Originally Posted by Bob Butler 54

I vaguely recall a number of stations vanishing from the record about the time the cold war ended. I don't know that the missing data will be totally random, but is apt to have military, political and economic causes. I don't know that this would really introduce a bias, though I'm particularly interested in the early 1990s, which is just when a bunch of stations vanished.

The site's down again and I cannot get the rest of the stations. I found a way to get around the changing stations. What I did was calculate the average July temperature stations at the same approximate lattitude over the entire period:

Lat Temp
54 18.7
57 18.4
60 16.3
63 15.8
66 13.9
69 10.7
72 8.4
75 3.2
78 2.3
81 0.7

I then subtracted the appropriate average from the raw data to get the temperature data relative to the long-term average for that lattitude. This removes the effect of location. Now missing stations won't affect the data. I then averaged these data for year year as before:

Year Temp
1960 0.64
1961 0.35
1962 -0.03
1963 -0.34
1964 0.35
1965 -0.63
1966 -0.33
1967 0.29
1968 -1.88
1969 -0.08
1970 0.16
1971 -0.27
1972 0.46
1973 -0.59
1974 0.85
1975 -0.27
1976 -1.16
1977 -0.23
1978 -1.43
1979 -0.68
1980 -1.25
1981 0.15
1982 -0.06
1983 0.06
1984 0.13
1985 -1.32
1986 -1.02
1987 -0.59
1988 1.49
1989 0.31
1990 0.27
1991 0.72
1992 -0.81
1993 0.37
1994 -0.49
1995 0.04
1996 0.06
1997 -1.08
1998 1.26
1999 0.89
2000 0.73
2001 1.58
2002 1.40
2003 1.17
2004 0.95
2005 0.83
2006 -0.25

Ran another regression, positive slope of 2.1 ± 0.8 deg/century, significance

The plot looks pretty much like the last one, a cloud of points with a positive trend only. Initially I calculated the average temperature for each station rather than lattitude. I got an even more disperse collection of points. The regression was still significant and still positive, but the data was less tight that the original approach and worse than the lattitude approach as well.

The data vary about the trend line with a standard deviation of 0.76. This is about the same size as the trend rise over the 46 year period covered.

Assuming all months trend this way, if one used all the stations and all twelve months, this should reduced the noise to about 0.15-0.2, in which case the trend should be quite visible in a plot. Assuming I can get back in I am going to try to do this to see,

Originally Posted by Brian Rush

Myself, I'm of course unsurprised that you found a warming trend from the data Justin linked, because that was the only likely outcome, but I'm also unsurprised that Justin is finding excuses to dismiss it -- and that's why I was unwilling to go to all that work myself. But then, you've done a better job, certainly with better graphics, than I would have, so all's well.

I have been watching your joust with JPT, notably post 433 in the myth thread, and will return your applause.

My recent decision to play here in Global Warming is in part as there is hard data and science. Yes, the skeptics can try to dismiss as much data as possible. Yes, they will draw 'trend lines' from a peak point of a forcing factor to a trough and believe such trends will lead them to the future. Still, the climate science is there. Even when Justin cherry picks his source of data, a specific part of the world and a specific time frame, the science is there.

He also hasn't dismissed it yet. He seems to be acknowledging the existence of the Arctic Oscillation, though he hasn't really rejected his Siberian peak to trough trend lines. We'll see if he speaks to my recent posts.

How can one force a libertarian to acknowledge main line economics? How does one convince a living in the past conservative not to confuse populism with communism? Navigating a data blizzard isn't easy, especially when access to the data blizzard is so intermittent, but it seemed preferable to trying to convince JPT that his perspective on American political history isn't the only possible perspective.

The New York Times (Thomas Friedman) writes The Copenhagen That Matters a Global Warming piece about how Denmark is building its green economy and thriving as a result. It's the usual stuff. Pass some carbon taxes and industry will adjust in a constructive way that produces jobs and stimulates the economy.

But one paragraph in particular speaks of turning theory. We are not in regeneracy yet. Period. Underline. Underline it again. While I still think there was a security related aspect to the crisis during the Bush 43 administration...

Espersen laughed. But I didn’t. How long are we Americans going to go on thinking that we can thrive in the 21st century when doing the optimal things — whether for energy, health care, education or the deficit — are “off the table.” They’ve been banished by an ad hoc coalition of lobbyists loaded with money, loud-mouth talk-show hosts who will flame anyone who crosses them, political consultants who warn that asking Americans to do anything important but hard makes one unelectable and a citizenry that doesn’t even ask for optimal anymore because it believes that optimal is impossible.

Originally Posted by Mr. Friedman

Espersen laughed. But I didn’t. How long are we Americans going to go on thinking that we can thrive in the 21st century when doing the optimal things — whether for energy, health care, education or the deficit — are “off the table.” They’ve been banished by an ad hoc coalition of lobbyists loaded with money, loud-mouth talk-show hosts who will flame anyone who crosses them, political consultants who warn that asking Americans to do anything important but hard makes one unelectable and a citizenry that doesn’t even ask for optimal anymore because it believes that optimal is impossible.

We'll be thinking that way until Millennials become a larger political coalition. Millennials aren't like that.

