Geoscience Reference
In-Depth Information
1 in the form of Relative Absolute Errors (RAEs). The reported RAEs are the absolute
erroroftheforecastsfromthehypothesisrelativetothecorrespondingabsoluteerrorofthe
forecasts from the persistence (no-change) model over the forecast horizon. Thus, a figure
of 0.5 means the error was only half as big as that from the persistence model forecast, and
2.0 means it was twice as big.
Evidence-based climate forecasts for the 21
st
century
Our testing used alternative data sources, different time periods, different starting points,
and different horizons. The findings were always the same. Forecasts from the more
conservative cooling hypotheses were more accurate than forecasts from the warming
hypotheses. Forecasts from the most conservative hypothesis, the no-trend model, were
always much more accurate. The no-trend model is consistent with evidence-based
forecasting principles and with the state of knowledge about the behaviour of the Earth's
climate.
Table 1: Relative accuracy of forecasts from alternative climate change hypotheses,
warming, and cooling, versus persistence
* Monthly forecasts. **Successive updating used.
The IPCC's alarming-warming model is not. Consistent with knowledge about the proper
modelforthissituation,thepredictivevaliditytestsfindsnosupportfortheglobalwarming
hypothesis for forecasting global mean temperatures over this century and beyond.
