The tool C2W (climate-to-weather disaggregator) is introduced which is
aimed at disaggregating climatological means and anomalies into reali
stic weather processes. For the core variables of minimum and maximum
temperature and precipitation, a probit normalization is conducted whi
ch transforms each quantity into one which is normally distributed wit
h mean 0 and standard deviation 1. In this way, the total spatial and
seasonal spectrum of climatological variability is filtered out, leavi
ng a 3-dimensional process of 'normalized weather'. The climatology is
contained in the set of parameters which define the probit function.
In a second step, a first order autoregressive model is fitted to the
normalized weather process which is then, by construction, globally ap
plicable for any time of the year. The determination of probit paramet
ers is achieved, roughly, by a parameterization of climate variability
in terms of climate mean. By way of Monte-Carlo simulations, a univer
sal map can be defined which transmits, in a 1-1 way, information betw
een the climate mean and the probit function in such a way that the me
ans of the simulated weather converge statistically, if simulated long
enough, to the given climate mean. The simulated variability, however
, is generally not preserved. Depending on the specific region and tim
e of the year, C2W exhibits deviations from observed variability, with
errors increasing in extreme climates; for most temperate climates, t
he simulated variability is comparable to the observed. The disaggrega
tion of climatological means is then extended to include Various aggre
gations, such as monthly or seasonal means, by interpreting them as an
omalies from the mean. Besides being a simple and handy weather genera
tor, C2W is best applied as a postprocessing scheme for gridded data s
ets, such as those from General Circulation Models (GCMs) or gridded c
limate maps. In that way, C2W works as a simple link between GCMs and,
for example, dynamic global vegetation models, C2W is available as a
Fortran program module.