Originally Posted by writerGrrl

We'll be thinking that way until Millennials become a larger political coalition. Millennials aren't like that.

I hope so, and I hope it's soon.

First, I'm about ready to resume the Cold War against the Russians. I am beginning to suspect that their climate data web site is designed as a subtle tool to frustrate capitalists. It was up again for a while. A while.

I got Justin's list of 133 stations into a spreadsheet, sorted by latitude, then started selecting sites for download starting from the north, and working my way south. I only got two stations downloaded before aisori.meteo.ru/climat went kaputchnik. Those two were...

20087, Golomjannyj, 79.55 N, 76.95 E
20476, Sterlegova, 75.42 N, 88.90 E

Just to make sure everyone is seeing the obvious, 79.55 and 75.42 are way far north. I mean, yes, it is really cold up there. For some reason they shut down both stations around 2000 A.D. Can't understand it. Anyway...

How far north are Golomjannyj and Sterlegova? Well, I had to subtract 17 from Mike's average of all Siberian stations to position it on the chart. I had to add 3 to Perevoz. I had to add 11 to Sterlegova, and 12 to Golomjannyj.

I'm including Perevoz again as it is representative of the 'gentle' decline since 1990 pattern suggested by Justin.

Anyway, I was sort of assuming that the temperature curves for all 133 of Justin's sites would be sort of similar. I figured there was a warm stretch in the early 90s. There was one in Perevoz, and Ika, and I assumed elsewhere, though judging from Mike's curve, it is less prominent further south..

Except the northern stations dive cold just as Perevoz goes up in that warm bulge. Note, the northern stations in the early 1990s time frame are moving in opposition to the AO and solar forcing curve.

I am now guessing that part of the reason Perevoz and Ika stayed kind of warm was because a lot of cold air stayed up around Golomjannyj and Sterlegova.

I note that in the 1960s and early 1970s the three station curves are moving more or less together. In the mid 1970s, It looks like the curves start doing the Thatch Weave. (WW II fighter reference.) As the green Perevoz curve loops in one direction, the other two stations go the other way. This weave pattern peaked huge in the early 1990s. They may be settling back together at this point. Maybe.

Not sure if this means anything to the discussion, other than I'm even less inclined to think the warm time around 1990 is a typical data point from which one starts trend lines.

I've downloaded all the monthly temperature data from all the stations for the period 1900-2008, inclusive.

I calculated the average temperature for the 1960-2008 period for each month and for each station. I used this subset because the data set gets increasingly spotty as you go further and further back. This 47 year period contained 23008 rows of data, 270,499 values. A full set of data would have 23,008 X 12 = 276,096 values, so it is missing 0.55% of the data.

I wanted a true average temperature and cannot include a bunch of zeros for months missing data. So I scrubbed this data set to remove all of the blank cells. I simply sorted the data for January, which puts all the rows in which January is blank on top, and then deleted all these rows. I repeated this for the other 11 months. Doing this left me with 21732 rows of data for which all 12 months have values, a total of 260,784 values. This cost the loss of 3.6% of the data.

I now had an array of station averages for each month. The Excel vlookup function is designed to fetch the appropriate average from this array when given the station number and month.

I partially scrubbed the pre-1960 data to remove all rows in which three or more cells were blank. I then appended it onto the 1960-2008 data and sorted by year to obtain a single data set with 37622 rows. I calculated relative temperatures for all the data in the table by subtracting the appropriate station average (obtained automatically by the lookup function) from each of the values in this table.

I now had the raw data put into the form of station values relative to their own average over the 1960-2008 period. I then calculated averages for each year and month (that is from all the stations that had data for that year and month). I now had an array with 109 rows (one for each year from 1900 to 2008) and 12 columns (one for each month), all expressed in terms of deviation for their own long-term average.

I then averaged the 12 monthly temperature to get an annual average (the first col is year, the second is the annual average temp, and the 3rd is the # of stations that provided data for that year):

