Conditional statistical models: a discourse about the local scale in climate simulations

The local scale of climate plays two different roles; it is the scale at which people experience climate, so that it is the dominant scale of applied climate research, ranging from climate impact to forecasting weather in the atmosphere and the ocean. On the other hand, the local scale is not important in its details for the formation of the global climate. For the understanding, and simulation, of the global climate, the small scales matter only in a statistical sense so that their influence may be described by means of parameterizations. In the present essay, we demonstrate that both processes, 'downscaling' (the derivation of local information in climate change and climate variability simulations and in weather forecasts) and 'parametrization' (the description of the net effect of small scales on the larger scales) may formally be understood as the building of empirical models whose parameters are conditioned upon larger-scale features of the state of the atmosphere or ocean. It is suggested to acknowledge the presence of unknown processes by building downscaling and parameterization procedures with a randomized design, conditioned upon the known resolved scales. (orig.)Reprinted from: Mueller, P.; Henderson, D. (eds.): 'Aha Huliko'a Monte Carlo simulations in oceanography, School of Ocean and Earth Science and Technology, University of Hawaii, Honolulu, HI (US), SOEST Special Publication, 1997, p. 49-58Available from TIB Hannover: RA 3251(97/E/59) / FIZ - Fachinformationszzentrum Karlsruhe / TIB - Technische InformationsbibliothekSIGLEDEGerman