Geoscience Reference
In-Depth Information
temperatures, we determined that the benchmark model that is most consistent with the
state of knowledge is one that forecasts no change in the level; in other words, no trend.
Wecomparedtheforecastsfromtheno-trendmodelwiththeforecastsfromthecooling
and warming hypotheses. We found that the average error of the no-trend forecasts was
smaller than the average errors of both the warming and the cooling forecasts for
all
forecast horizons (Figure 1).
Figure 1: Average absolute errors of 0.03ºC warming, 0.01ºC cooling, and persistence
forecasts
Forecasts for 1851 to 1975 by forecast horizon.
Source: K.C. Green and J.S. Armstrong
The average errors of the warming forecasts (dashed line) and the cooling forecasts
(dotted line) over the short-term (one to ten years) were 45 per cent and 10 per cent
larger, respectively, than the average errors of the no-trend forecasts (solid line). The
average error of the no-trend forecasts for the longer-term horizons, from eleven to 100
years, was roughly one-quarter of the average cooling forecast error, and one-eighth of
the warming forecast error. In absolute terms, the average errors of the no-trend forecasts
were less than 0.20°C for all horizons out to 75 years; beyond that, the average errors
did not exceed 0.24°C . The small and steady forecast errors from the persistence model
suggest that the Earth's climate is remarkably stable over human-relevant timescales.
This is particularly remarkable given the claims by warming alarmists that we have been
experiencing 'unprecedented' changes in the climate over the period of the test.
34