1900 -1.07 65
1901 -0.29 68
1902 -2.03 72
1903 -0.11 76
1904 -0.70 72
1905 -0.20 67
1906 0.06 68
1907 -1.30 74
1908 -1.50 72
1909 -0.55 73
1910 -0.41 74
1911 -0.90 79
1912 -1.57 82
1913 -0.54 83
1914 -0.40 90
1915 -1.06 91
1916 -0.82 92
1917 -0.74 91
1918 -0.83 78
1919 -1.14 63
1920 0.06 68
1921 0.02 71
1922 -0.43 74
1923 -0.28 82
1924 -0.26 81
1925 0.18 91
1926 -0.78 96
1927 -0.76 114
1928 -1.00 119
1929 -1.43 119
1930 -0.42 120
1931 -0.78 124
1932 0.21 125
1933 -1.35 139
1934 0.26 156
1935 -0.18 168
1936 0.08 187
1937 -0.13 185
1938 0.39 186
1939 0.08 194
1940 -0.72 311
1941 -1.64 300
1942 -0.89 295
1943 0.53 302
1944 0.25 309
1945 -0.90 321
1946 -0.70 336
1947 -0.67 340
1948 0.36 350
1949 -0.13 349
1950 -0.54 361
1951 -0.18 357
1952 -1.07 360
1953 -0.17 362
1954 -0.56 358
1955 -0.61 370
1956 -1.13 368
1957 -0.28 372
1958 -0.97 375
1959 -0.14 419
1960 -0.89 422
1961 0.08 454
1962 0.45 461
1963 -0.18 458
1964 -0.88 463
1965 -0.90 468
1966 -1.13 461
1967 0.24 460
1968 -0.70 449
1969 -1.95 457
1970 -0.60 462
1971 -0.25 457
1972 -0.57 462
1973 -0.09 455
1974 -0.46 438
1975 0.67 457
1976 -1.08 448
1977 -0.63 455
1978 -0.53 457
1979 -0.68 441
1980 -0.57 448
1981 0.57 422
1982 -0.14 406
1983 0.75 443
1984 -0.44 455
1985 -0.83 436
1986 -0.11 439
1987 -1.33 463
1988 0.32 456
1989 0.98 428
1990 0.94 435
1991 0.62 447
1992 0.07 448
1993 0.02 443
1994 -0.01 433
1995 1.57 433
1996 -0.02 430
1997 0.51 416
1998 -0.20 433
1999 0.22 429
2000 0.44 435
2001 0.40 417
2002 0.86 409
2003 0.83 442
2004 0.65 436
2005 1.13 439
2006 0.20 447
2007 1.72 443
2008 1.48 436

This data when plotted has an obvious postive slope. Linear regression so a rising trend of 1.2 deg per century that is significant and 99.9999994% significant.

A linear fit isn't the best fit, the r squared is 027. A second order polynomial does a bit better, giving r squared = 0.33. This plot shows a flattish trend for the early years and an accelerating rise in the latter years. A third degree polynomial boosts r squared to 0.39. It shows a rising trend early in the 20th century, ending in the 1920's, then a flattish period from the 1920's to the 1970's and a rising trend afterward.

Going to a 4th degree polynomial gives a modest improvement to r aquared of 0.40. It shows the same basic shape, rising trend to around 1930, flat to the 1970's and then rising afterward.

Going to a 5th degree polynomial has an even smaller increase on r squared, it shows the same rising trend to the mid 1930's and a gentle decrease to about 1970, and then a rising trend afterward. Going to sixth degree polynomial adds nothing to r-squared; nothing more is explained by going to higher order than 5.

These Russian data show the same broad trends shown by the global data. There is a rising trend in the early portion of the 20th century, a flat-to-slightly down period in the middle and a rising trend starting around 1970. On difference is the Russian data shows warming continuing throughout this decade; that is, no evidence of any leveling off over the past 10 years. For example, the warmest year was 2007, third warmest was 2008, fourth warmest was 2005; seventh warmest was 2002 and eighth warmest was 2003.

There were a few surprises in this data set.
The basic data set contains 391,496 values. The highest temp was 30.0 C and the lowest -55.1 C. The average was -1.3 and the standard deviation was 14.7 degrees. Based on this max temperature data it appears that these data are not from individual days in a month, but are some sort of monthly average. I would expect the hottest day on record would have an average temperature higher than 30 C (86 F). Similarly I would expect the coldest day to have a temperature less than -55 C (-67 F).

After subtracting the station averages, I get the following properties:

average -0.17 C, standard deviation 2.7, max +20.5, min -17.7

The standard deviation has shrunk about 5.5-fold, meaning that I have removed a great deal of the variation, that variation caused by the location of the stations and the time of the year. Stations farther north will tend to have colder temperatures than stations farther south. Similarly temperatures in the summer will tend to be warmer than in the winter. Bu subtracting out the station average for a given month I largely remove these effects and leave the variation due to other things.

The biggest remaining factor is weather. Weather is fundamentally unpredictable more than a few weeks out, because of chaos. The assumption is the effects of chaos over sufficiently long period will be white noise, random fluctuations around the mean value. By averaging data from various stations for a given month and year the effect of weather is averaged out, leaving the effects of climate.

The standard deviation of the individual relative temperatures was 2.7 degrees. Each of the temperature values in the post above is an average or at least 750 values. Based on the standard deviation of the raw data, the standard error of the mean of averages of these data is about 0.1 C at most. That is, random factors (i.e. weather) can contribute a dispersion to the final data described by a standard deviation of about 0.1 C.

The actual dispersion is about six times larger. This means there are some nonrandom factors besides global warming that makes some years significantly different from other nearby years (e.g. 1960 and 1961). The rest might be described by other climatic factors such as ENSO and this artic oscillation. There could also be non random artifacts in the data that could cause unusually large dispersion. Whatever they are, they make the averaged temperarate profile I obtained more noisy that the sorts of plots I have seen elsewhere.

The presence of this greater noisiness does not change the global warming result. This long-term trend is real, highly significant in the Russian data, and consistent with what has been reported elsewhere.

