statn.use.Ts.GHCN.CL.PA.1200
that tells you what stations are used for what GRID to fill in boxes. So, for example, in “region 6” we find:
used station 742070010 41 times
used station 742300010 42 times
used station 742300020 43 times
used station 742300030 44 times
used station 742300040 45 times
used station 743410010 22 times
42574207001 GRAPEVIEW 3SW 47.30 -122.87 15 43R -9FLxxCO 1x-9COASTAL EDGES A1 8
42574341001 NEW ULM 44.30 -94.45 262 282S 13HIxxno-9x-9COOL CROPS C2 38
Examples of both a Rural and a Suburban station being ‘used’. So they are making it into the anomaly maps. I’ve just lost track how how as they shuffle from file to file to file…
KevinUK
I’m also going to be looking at the ‘missing months’ problem as vjones has shown that GISS (and NOAA) appear to derive seasonal means and so year means for some years for stations where some of the monthly data for that seasonal are missing i.e. are set to -9999. For example if the value for January is missing for a given year e.g. 2001, they appear to still be able to calculate a mean for the DJF seasonal mean and then go on to use this value to derive the year mean
Yes, they do…
The code is toANNanom.f and a listing is here:
in it you will find that it averages, for each month of the years it has, whatever data it has. It then uses those to compute a monthly and seasonal anomaly. Eventually even an annual anomaly. What happens if you only have ONE month of the time series with a value? Well, that’s going to be your station average for that month…. and any future data for that month better be the same or it’s going to be found to be an ‘anomalous’ value…
Down in the subroutine that does that ‘work’, it looks over the twelve months, and for each month it works through all the years data. IFF a valid value is found, it is added to the “av(m)” or average of all data for month “m”. Notice there is no count for number of valid months. If there are 10 years data, but 9 are missing December. That one December value becomes the “mean”. A count of years with valid data is kept, and that lone Dec value will be divided by 1. The same is true for seasonal means where a missing month is just glossed over and for annual means where a missing season is just glossed over.
subroutine annav (mon,nyrs, iann, iy1, iy1n,iy2n, ibad) dimension mon(12,nyrs),iann(nyrs),seas(4),av(12) do m=1,12 ny=0 av(m)=0. do n=1,nyrs if(mon(m,n).ne.ibad) then ny=ny+1 av(m)=av(m)+mon(m,n) end if end do if(ny.eq.0) stop 'station too short - impossible' av(m)=av(m)/ny end do Cw write(*,*) av iy1n=ibad iy2n=ibad nok=0 do n=1,nyrs C**** find 4 seasonal means ; mon(1,.)=dec ... mon(12,.)=nov-data do is=1,4 seasis=0. nms=0 do m=1+(is-1)*3,3+(is-1)*3 if(mon(m,n).ne.ibad) then nms=nms+1 seasis=seasis+(mon(m,n)-av(m)) end if end do seas(is)=ibad if(nms.gt.1) seas(is)=seasis/nms end do C**** find annual mean anomalies from seasonal means sann=0. nss=0 do is=1,4 if(seas(is).ne.ibad) then sann=sann+seas(is) nss=nss+1 end if end do iann(iy1+n-1)=ibad if(nss.gt.2) then iy2n=iy1+n-1 iann(iy2n)=nint(10.*sann/nss) if(nok.eq.0) iy1n=iy2n nok=nok+1 end ifBut at least they require you to have at least 3 seasons to make an annual average ;-)
Wonder which season has more dropouts in the polar regions. Winter or summer? …
Hey, it’s not like a single August could make a whole annual mean. It would need to have at least an April & May, and a July, September, October to go with it ;-) It’s not like leaving out November, December, January, February and March would impart a bias to the data or anything…
(Or at least the number of days in those months that NCDC deems ‘enough’ to make the monthly average that is the input data to GIStemp; unless of course GIStemp can just make up a monthly average from somewhere ‘nearby’ in homogenizing in STEP1…)
That’s a great question; how many daily temps are needed for NCDC to declare a monthly temperature?
Is this knowable?
As I decode your comments, 1 month per season is deemed to be enough to declare a seasonal ‘average’. and 3 ‘seasons’ is enough to call it a year.
Or are you saying that they want to see more than 6 months to declare a year?
I know you’re behind on sleep. I have no excuse for fuzzyheadedness.
RR
REPLY: [ If you look closely, the code requires 2 months to make a season. I’ll bold that bit too… So the least you can get by with is 6 months to man an annual mean. 2 each in 3 seasons. Oh, and given how the wrap the year around, the list I gave above is, I think, the optimal for a biased outcome. Per NCDC and days in an acceptable month. It ought to be knowable. -E.M.Smith ]
I’ve put bold on the bits of the code that set the limits on how little data you can have.
