Here’s everything you need to know about global warming on one chart:

The top pane shows the decimated Hadcrut3 data series we’ve been using. The second pane shows mode 2 of the SSA decomposition. The third pane reconstructs the series with all 35 modes, excluding mode 2. The last pane shows that if we add mode 2 back in, we can re-construct the original series exactly (both are overlayed on the plot).

Focusing on pane 3, where we’ve removed the mode 2 contribution, we see a linear trend with a slope of ~.5 degsC per century which has remained constant across the entire 113 year span, with no sign of acceleration due to anthropogenic effects. Thus, if there is any signature of AGW to be found, it must be found in the contribution we’ve excluded, mode 2 of the SSA decomposition.

Note that this conclusion is completely independent of the method of decomposition. If I claim an angel from on high appeared suddenly and handed me the extracted mode plotted on pane 2, the conclusion above would be scientifically valid, as long as you accept the laws of addition. This is shown in the last pane which proves that pane 3 (the reconstruction excluding mode 2) + pane2 (mode 2) = Observed Temperature Record, as it must. Pane 3 shows no AGW signature, therefore it must be in mode 2 if it is in the observed temperature record at all.

The conclusion that any sign of AGW must appear in the mode 2 component is also free of any assumptions about the nature of mode 2 or of the underling climate system as a whole. It depends only on the observed temperature record.

Looking at mode 2 (SSA[35,2,t] plotted as a time series in pane 2), we see a periodic, sine-ish looking signal tilted upward. I’ve shown a linear regression with a slope of about .25 degC per century but the slope of this line depends on where we choose to start and stop the record. In fact it is just the average value of the instantaneous slope (year-to-year difference) plotted below.

First difference SSA Mode 2

First difference SSA Mode 2

Note that the maximum slope achieved in 1993 barely exceeds that of 1923. That alone disproves the AGW effect (remember – if there’s AGW it has to be seen in mode 2). But look closely at 1980. There is a suspicious step change in the slope that is undoubtably something hinky with the dataset. If one were to look back at the history of Hadcrut3, I’d bet dollars to donuts there was an “adjustment” made to the data from 1980 onwards – probably because the data wasn’t matching the GCM models.

Hinkiness aside, if somehow the climate should assume the maximum temperature slope ever seen in the past 113 yeas (actually the past 163 years since the slope from 1850 to 1900 was down to flat) and stayed constant at this extreme, the most we’d see is 1.85 degs/century (.5 degsC/century from the overall trend in pane 3 plus the worst-case 1.35 deg/century above).

Of course that’s not going to happen. The mode 2 slope is negative, as is the mode 1 slope  (see the trend analysis in part 2) but future trends aside, the main point is, there is absolutely no AGW signature to be found in the observed surface temperature record. There is nothing to indicate the last half of last century warmed any faster than the first half, and if my suspicions about the dataset are correct, the maximum rate in the second half was substantially lower. This obviates the need for forecasts altogether, since there is no evidence mankind is having any effect whatsoever on the global surface temperature.

Note to SkepticalScience readers The foregoing is not a model, it is a deconstruction. It makes no predictions, projections or divinations. It cannot be validly backcast, forecast (but it should be broadcasted – widely). It makes no assumptions about forcings, debunked or otherwise, and is valid whether the extracted mode is internal,external or a fluke. It depends only on observed data and the Law of Addition.

Correcting the Record

Just for fun I’ve taken the step out of the mode 2 first difference (it turns out it was between 1979 and 1980), reintegrated to create a “corrected” mode 2, and used the corrected mode 2 to reconstruct the dataset. The difference is a significant 0.12 degrees overestimation of current warming caused by the adjustment.

Adjusted Hadcrut3

Adjusted Hadcrut3