I took the data I posted two up from here. I fitted it to a 3rd degree polynomial (the smallest degree that captured almost all of the variation that can be captured by a polynomial fit). This polynomial represents the effect of global warming. I then subtracted this out to obtain the residuals (listed below). I plotted them out. They have no trend (slope zero). They shouldn't because by subtracting out the global warming trend I have removed the trend. What is left is variation from year to year that has a standard deviation of 0.56, compared to 0.73 for the data before detrending. The regression coefficient for the polynomial fit I used to represent the effects of global warming shows that this trend accounted for about 40% of the total variation. Most of the variation of this data (60%) is still in these trendless residuals given below. I applied various moving averages to the data to look for cycles. I did 5, 6, 7 and 9 year centered moving averages. What I saw were what looks very much like Yule-Slutsky cycles.

That is, these residuals appear to be random. But they are not the effect of the short-term chaotic (random) processes (i.e. weather) that give rise to daily weather variations because these processes would produce a standard deviation of these residuals of less than 0.1 C, compared to the 0.56 that is observed. There are two possibilities. One is there are longer-term chaotic processes that produce seemingly-random perturbations in the temperature record on the scale of years, or there is some systemic artifact that I am carrying forward in my analysis that I lack the statistical chops to identify.

The official data shows the same trend I find in the Russian data. The dispersion is tighter, however. I took the HAD crut3 data from 1900-2006 and fitted it to a cubic polynomial. I then subtracted the polynomial fit to obtain the residuals. They had a standard deviation of 0.11, fives times less than what I get from the Russian data.

So what I have learned so far is the Russian data show the same global warming trend as do the HADcrut3 data. This trend can be well represented by a cubic polynomial for both data sets. The fit is very significant for both data sets. The general shape (rising in the early 20th century, flat in the middle of the century, and rising afterward) is the same for both. The residuals are much smaller for the HADcrut data than for mine (standard deviation of 0.11 versus 0.56). As a result the polynomial has a higher regression coefficient for the HADcrut4 data than for my data. The fit explains more than 80% of their variation versus about half that for mine.

Their standard deviation of the residual is so small that the effects of short-term chaotic fluctuations (i.e. weather) can explain the residual. For my data I have to invoke some new phenomenon to explain my greater dispersion. This is why I suspect I have a statistical problem.

On the other hand, the HADcrut3 data is for the entire planet. There may be real processes that produce perturbations in some parts of the planet that are balanced by perturbations in other parts of the planet. Such perturbations would average out in a database that covers the whole planet, whereas they would not in a database that covers about 4% of the planet, mostly in upper northern hemisphere.

Year Residual

1900 -0.06 1901 0.67 1902 -1.10 1903 0.78 1904 0.15 1905 0.62
1906 0.85 1907 -0.54 1908 -0.77 1909 0.15 1910 0.27 1911 -0.24
1912 -0.94 1913 0.07 1914 0.19 1915 -0.48 1916 -0.26 1917 -0.19
1918 -0.29 1919 -0.62 1920 0.57 1921 0.52 1922 0.06 1923 0.20
1924 0.22 1925 0.65 1926 -0.32 1927 -0.30 1928 -0.55 1929 -0.98
1930 0.03 1931 -0.33 1932 0.65 1933 -0.90 1934 0.70 1935 0.27
1936 0.53 1937 0.32 1938 0.85 1939 0.54 1940 -0.26 1941 -1.18
1942 -0.43 1943 0.99 1944 0.72 1945 -0.43 1946 -0.23 1947 -0.20
1948 0.85 1949 0.35 1950 -0.05 1951 0.31 1952 -0.58 1953 0.32
1954 -0.06 1955 -0.12 1956 -0.64 1957 0.22 1958 -0.47 1959 0.36
1960 -0.39 1961 0.57 1962 0.94 1963 0.30 1964 -0.40 1965 -0.42
1966 -0.66 1967 0.70 1968 -0.25 1969 -1.51 1970 -0.17 1971 0.18
1972 -0.16 1973 0.30 1974 -0.08 1975 1.04 1976 -0.73 1977 -0.30
1978 -0.22 1979 -0.39 1980 -0.31 1981 0.81 1982 0.07 1983 0.93
1984 -0.29 1985 -0.71 1986 -0.02 1987 -1.27 1988 0.34 1989 0.95
1990 0.88 1991 0.51 1992 -0.08 1993 -0.18 1994 -0.26 1995 1.27
1996 -0.37 1997 0.10 1998 -0.67 1999 -0.31 2000 -0.16 2001 -0.26
2002 0.14 2003 0.03 2004 -0.22 2005 0.18 2006 -0.82 2007 0.61
2008 0.29

Just plotting the numbers Mikebert just sent.

I note that as Mikebert's Russian data has more than Justin's list of 133 stations, he is likely still gathering data for all of Russia, not just for Siberia. He is not really testing Justin's hypothesis.

While the red all of Russia chart shows the familiar global warming upswing in temperature, the HADcrut3 curve looks flatter than I'm used to seeing it.

I added 3 degrees to the HADcrut3 data, just to keep the two curves from over writing each other. I suspect I ought to do rolling average versions at some point, the better to see the trends...

James Hanson, one of the keepers of the climate data from NASA, posted another response to the 'skeptics'. (pdf) It sounds familiar. His summary ends...