If ANY one month (i.e. ANY December) of a series of years has a value, that will contribute to the “mean for that month” with the minimum required to get a December annual mean being a single data point.
For each Seasonal mean, you must have at least 2 months with data (said months being offset by the ‘annual mean’ to get the anomaly value). Any two will do.
For each Annual mean, you must have at least 3 seasons with seasonal means, any seasons, (that can be from 2 months, any two).
So you need at least 6 months with data to make 3 seasonal means of 2 months each; to get your ‘annual mean’.
So you can lose an entire season plus the month on each side plus one more random month and still call it an ‘annual mean’…
And folks wonder why I don’t trust the anomaly process to be our accuracy saviour…
So, the ‘mean month of multiple years’ only requires one ‘valid’ month of data in each of the 12 monthly slots,
to declare a multiyear mean month.
For a 30-year ‘baseline’ period, this routine will generate monthly mean values with less than 4% temporal coverage.
Wow, that quite a stretch if I am capturing this correctly.
It seems that the seasonal and annual mean are somewhat independent, and have the criteria you describe.
(at least 2 months per season, and at least 3 seasons to declare an annual mean=>
Therefore they require a minimum of 6 months per year, no more than 4 or 5 sequential missing months; 1/3 of possible gap-start-months allow the 5-month gaps.
Am I missing something here?
RR
REPLY:[ Looks to me like you more or less have it right. I’d only add one enhancement on the positive side. When making the seasonal means it does require that there be data in that particular seasons months in order to divide by that average month. So a 4% temporal coverage average month over years can be made, but in the step where it tries to divide it into the seasons Y/Month it will have an ‘ibad’ flag, so be dropped. You will rapidly get a lot of dropped months and so dropped seasons… I have no idea where those two curves will end up crossing (at what % temporal coverage you start getting lots of seasons that succeed but on too little data). -E.M.Smith ]
Well, what this does not answer is the question of how they treat the baseline period.
I think it is the very sparse way.
(a single good month in any of the years of the ‘baseline’ period is enough to be applied to all the years.)
I’ll recheck my work, and then post an example I hope.
I tested this by looking at the beginning of recordings in Anarctica, in about Dec 1957.
I get the same purple antarctica for a 1956-1957 base, as a 1950-1957 base.
RR
Ok, Check these two plots;
So, it was Dec1958 compared to 1957-1957 vs
Dec1958 compared to 1900-1957
using ‘unadjusted’ data and no oceans.
Too bad only 1200km is the only viable option.
RR
http://data.giss.nasa.gov/cgi-bin/gistemp/do_nmap.py?year_last=2009&month_last=12&sat=0&sst=0&type=anoms&mean_gen=12&year1=1958&year2=1958&base1=1900&base2=1957&radius=1200&pol=reg
http://data.giss.nasa.gov/cgi-bin/gistemp/do_nmap.py?year_last=2009&month_last=12&sat=0&sst=0&type=anoms&mean_gen=12&year1=1958&year2=1958&base1=1957&base2=1957&radius=1200&pol=reg
Where did we get to with deciding HOW the anomaly was calculated over on the Air Vent?
I have the spaghetti graphs done and other graphs; if you do a straight average of the anomalized data then the original temperature of the data is not important (but the adjustments make a huge difference – as does the quality of the data).
You quoted on tAV “As a final step, after all station records within 1200 km of a given grid point have been averaged, we subtract the 1951-1980 mean temperature for the grid point to obtain the estimated temperature anomaly time series of that grid point.”
You know I think I’ll leave that side out for the present as my head hurts and I can see wiggly lines when I close my eyes.
The Third Shoe
Adding 1900-1957 to the baseline didn’t change anything ,but
making the baseline 1900-1956 makes some purple go away.
http://data.giss.nasa.gov/cgi-bin/gistemp/do_nmap.py?year_last=2009&month_last=12&sat=0&sst=0&type=anoms&mean_gen=12&year1=1958&year2=1958&base1=1900&base2=1956&radius=1200&pol=reg
REPLY: [ Interesting graphic. LOTS of purple. in the part of the arctic that is all red all the time today. Gee, very cold year in the default baseline, very hot anomaly… who knew ;-) -E.M.Smith ]
OK2seasons of >1months criterion, for this particular testcase. That will depend on when the station first came on line. More digging needed here.
This particular sentence was going to start with the word ‘Presumably’, but then I remembered we are talking about Gistemp here, so it would be an oxymoronic assertion if I had
not reconsidered it.
