OT: Global Temperatures analysis. Was: Breaking News

[Chris Hogg]

I thought it worth starting a new thread for this.

In the Breaking News thread I said I was going to do analyses of the Global Temperatures vs. Date data, for my own satisfaction. I have no doubt that there are plenty of highly qualified statisticians out there who do a much better job, but this is uk.d-i-y after all!

The data was downloaded from here:

and the individual columns are explained here: . The data was copied and pasted into EXCEL 2003, and converted into separate columns using the 'Text to Columns' function. All but columns

1 (Date) and 2 (median temperatures) were deleted. There are 1990 pairs of data.

When it came to analysing the data, I had to improvise to get them into some sort of numerical form because EXCEL doesn't automatically handle dates before 1900. Fortunately the temperature records are at regular monthly intervals, so I just numbered them 0 - 1998, and converted those numbers to date-related numbers: B(n)/12+1850, where n is the record number between 0 and 1989. In effect this converts the year and month to a year and decimal fraction of that year, so March

1912 (record 748) becomes numerically 1912.167 and September 1944 (record 1136) becomes 1944.667, IYSWIM. Graph plotting (scatter graphs) and linear regression of the data (using the LINEST function) was then straightforward, if tedious.

The resulting graph and fitted lines are shown here:

Visual inspection of the data suggested six approximately linear portions of temperature change vs. date. The graph shows linear regression fits to those sections, in red, and a single fit to all the data, in blue, together with their respective gradients (deg.C/year) and the figures for 2 standard deviations, 2-sigma, on those gradients.

The results are summarised here (if it tabs OK!): Date Range Mean 2-sigma Max Min rise rise rise °C/y °C/y °C/y Jan 1850 - Dec 1879 0.0039 0.0025 0.0064 0.0014 Jan 1880 - Dec 1909 -0.0053 0.0018 -0.0035 -0.0072 Jan 1910 - Dec 1944 0.0142 0.0013 0.0155 0.0128 Jan 1945 - Dec 1974 -0.0012 0.0017 0.0005 -0.0029 Jan 1975 - Dec 1999 0.0189 0.0021 0.0210 0.0168 Jan 2000 - Oct 2015 0.0103 0.0036 0.0139 0.0068

Jan 1850 - Oct 2015 0.0049 0.0002 0.0051 0.0047

The figures for Max rise and Min rise are obtained by adding or subtracting 2-sigma values from the Mean rise.

The overall results show a temperature rise from 1850 to 2015 of between 0.0047 and 0.0051°C/y. It has been suggested that this is simply on-going slow recovery from the last ice age. I am not qualified to comment.

Shorter periods of faster rising temperatures occur from 1910 to 1945 and from 1975 to 2000. The former rises between 0.0128 and 0.0155°C/y and is attributed to increasing solar activity over that period

, while the latter rises a little faster at between 0.0168 and 0.0210°C/y, but this is attributed to rising CO2 levels, although I note that solar activity was also increasing over that period, arguably more so than between 1910 and 1945.

Between these two periods of rising temperatures, i.e. between 1945 and 1975, is a period of slightly declining temperature (but virtually horizontal within the statistics). Sunspot activity peaked during this period, but aerosols are said to have peaked after the war, although their influence seems uncertain

. I would have thought aerosols and dusts would behave similarly, and the influence of dust is apparently also uncertain What we've been led to believe, that the climatologists have a good understanding all the factors affecting global temperatures, doesn't seem to be the case. So what else haven't they discovered that they need to include?

There's an earlier period of cooling, between 1880 and 1910. I don't know of an explanation, but that doesn't mean there isn't a perfectly adequate one; I've just not seen it.

The last section, between 2000 and present, shows a continuing rise in temperature, although at about half the rate of the preceding period. Commencing the last section at Jan 2001 rather than Jan 2000 and ending at Dec 2014 results in an almost horizontal line (not shown),

0.0026°C/y, 2-sigma 0.0038, consistent with the oft-stated claim in recent years that the temperature has stopped rising. This implies that the warming in the latter part of the last century ceased at the end of 2000 (rather than the end of 1999 as I've chosen to plot it) but may have started again at the beginning of 2015, but of course, you can't say anything based on just one year's data.

