Last spring, I noted that recent global temperatures seemed to be primarily a combination of a long-term trend and modulation by El Niño and La Niña (here and here). In September, I took this idea one step further and issued global temperature anomaly forecasts for 2012 and 2013 based entirely on a linear trend plus ENSO (here). Today, we see how the 2012 forecast turned out, and update the forecast for 2013.
The forecast method is about as simple as I can make it. I take computer model forecasts of the ONI (Optimal Niño Index) average value for 2013. With about 20 different forecasts available, that gives me 20 possible ONI forecast values. I compare those values to the period 1979-present to see how many fall into the warmest third, middle third, and coolest third historically. Then, I regress (fit a straight line to) the observed global temperatures during the warmest third ONI years, the middle third, and the coolest third, with ONI leading temperatures by three months.
The ONI forecast ensemble, the trend lines, and the scatter about the trend lines give me a forecast of global temperature complete with error bars. The forecast error bars below span the 10% to 90% probabilities.
Note that this approach should fail miserably if there is no underlying warming trend, so it constitutes a test of whether the long-term warming trend is still there.
According to linear regression, this ongoing warming trend is 0.12-0.18 C per decade.
I make separate forecasts for each global data set. The data sets use different base periods to calculate anomalies, and I do not attempt to align them to a common base period