I only have evidence for a single year’s monthly data being declared sufficient for Gistemp to report a multi-year baseline ‘average’ when that data is in the final year of an epoch. The bolded code segment in the above post shows no hint that the final year of a baseline epochd is afforded unique treatment.
What are the implications of this finding?
I think that those clever Euro DBwizards had implemented a more conservative, less ‘inclusive’ approach to the treatment of partial years and holes in the record.
I don’t know how hard it would be for them to build a comparison version which implemented the ‘generously inclusive’ gistemp averaging rules,
and then
make a ‘differencing’ analyzer, to automatically quantify some measure of the net change of the ‘zonal means’ (?),
and then even more grandiosely,
run a ‘baseline sensitivity sweep’
which plots the net difference parameter as a function of the swept baseline parameter.
Sweeps could be ‘rolling N year baselines’ where the start year is rolled through time,
or
N is swept ,
or
Your Ad Here…
Talk is cheap, some less valuable than other…
Anyway, I still need to pin down the question of whether the extant on-line version of gistemp requires a ‘valid’ single year average to be available before including a year in a multiyear baseline calculation. I think I have shown that data within a single year is sufficient to cause gistemp to report that single year data as the multiyear average baseline. The calculated multiyear baseline average is not changed by the inclusion of >50 years without data.
Please flag any overbroad claims or ambiguity.
RR
OK
Whoops, the prior post did not include the intro section.
So, to at least my own satisfaction, I have herewith shown that the Gistemp baseline average calculator (at present) does a very ‘inclusive’ style of multi-annual averaging;
the presence of a single month of data,
(possibly in ‘sufficiently nonsparse year’)
for a particular station,
in a 57 year baseline,
is sufficient for Gistemp to declare an average temperature for that station/month, over the baseline period.
I concluded this after
1. finding the onset of data for a particular part of the Antarctic (Dec 1957), shown by change from gray before to color thereafter.
2. Comparing Dec 1958 to a baseline of Dec 1957 to see the purple color report.
3. Changing the baseline from December of a single year 1957 to December of a multiyear epoch (1900-1957),
4. Observing the same station color report as step 2.
5. Changing the baseline Epoch to December of 1900-1956
6. Observing station color report revert to gray (no data) because the Dec 1957 first and only monthly data for that station was no longer in the baseline epoch.
So, the code you provided does seem to require at least 3
‘seasons’ of containing at least 2 ‘good months’ of data when calculating single-year ‘annual’ means.
But these restrictions might not apply to the multiyear baseline ‘averaging’ process.
Hmmm, I have NOT yet demonstrated whether 1957 was required to pass the { G.T. 2seasons of G.T.1months } criterion, for this particular testcase. That will depend on when the station first came on line. More digging needed here.
This particular sentence was going to start with the word ‘Presumably’, but then I remembered we are talking about Gistemp here, so it would be an oxymoronic assertion if I had
not reconsidered it.
I only have evidence for a single year’s monthly data being declared sufficient for Gistemp to report a multi-year baseline ‘average’ when that data is in the final year of an epoch. The bolded code segment in the above post shows no hint that the final year of a baseline epochd is afforded unique treatment.
What are the implications of this finding?
I think that those clever Euro DBwizards had implemented a more conservative, less ‘inclusive’ approach to the treatment of partial years and holes in the record.
I don’t know how hard it would be for them to build a comparison version which implemented the ‘generously inclusive’ gistemp averaging rules,
and then
make a ‘differencing’ analyzer, to automatically quantify some measure of the net change of the ‘zonal means’ (?),
and then even more grandiosely,
run a ‘baseline sensitivity sweep’
which plots the net difference parameter as a function of the swept baseline parameter.
Sweeps could be ‘rolling N year baselines’ where the start year is rolled through time,
or
N is swept ,
or
Your Ad Here…
Talk is cheap, some less valuable than other…
Anyway, I still need to pin down the question of whether the extant on-line version of gistemp requires a ‘valid’ single year average to be available before including a year in a multiyear baseline calculation. I think I have shown that data within a single year is sufficient to cause gistemp to report that single year data as the multiyear average baseline. The calculated multiyear baseline average is not changed by the inclusion of more than 50 data-free years.
Please flag any overbroad claims or ambiguity.
RR
More Madagascar Madness
Well Madagascar has a bit a of further story to tell. I had offered to plot a spaghetti graph of the temperatures from the 10 Stations used on Madagascar. For a start, the annual mean temperatures plotted on a graph show clearly the differences between the stations – Antananarivo is high altitude and relatively cool; some have cooling trends, most are warming. Very sparse data after 1990. Note the darker blue data for Maintirano, of which more later.