I remain unconvinced about AGW. Other hypotheses, such as Abdussamatov's

or Zharkova's, that I've referred to before, which predict a cooling over the next few decades or so, have been largely eclipsed by the AGW hypothesis, but their effects should start to be seen in the next few years if they're correct. Time will tell.

[Chris B]

Not at all. Any reductions in temp in the next few years will all be to do with this wonderful agreement being finalised this very morning.

[RJH]

I couldn't begin to understand how to extrapolate AGW from domestic weather, but it does look interesting. I'd like to know far more about how the geographies are chosen - looks suspiciously like convenience to me. Those other models rightly question the cause-correlation trap, but as you say, remain to be tested properly.

What are R the R2 figures? From a quick look at your scatter plot it all looks pretty weak as a model of anything.

[Chris Hogg]

This may give you an idea of the geographic spread of monitoring stations used

It's supposed to come from but it's not there AFAICT, although the Q/A are interesting in themselves. Otherwise, you may find something in this lot

Here are the EXCEL outputs for each of the six sections, the shorter section for this century, and the whole data set. I hope they tab OK. If not, they're in four columns: the middle two columns are the numbers, first and fourth columns are what those numbers refer to. Hope it answers your query.

Jan 1850 - Dec 1879 Slope 0.003899086 -7.58 Intercept St.Dev. Slope 0.001247619 2.33 St.Dev. Intercept R^2 0.026557598 0.21 St.Dev. Y Fisher F Stat 9.767008517 358.00 Degs. Freedom n-2 Regression SS 0.410474291 15.05 Errors ss unexplained variat'n

Jan 1880 - Dec 1909 Slope -0.005318427 9.72 Intercept St.Dev. Slope 0.000919802 1.74 St.Dev. Intercept R^2 0.085412194 0.15 St.Dev. Y Fisher F Stat 33.43316535 358.00 Degs. Freedom n-2 Regression SS 0.763707036 8.18 Errors ss unexplained variat'n

Jan 1910 - Dec 1944 Slope 0.014163456 -27.50 Intercept St.Dev. Slope 0.000656914 1.27 St.Dev. Intercept R^2 0.526538273 0.14 St.Dev. Y Fisher F Stat 464.8591114 418.00 Degs. Freedom n-2 Regression SS 8.600826166 7.73 Errors ss unexplained variat'n

Jan 1945 - Dec 1974 Slope -0.001187159 2.27 Intercept St.Dev. Slope 0.000842755 1.65 St.Dev. Intercept R^2 0.005512276 0.14 St.Dev. Y Fisher F Stat 1.984333096 358.00 Degs. Freedom n-2 Regression SS 0.038052052 6.87 Errors ss unexplained variat'n

Jan 1975 - Dec 1999 Slope 0.018878279 -37.39 Intercept St.Dev. Slope 0.001059615 2.11 St.Dev. Intercept R^2 0.51577457 0.13 St.Dev. Y Fisher F Stat 317.4158405 298.00 Degs. Freedom n-2 Regression SS 5.568522837 5.23 Errors ss unexplained variat'n

Jan 2000 - Oct 2015 Slope 0.010337477 -20.27 Intercept St.Dev. Slope 0.001791678 3.60 St.Dev. Intercept R^2 0.150434709 0.11 St.Dev. Y Fisher F Stat 33.28964293 188.00 Degs. Freedom n-2 Regression SS 0.424164349 2.40 Errors ss unexplained variat'n

Jan 2001 - Dec 2014 Slope 0.002593417 -4.72 Intercept St.Dev. Slope 0.001934306 3.88 St.Dev. Intercept R^2 0.010712932 0.10 St.Dev. Y Fisher F Stat 1.797604397 166.00 Degs. Freedom n-2 Regression SS 0.018454969 1.70 Errors ss unexplained variat'n

