I thought I’d announce my prediction for the temperature anomaly for 2008. I predict the average temperature anomaly for 2008, as reported by GISS Land/Ocean measurements will be 0.70 ± 0.11 C.

The ±0.11C is my hookey claim for the standard error. That means if my model is halfway decent, there is roughly a 2/3 chance the temperature will fall in that range. (I figure if I’m going to predict, it’s silly to give myself 95% error bands. I don’t know how to calculate them anyway. )

For your enjoyment, I’ve plotted my best estimate predictions of the temperature anomaly in orange shown with ±0.1C uncertainty intervals.

The green curve indicates the “steady state temperature” my model predicts if a level of forcing for a given year were sustained be sustained forever rather than increasing or decreasing the following year. I don’t have forcing data after 2003; the curve shown assumes that the forcing will increase at a constant rate of 0.48 W/m² after 2003.

Explanation of uncertainty intervals.

Assuming my semi-empirical lumped parameter model is plausible, forecasts have four sources of uncertainty: These are:

1. Uncertainty in my two model parameters. I don’t know the magnitude of these. For the purpose of today’s forecast, I’m going to assume the uncertainty associate with these parameters is zero. (The real uncertainty is not zero. )
2. Uncertainty in recent values of forcing. I currently only have forcing values up to 2003. Forecasting future temperature using my model requires me to estimate the increase in forcing since that time. I did this three ways: 1) I assumed the forcing would remain constant, 2) I assumed forcing would increase at the average rate for the past 5 years, 3) I assumed forcing would increase at the average rate for the past 10 years. This resulted in low, medium and high forecasts of 0.69 C, 0.70 C and 0.71 C.

   That is to say, unless a major volcano erupts, the uncertainty in recent rate of increase in GHG’s makes little difference to my prediction for next year.

   My model predicted relatively low sensitivity to doubled CO₂, but it also has a long time constant. As a consequence, it predicts there is warming ‘in the pipeline’.

3. Random factors not captured by the model. My model treats the climate as one big lump. Needless to say, it’s not perfect. The residual errors for my model is 0.1 C . This includes the 0.05C uncertainties. associated with measurement error anticipated by GISS. I don’t know what other factors contribute to this variation; they may include La Ninas, El Nino’s or persistent weather patterns.
4. Volcanic eruptions. My model can’t predict when a volcanic eruption will occur. Predicting temperature with my model requires the user to first predict the forcing; large volcanic eruptions dramatically affect that forcing. If a major volcano erupts, particularly early in the year, the average annual temperature will be lower than forecast based on my model.