This is what it looks like if you do just average the mean annual temperatures, which, clearly does not work as an average temperature for the island.
Normalizing each of the temperature series by calculating the mean temperature for that station for the baseline period of 1951-1980 allows plotting of an anomaly-based spaghetti graph. This shows what looks like warming-cooling-warming climate cycles very clearly and it is possible to fit a third order polynomial trendline though the averaged data. I’ve seen this again and again for data I’ve plotted (for TonyB incidentally) around the world.
Now for the interesting bit – how GIStemp adjusts the data. GIStemp takes rural datasets and uses them to correct for urbam warming. In this set of 10 unadjusted stations there were 3: Maintirano and two overlapping ones but separate ones for Antalalava (why kept separate?). In the homogenized set, only Maintirano, which had a large warming trend of 1.16 deg. C/century remains undjusted and all the other stations have the trend increased – it seems to match this station. E.M.Smith above finds 7 other rural stations within 1000km that may contribute to homogenization. They also show cooling to about 1965-1975, then a warming trend. This is lost from the homogenized data.
So overall what effect does homogenization have – well a big one. Having started into a better understanding of calculation of anomalies, I decided it was better to leave that for the present, but a straight average of the normalized unadjusted and homogenized overlaid with a 10 year moving average for each shows just what homogenisation does for the ‘anomaly’ value for Madagascar calculated this way – it stabilises the base period and significantly warms the subsequent years.
Given that several of the stations show a cooling trend prior to homogenization, and that UHI correction should NEVER be in the wrong direction, this is nothing short of scandalous.
I originally looked at the temperature strends using the TEKtemp database, but when I grabbed up-to-date data from the GISS site, I found the trends in the data were different. We’ve now found the reason for that and that is worth investigating in its own right (Kevin is looking into it). The answer is simple – bad data. The TEKtemp QC system throws out years with missing months of data and after 1990 the data in most of the Madagascar stations is patchy at best, so TEKtemp ingnored the data in plotting the temperature trends. It is amazing how much warmer Madagascar is with that patchy data included.
http://2.bp.blogspot.com/_vYBt7hixAMU/S2ynNoL1S4I/AAAAAAAAARc/ZC_kGLn_2aQ/s1600-h/Madagascar+4.bmp
Once final thing. Even the patchy data stops in 2005 so after this date Madagascar too gets ‘filled in’ data from elsewhere – it seems from the rural stations up to 1000km away – again. And even they have patchy data – many have a gap, then ONE DATA POINT in 2009. This is unbelieveable. I was going to post an example, but better still here are the hyperlinks – check for yourself:
http://data.giss.nasa.gov/cgi-bin/gistemp/findstation.py?lat=-18.8&lon=47.48&datatype=gistemp&data_set=1
16861976000 SERGE-FROLOW
11161972000 ILE EUROPA
13167215000 PORTO AMELIA
12567073000 MAINTIRANO
16367005000 DZAOUDZI/PAMA
11161968000 ILES GLORIEUS
12567019000 ANALALAVA
11161970000 ILE JUAN DE N
11167004000 OUANI (ANJOUA
This table might or might not work:
Ile Juan De N 17.1 S 42.7 E 111619700000 rural area 1973 – 2009
Dzaoudzi/Pama 12.8 S 45.3 E 163670050000 rural area 1951 – 2009
Iles Glorieus 11.6 S 47.3 E 111619680000 rural area 1956 – 2009
Ouani (Anjoua 12.1 S 44.4 E 111670040000 rural area 1963 – 1984
Serge-Frolow 15.9 S 54.5 E 168619760000 rural area 1954 – 2009
Ile Europa 22.3 S 40.3 E 111619720000 rural area 1951 – 2009
Porto Amelia 13.0 S 40.5 E 131672150004 rural area 1987 – 200
All – my previous comment, to which the table relates, is in the moderation queue due to the number of hyperlinks.
REPLY: [ not any more ;-) And I’ve upped the ‘link limit’ by 2 more so you can escape it in the future. It’s now 8 but let me know if you want more … -E.M.Smith (great stuff, BTW) ]
@vjones:
It was just that kind of WT… from looking at places where there was no data and finding them made up from nothing (and finding flat or roller trends turned into warming) that caused me to shift from skeptical to flat out not buying it at all…
We only have about 1000 stations from 2009 making it into the final steps of GIStemp, and now we’re finding out that a lot of them have ‘patchy data’ that are just filled in and made up too.
With every step, we end up with more fantasy and less reality.
From this folks expect 1/10 C precision AND accuracy?
Just nuts.
Those by altitude/latitude charts look much better and more informative as stacked area charts.
Pingback: Munging Madagascar « Watts Up With That?