Originally Posted by James Hansen

Summary

The nature of messages that I receive from the public, and the fact that NASA Headquarters received more than 2500 inquiries in the past week about our possible “manipulation” of global temperature data, suggest that the concerns are more political than scientific. Perhaps the messages are intended as intimidation, expected to have a chilling effect on researchers in climate change.

The recent “success” of climate contrarians in using the pirated East Anglia e-mails to cast doubt on the reality of global warming* seems to have energized other deniers. I am now inundated with broad FOIA (Freedom of Information Act) requests for my correspondence, with substantial impact on my time and on others in my office. I believe these to be fishing expeditions, aimed at finding some statement(s), likely to be taken out of context, which they would attempt to use to discredit climate science.

There are lessons from our experience about care that must be taken with data before it is made publicly available. But there is too much interesting science to be done to allow intimidation tactics to reduce our scientific drive and output. We can take a lesson from my 5-year-old grandson who boldly says “I don’t quit, because I have never-give-up fighting spirit!” http://www.columbia.edu/~jeh1/mailin...tingSpirit.pdf

There are other researchers who work more extensively on global temperature analyses than we do – our main work concerns global satellite observations and global modeling – but there are differences in perspectives, which, I suggest, make it useful to have more than one analysis. Besides, it is useful to combine experience working with observed temperature together with our work on satellite data and climate models. This combination of interests is likely to help provide some insights into what is happening with global climate and information on the data that are needed to understand what is happening. So we will be keeping at it.


*By “success” I refer to their successful character assassination and swift-boating. My interpretation of the e-mails is that some scientists probably became exasperated and frustrated by contrarians – which may have contributed to some questionable judgment. The way science works, we must make readily available the input data that we use, so that others can verify our analyses. Also, in my opinion, it is a mistake to be too concerned about contrarian publications – some bad papers will slip through the peer-review process, but overall assessments by the National Academies, the IPCC, and scientific organizations sort the wheat from the chaff.

The important point is that nothing was found in the East Anglia e-mails altering the reality and magnitude of global warming in the instrumental record. The input data for global temperature analyses are widely available, on our web site and elsewhere. If those input data could be made to yield a significantly different global temperature change, contrarians would certainly have done that – but they have not.

The bottom line is that there is only one way this can be made to "work" - and that is the establishment of a one-world government - since if everybody doesn't go along with the proposed restrictions on industrial development, the effect on the planet is the same as if nobody does.

I'm surprised that the usual suspects on the far right haven't latched onto this idea yet.

Originally Posted by Bob Butler 54

Just plotting the numbers Mikebert just sent.

I note that as Mikebert's Russian data has more than Justin's list of 133 stations, he is likely still gathering data for all of Russia, not just for Siberia. He is not really testing Justin's hypothesis.

While the red all of Russia chart shows the familiar global warming upswing in temperature, the HADcrut3 curve looks flatter than I'm used to seeing it.

I added 3 degrees to the HADcrut3 data, just to keep the two curves from over writing each other. I suspect I ought to do rolling average versions at some point, the better to see the trends...

The HADcrut data doesn't look right to me. Where did you get it?

Originally Posted by Bob Butler 54

I note that as Mikebert's Russian data has more than Justin's list of 133 stations, he is likely still gathering data for all of Russia, not just for Siberia. He is not really testing Justin's hypothesis.

Bob is right. The data I presented so far is for all the stations. It is simple matter to do the same for any subset of those stations. Here is the corresponding data for Justin's 133 station subset (Bob, I sent you a 1K text file with this data in it):

Year Temp #stations
1900 -1.13 10
1901 -0.53 11
1902 -1.97 13
1903 -0.76 14
1904 -0.66 12
1905 -1.47 12
1906 -0.27 13
1907 -1.12 15
1908 -0.88 14
1909 -0.98 16
1910 -1.17 13
1911 -0.79 15
1912 -2.09 17
1913 -0.92 15
1914 -0.25 16
1915 -1.64 18
1916 -1.58 17
1917 -0.31 18
1918 -0.73 16
1919 -0.98 10
1920 -0.36 14
1921 0.26 16
1922 -0.71 17
1923 -0.67 16
1924 0.13 15
1925 -0.28 18
1926 -0.04 16
1927 -0.36 23
1928 -0.67 23
1929 -1.24 28
1930 -0.40 29
1931 -0.85 29
1932 0.31 31
1933 -1.04 40
1934 0.41 46
1935 -0.30 52
1936 -0.18 62
1937 -0.17 63
1938 0.32 63
1939 0.44 70
1940 -0.37 103
1941 -1.07 105
1942 -0.04 109
1943 1.01 109
1944 0.53 107
1945 0.03 114
1946 -1.12 119
1947 -0.40 120
1948 0.55 126
1949 -0.20 122
1950 -0.65 128
1951 -0.11 131
1952 -1.39 131
1953 0.34 132
1954 -0.45 130
1955 -0.65 131
1956 -0.46 131
1957 -0.67 132
1958 -1.37 132
1959 0.20 132
1960 -0.98 133
1961 -0.33 132
1962 0.54 133
1963 0.15 131
1964 -1.07 131
1965 -1.12 133
1966 -1.65 131
1967 0.37 131
1968 -0.60 123
1969 -1.53 130
1970 -0.73 130
1971 0.00 129
1972 -1.03 133
1973 0.09 129
1974 -1.37 124
1975 0.54 131
1976 -0.99 125
1977 -0.92 128
1978 0.05 130
1979 -1.15 128
1980 -0.28 133
1981 0.72 119
1982 -0.42 113
1983 1.02 130
1984 -0.50 132
1985 -0.23 123
1986 0.39 129
1987 -1.41 133
1988 0.59 128
1989 0.64 117
1990 0.79 124
1991 0.69 126
1992 -0.05 117
1993 0.49 116
1994 0.11 115
1995 1.89 116
1996 0.14 113
1997 0.97 119
1998 -0.45 120
1999 -0.15 115
2000 -0.01 115
2001 0.46 105
2002 1.13 106
2003 1.28 114
2004 0.19 118
2005 1.43 118
2006 0.05 117
2007 2.13 118
2008 1.52 113

