RANDOM-WALK EXPECTANCIES FOR RECENT GLOBAL CLIMATE, AND IN AN ENHANCED GREENHOUSE WARMING

We partition the United Kingdom Meteorological Office Global Temperatu re Series (T(k)) using an exponential decay filter into a filtered ser ies (T(k)) and a difference series (T(k)' = T(k) - T(k)). For a decay time constant, tau almost-equal-to 0.85 years, T(k) is shown to be a g ood approximation to a random walk generated by a cumulation of normal ly distributed interannual temperature transitions, and hence T(k)' co ntains the predictable temperature signal. The standard deviation of t he T(k) series, sigma = 0.083K, which is about 1 1/2 that of the T(k)' series. From this partition, it is argued that tau is the decay time constant (e-folding time) for the global temperature series, and also by the elementary theory of damped oscillations, that the global cimat e system as represented by the global temperature) can only support fr ee oscillations of natural period less than T = 2pitau almost-equal-to 5 years, i.e. the QBO and ENSO signals. On assuming that sigma does n ot vary significantly over periods up to 20,000 B.P. we find that the expected maximum excursions of the random walks are consistent with th e actual inferred temperature variability. On the other hand, the proj ected temperature rise due to the enhanced Greenhouse effect possibly cannot be supported as a random walk by sigma. This suggests that the interannual structure of the climate system would change under these c onditions. This conjecture can be tested adequately only with climate models which correctly reproduce random walk behaviour. This is inhibi ted in published simulated temperature series from coupled models, pos sibly because of flux correction. An assessment of the likelihood of a change in the interannual variance, and of the ratio between its pred ictable and random proportions is clearly of utmost significance in th e Greenhouse debate, yet it appears to have received very little discu ssion.