Tuesday, February 17, 2015
I computed a similar forecast for the global ocean temperature anomaly using data from Jan 1880 to the Aug 2013 and got similar results. The monthly data used, 20 year averages, fit and error bounds are shown in red and the additional data to the end of 2014 is in blue.
The actual anomaly data looks more like a random walk in this case and there is a more pronounced decrease in the projected temperature anomaly. It remains to be seen how reliable the prediction is. The global ocean anomaly appears to be much more stable than the global land anomaly.
Wednesday, February 11, 2015
The observed global land temperature anomalies agree quite well with the a prediction made in October 2013. The observed values are within the error bounds of the computed curve. The forecast was based on the data in red and the blue points are for the observed data after the prediction. The solid red curve is a smoothed version of the curve obtained by taking 20 year averages. Combining the deviations from the 20 year with the 20 year average produced a much smoother curve. The doted lines are the fit to smoothed curve and the error bounds.
Supplemental (Feb 11): One might ask why the "corrected" 20 year average works so well and the answer might be that there is a feedback mechanism involved. The global temperatures are somewhat volatile but there also appears to be some resistance to change which is what one would expect if the temperatures are near an equilibrium point. Canceling the deviations from the 20 year average is a mathematical way of adding some resistance to change which results in a stiffened curve.