All Data (Jan 1850 - Oct 2015) Slope 0.004881359 -9.54 Intercept St.Dev. Slope 9.34355E-05 0.18 St.Dev. Intercept R^2 0.578575957 0.20 St.Dev. Y Fisher F Stat 2729.338828 1988.00 Degs. Freedom n-2 Regression SS 108.6668015 79.15 Errors ss unexplained variat'n

[newshound]

The data which I would like to get hold of is that behind the second graph in this article

The *first* graph shows the same sort of trend as your data, but the

*second* shows a gradual increase over the past 8000 years. To my mind, it is the comparison of (say) the last 300 years with the previous 8000 which would be interesting.

It's a basic principle in predicting from any data set which shows significant historical variations that you need to look at the longer as well as the shorter timescales. This is what I find most frustrating about the "soundbite" analyses based on ten or 20 years of rainfall, temperature, or whatever which are so popular with the lobbyists and the media. And they are sometimes advanced by competent and reputable "scientists" who should damned well know better.

[newshound]

As has been pointed out elsewhere, these are dominated by industrialised regions and a proportion of them will be in urban areas. The urban "heat island" effect has been well documented and unless all such measurements are removed from the analysis, I find it hard to give any credibility to the data.

Maybe someone has done this, but I don't think it is well publicised.

Don't get me wrong, carbon dioxide levels have risen significantly, it is a greenhouse gas, and there could well be a real effect. I'm just not convinced that we know what it is.

[The Natural Philosopher]

Probably about 0.1-0.2 degrees C.

Completely insignificant.

[Tough Guy no. 1265]

Good grief, some common sense. Go tell the government to stop wasting m= oney!

-- =

"I'm prescribing these pills for you," said the doctor to the overweight= patient, who tipped the scales at about three hundred pounds. "I don't want you to swallow them. Just spill them on the floor twice a= day and pick them up, one at a time."

[Tough Guy no. 1265]

Indeed. The only reasons for "global warming" I can see are bribed scientists, corrupt governments, and company desires.

[michael adams]

Not sure whether its of interest or not but on an accompanying abstract

" Quantifying uncertainties in global and regional temperature change using an ensemble of observational estimates: the HadCRUT4 data set "

One conclusion at the foot of page 1 is as follows

Climate diagnostics computed from the gridded data set broadly agree with those of other global near-surface temperature analyses.

Fitted linear trends in temperature anomalies are approximately

0.07 degC/decade from 1901 to 2010

and 0.17 degC/decade from 1979 to 2010 globally.

Northern/southern hemispheric trends are

0.08/0.07 degC/decade over 1901 to 2010 and 0.24/0.10 degC/decade over 1979 to 2010. Linear trends in other prominent near-surface temperature analyses agree well with the range of trends computed from the HadCRUT4 ensemble members.

They're claiming temperature anomalies twice as large, post 1979.

AIUI anyway the claim is that the rate of change in temperature over a specific period is the largest on record, while only CO2 concentrations while only a small component in greenhouse gasses overall among measurable factors show a similar increase.

Whether this is nonsense or not is another matter, but that AIUI that is what's being claimed.

michael adams

...

[Chris Hogg]

No disagreement about the relative sizes of the temperature anomalies over the two time periods. It's worth noting that another anomaly, from 1910 to early 1940's also occurred, and is also regarded by Jones as significant, but explained by increasing solar activity (

in his reply to the 1st question).

But it does seem a bit desperate to invoke CO2 as the explanation for the rise in temperatures in the later part of the 1900's simply because it's the only thing that increased sharply over that time, and especially since there seems to be some doubt as to whether CO2 could actually bring about the changes in global temperatures that have been seen, bearing in mind that the most abundant greenhouse gas, water vapour, absorbs 68% of the solar radiation in the infra-red region, and CO2 only 12%

Fig.4.