Justin's subset of Siberian stations shows the same basic characteristics of the larger data set. Linear regression shows a positive slope and has an r-squared of about 0.27 as before. Moving to a cubic increases the r-squared to 0.39, higher degrees don't increase the fit much. The cubic fit shows a rising trend to about 1930 and a rising trend starting in the mid 1970's with a broad flattish section in between. Perhaps Bob can post a plot.


What is Justin's hypothesis?

Originally Posted by Justin '77

Umm. Siberia isn't.

The link won't load.

What's really fun is looking at the areas that the warmists' models leave blank (and then fill in with warming).

Looks like lots of warming in Siberia

Here's an app that will give you raw data (direct from the Russian meterological service; their page) from 12-some-odd percent of the earth's land surface. Station list with coordinates.

I downloading all the montly data for all stations for all 12 months for the 1900-2008 period. I see warming.

All those warming Siberian regions turn out to have not been warming particularly much at all.

You provided a list of 133 Siberian stations. These show warming just like the the entire set of stations does. What are you talking about?

Originally Posted by Justin '77

...HADCRUT cherry-picked a tiny sample from the 400-some-odd meteostations throughout the country. And that their cherry-picking leads to conclusions directly at odds with the data given by the stations as a whole,

I analyzed the data for all 400+ stations. It shows a warming trend constitent with that shown by the HADcrut data.

or by picking based on scientifically-valid criteria (like station longevity, avoidance of UHI, station quality, maximum coverage, or so forth).

I did the same analysis for the stations YOU picked and got the same result. What are you talking about.

So it makes sense that the cherry-picked data would give the result that people say they got from the cherry-picked data. You need to get the raw data to even see what the point is of the criticism. And it's there and open.

Yes and it doesn't say what you claim it says.

Of course, your dismissal of data in favor of ideology is par-for-the course.

You are the one doing the dismissing.

Originally Posted by Justin '77

Here's four stations (fourth one on the next post; the forums software doesn't like more than four pics in a post) chosen for having been in the first few that I did (I started working backwards by station numbers).

These all show warming. I'm not following you.

Originally Posted by Justin '77

...the world has seen a long-term gradual warming trend which shows up in the overall slope of every graph's trend.

This is precisely the point.

BUT, when you look at the timeframe where -- according to AGW canon -- global warming took off and the siberian permafrosts started catastrophically melting, what I keep seeing over and over is a multi-decade level/cooling.

Huh? The first two graphs show 2007 as the warmest year, which is what the larger set of data show, had you bothered to look at it.

There's no '1998-was-the-hottest-year' spikes at all, either. Which is kind of a big surprise to me, even.

Why? Do you seriously think that the warmest year on record globally should translate to the warmest year on record locally? Particularly when the warmest year on record in the general region of your concern was not in 1998, but in 2007?

Originally Posted by Justin '77

TExcept that in the Siberian ones I've gone through so far... it doesn't. It pokes up from 60-80, and then level or down from mid-80s through now.

Those you posted didn't show this pattern. Not does the composite of the 133 stations your gave. What are your talking about?

The AGWarmist line is currently working to hand-wave away the 2000's cooling trend, but the Siberian data seems to challenge even the 1990s AGW trend.

The Siberian, indeed the entire database shows a warming trend into the 2000's with record temperature in 2007.

Originally Posted by Justin '77

By the way, Tara, since it appears in the CRU dataset, is exactly one of those cherry-picked stations that shows warming which is not characteristic of the climate trends in siberia as a whole.

How do you know this? After all the entire set of Siberian stations that YOU provided DOES show warming.

So the fact that it shows warming is... duh... completely noncontroversial.

No its completely expected, unlike the stuff about no warming in Siberia that you are apparently making up.

If you want to have something to say -- be it criticism or confirmation -- you really need to have something better than a total cock-up pile of garbage for data.

I used YOUR data source and YOUR selected stations and got results in accord with what Bob says.

[quote[It's kind of sad that, having access to as many datasets as we now have, you seem to think that presenting two pictures of god-knows-what (you said yourself -- even you don't know what they are!) are an appropriate accompaniment to a reasoned argument.[/quote]
Where are YOUR data? I am getting tired of your abject lying.

