Abstract
An information approach to the study of climate predictability is introduced. This approach uses various information measures based on entropy to define the limits of predictability and to quantify the information given by different climate components and boundary forcings about the prediction of future climate anomalies. Together with a set of basis functions defined through frequency dependent empirical orthogonal functions, this method provides a systematic way of studying the regional predictability of a complex system. The method is applied to study predictability of the atmosphere simulated by the NCAR CCM0 on a zonally symmetric land planet. By removing the ocean in the simulation, the time scale of the system has been greatly reduced. Hence, this simulation provides sufficient time samples for realistic study of the properties of the atmospheric component. Atmospheric variability as simulated under the various idealized boundary conditions is studied. This provides noise climatologies useful for climate sensitivity experiments. The space-time statistics of the simulated surface temperature field are fitted satisfactorily by a stochastic climate model with only five parameters. This suggests the usefulness of simple statistical models as guidance for climate sensitivity studies, and sampling considerations for model simulation or data collection. The limit of predictability of the first kind of the surface temperature field is found over two zonal bands: tropics (0°-20°) and mid-latitudes (40°-60°). The predictability as a function of spherical harmonic degree is also evaluated. Results are consistent with past studies of atmospheric predictability using random perturbation experiments. Extension of the above study to climate predictability of the second kind is discussed.
Leung, Lai-yung (1991). Variability and predictability of the surface temperature field on a zonally symmetric land planet. Texas A&M University. Texas A&M University. Libraries. Available electronically from
https : / /hdl .handle .net /1969 .1 /DISSERTATIONS -1235790.