Geoscience Reference
In-Depth Information
Very long-term testing of predictive validity
In order to assess the validity of the hypotheses over very long horizons, we tested the
accuracy of forecasts from warming, cooling, and no-trend model hypotheses against the
naturallyoccurringrecordsofbiologicalandphysicalprocessesthatvarywithtemperature.
The Loehle series was constructed from eighteen series obtained and calibrated by
other researchers who used such proxy records as boreholes and pollen counts that each
covered most of the Common Era and, between them, covered much of the globe. The
resulting Loehle series extends from AD 16 to AD 1935, allowing us to test forecasts from
variationsofthehypothesesforhorizonsofuptonearly2000years.Theseriesincludesthe
Medieval Warm Period and the Little Ice Age. Evidence suggests that the current climate
is not as warm as that of the Medieval Warm Period when cows grazed and willows grew
in Greenland and seals basked on the shores of Antarctica.
36
A forecaster living 100 years after the beginning of the Loehle series in AD 115 might
reasonably have forecast that the average temperature trend that had prevailed over the
previous 100 years, an increasing one of roughly 0.003°C per year (0.3°C per century),
wouldprevailindefinitely.Indeed,someresearchershavesuggestedthattheEarthhasbeen
increased as the forecast horizon lengthened as the dashed line in Figure 2 shows.
Figure 2: Absolute errors of warming, cooling, and no-trend forecasts
By year from AD 115 to AD 1935, in degrees Celsius.
Source: K.C. Green and J.S. Armstrong