Since you've expressed a distaste for finding answers if they are not handed to you on a platter

I did the work, and my results agree with what Bob's position. Did you do any work yourself, or are you talking bullshit as usual?

Originally Posted by Mikebert

I did the work, and my results agree with what Bob's position. Did you do any work yourself, or are you talking bullshit as usual?

Mike, as best as I can figure out, most of the 133 stations on Justin's list should be strongly effected by the Arctic Oscillation. In the early 1990s, there was a solar maximum stacked on top of an extreme warm time in the Arctic Oscillation. Currently, the solar cycles are near a minimum and the Arctic Oscillation is more or less neutral. For many of Justin's 133 Siberian stations, I suspect there will be a modest decrease in temperature from the early 1990s to present.

(Not all of the 133 stations, though. Way up north, Glomjannyi and Sterlengova do not show the early 1990s warm time found in stations further south. I suspect there will be other far north and perhaps far south stations that have not had in a cooling trend since 1990.)

You might want to see if there is a statistically significant slight drop in temperature over Justin's 133 stations from 1990 to present. That is Justin's hypothesis. I think everyone agrees that there is a warming trend from 1950 to present.

However, I don't think, even if a cooling trend from 1900 to present is there and significant, which it might well be, that this disproves the global warming hypothesis. Justin's down trend is only generated by cherry picking peak to trough tend periods. He is arbitrarily choosing a start time when two major forcing factors are at an extreme high point.

All in all, both the solar forcing and arctic oscillation are flat in the long term. Well, solar forcing might possibly be coming slightly down. Arctic oscillation might possibly be going slightly up. But neither can account for the long term warming trend.

Originally Posted by Bob Butler 54

Mike, as best as I can figure out, most of the 133 stations on Justin's list should be strongly effected by the Arctic Oscillation. In the early 1990s, there was a solar maximum stacked on top of an extreme warm time in the Arctic Oscillation. Currently, the solar cycles are near a minimum and the Arctic Oscillation is more or less neutral. For many of Justin's 133 Siberian stations, I suspect there will be a modest decrease in temperature from the early 1990s to present.

(Not all of the 133 stations, though. Way up north, Glomjannyi and Sterlengova do not show the early 1990s warm time found in stations further south. I suspect there will be other far north and perhaps far south stations that have not had in a cooling trend since 1990.)

You might want to see if there is a statistically significant slight drop in temperature over Justin's 133 stations from 1990 to present.

The slope over 1990-2008 is positive with slope 0.043 plus or minus 0.030. Significance is 86%.

To get a better assessment of the slope a longer time period is needed.

Originally Posted by Mikebert

Bob is right. The data I presented so far is for all the stations. It is simple matter to do the same for any subset of those stations. Here is the corresponding data for Justin's 133 station subset (Bob, I sent you a 1K text file with this data in it):

Year Temp #stations
1900 -1.13 10
1901 -0.53 11
1902 -1.97 13....


Justin's subset of Siberian stations shows the same basic characteristics of the larger data set. Linear regression shows a positive slope and has an r-squared of about 0.27 as before. Moving to a cubic increases the r-squared to 0.39, higher degrees don't increase the fit much. The cubic fit shows a rising trend to about 1930 and a rising trend starting in the mid 1970's with a broad flattish section in between. Perhaps Bob can post a plot.

What is Justin's hypothesis?

Plotting out the data from post 1813...

I'd agree that the post 1813 curve looks quite compatible with the global data. Justin has been picking specific cities which show warm years in the early 1990s, and claimed that this would hold over the 133 stations. Assuming Mike gathered his data correctly, Justin's claim of multi decade cooling in Siberia as a whole seems to be falsified. There might be a few stations that have been cooling since 1990, but Siberia as a whole as represented by Justin's 133 stations has not been cooling.

Mike, let's see if I might restate Justin's hypothesis (as I understand it) in a way which might be examined.

There are stations in Siberia where it was slightly warmer from 1989 to 1998 than it was from 1999 through 2008. Taishet seems to be one such example. My favorite Perevoz is another. Justin has posted several others. I believe these stations will be those where the effects of the Arctic Oscillation are strongest. At this point, I am mildly curious as to where the Arctic Oscillation might be strong.

Could you generate a number for each of the 133 stations showing whether temperature rose or fell from 1998 to present? Use whatever method seems appropriate for judging the slope of the temperature over that time period. This would allow us to judge just how many stations Justin's hypothesis holds up on, how strongly, and where they are located. If you could include longitude and latitude of each station in the sorted data as well, this would be good. If you could provide an estimate of statistical significance, such as we can divide the stations between those showing significant warming, significant cooling and those unclear in the middle, this might be good.

If you do the sort, I'll generate the charts and crude maps of station locations trending in either direction.

Mind you, I do not believe that the existence of the Arctic Oscillation disproves the global warming hypothesis. Identifying a limited area where the temperature fell over a specific time interval does not invalidate the global trend.

James Hansen's The Temperature of Science (pdf) is mostly about checking the influence of the skeptics. It does, however, review some points of controversy over the years. I'd like to work one point that few want to talk about. Let's start with a pair of diagrams from The Temperature of Science.

