Rather than take only the first 6 hours of each SCM run, all 24 hours are used to compute the
SCM results. Thus, the SCM value at a particular hour, say 2100, is an average of that value
from four runs:
Run 1 that starts at 0000 hours
Run 2 that starts at 0600 hours
Run 3 that starts at 1200 hours
Run 4 that starts at 1800 hours
This format is used for two reasons:
1) To prevent the masking of some model errors that are strongly dependent on the diurnal cycle.
As and example, assume the SCM begins to overestimate the cloud amount and that this
error increases as the run progresses. If we used a single 24-hour SCM run that started at
0000GMT, the effect on the solar radiative fluxes would be
much different at a site where 0000GMT is local midnight compared to another site where 0000GMT
is local noon.
2) Using a run length of 24 hours generally allows for deficiencies in most model parameterizations
to become evident. Of course, using a longer run length would increase the chances for parameterization
deficiencies to show up, but the longer run would also increase the chances for unrealistic model
temperature and humidities that would also influence the performance of the model parameterizations. The
choice of a 24-hour run is a compromise between these two competing factors.