On the left hand chart, one sees a pause in the increase in global temperatures, or something of a decline, from 1940 to about 1965. No one doubts more fossil fuel was being burned. How does one care to explain this? Hansen suggests...

Originally Posted by James Hansen

The different hemispheric records in the mid-twentieth century have never been convincingly explained. The most likely explanation is atmospheric aerosols, fine particles in the air, produced by fossil fuel burning. Aerosol atmospheric lifetime is only several days, so fossil fuel aerosols were confined mainly to the Northern Hemisphere, where most fossil fuels were burned. Aerosols have a cooling effect that still today is estimated to counteract about half of the warming effect of human-made greenhouse gases. For the few decades after World War II, until the oil embargo in the 1970s, fossil fuel use expanded exponentially at more than 4%/year, likely causing the growth of aerosol climate forcing to exceed that of greenhouse gases.

On the right hand chart, one sees this pause is location dependent. There is a definite decline in temperature in the Northern Hemisphere, while there is a slow steady rise in the Southern. The hypothesis is that sooty sulphates fall out of the atmosphere faster than CO2 is absorbed. Sulphates are temporary, where CO2 and other greenhouse gasses linger far longer. Thus, Sulphates do not stay in the air long enough to produce global dimming in the southern hemisphere.

It seems to follow that burning fossil fuels in a dirty way results in cooling. One sees pictures cities like London and Pittsburgh during the height of their polluting days. They were very sooty places. Modern Peking and other cities in China and India are not that bad, but still release considerable soot.

I note that it is easier to scrub soot than prevent the release of CO2. As soot is a grave health hazard, it is expected that the developing world will cut soot sooner and more drastically than it cuts CO2.

Let's switch from 'official' graphs to a spreadsheet I invented on the fly. Note, my chart goes back a bit further in time in order to pick up the Little Ice Age. Everything on the following chart is global.

Blue is global temperature. It's a hockey stick chart. One notes that the shaft of the hockey stick is really crooked. Roughly from 1625 to 1725 we have the little ice age which puts a significant dip in the hockey stick. There is a similar even littler ice age around 1790 to 1830 which also bends the hockey stick.

In red we have solar forcing. How much energy did we receive from the sun? It is fairly obvious that the Little Ice Age came from a strong reduction in solar energy. If one looks at the red curve, one can make a case that there has been a slow (and more or less) steady increase in solar energy since the Little Ice Age. As one sees the curves move up and down more or less together, it is very hard to say solar is not an important factor in global climate.

In brown, we have global temperature minus solar forcing. I deliberately chose scaling factors for the global temperature blue line to generate the straightest and purist hockey stick I could. My objective was to straighten the shaft of the hockey stick so as to get rid of the down bump at the Little Ice Age. To do this, I played with the scale of the global temperature curve... which is in degrees C multiplied by 100 plus 1358. The times 100 scaling factor was chosen by trial an error to deliberately straighten the shaft of the hockey stick during the 1600 through 1800 time frame, when man wasn't burning all that much fossil fuel.

I'm willing to send a copy of the spreadsheet to anyone who wants to try for a straighter shaft. I think I did OK, but if anyone else wants to give it a try, sure....

The black is just a rolling average of the brown. It gets rid of a lot of noise to make the long term hockey stick more visible.

Now, if solar forcing were the dominant cause of the warming since 1900, if solar effects caused both the Little Ice Age and the after 1900 global warming, then it should have been possible to get rid of both the hockey stick blade and the Little Ice Age. It can't be done. Again, I'll send a copy of the spreadsheet to anyone who wants to try. If anyone wants to start from scratch, I'd recommend NOAA's paleoclimatology pages as a good place to grab data.

I note that from 1800 through about 1900, the temperature drops down. This is true in the blue temperature record and more clearly true for the brown and black temperature minus solar curves. I shall propose a hypothesis. Burning fossil fuels without scrubbing out the sulphate soot produces a cooling effect. This is true at a global level. However, as soot falls out of the air faster than it travels from the northern to southern hemispheres, one might see warming in the south during times when there is cooling in the north.

In the 20th Century, the soot became unacceptable. Efforts were made to clear it up. At that point, the ratio of CO2 released with respect to sulphate soot reduced drastically. Thus, global warming only really started as ecological and medical concerns led to an attempt to get rid of the soot.

If places like India, China and Brazil also start cleaning soot out of their emissions, we might find ourselves with a sharp increase in global warming.

At the bottom, we have a graph for sulphates released from volcanoes. Note, the scaling factor is negative. The dips in that curve come after major volcanic eruptions. I wanted the curve to go in the same direction as the temperature, thus I flipped the volcanic curve upside down. Alas, the data is released in 5 year intervals, rather than yearly. I decided not to attempt to subtract volcanic forcing factors from the temperature. However, note there are significant downturns in temperature following big eruptions.

There are still lots of bumps in the temperature record that cannot be attributed to either volcanic or solar effects. I suppose I could plot El Nino effects as well. Still, a lot of the noise might well be just weather. There also seem to be distinct limits to spreadsheet climate models. I don't think I want to get more complex that the above.