Yves Rochon
In the absence of 6-hour forecast error statistics for the initial data
assimilation experiements with CMAM, statistics used were derived from
the output of a model run. More precisely, the derived statistics were
determined from differences of the fields between successive 6-hour
output. These statistics were applied in the framework of the MSC 3D-var
assimilation system.
A brief review of the assimilation variables and the representation of
the background statistics in the 3D-var system will be provided.
Statistics derived from the 6-hour difference fields will be shown and
compared with those based on the lagged-forecasts scheme currently used
by the operational system and on the ensemble-prediction scheme being
investigated at MSC. The resulting OmF and OmP standard deviations and
biases will be provided in comparison to those of the operational system
for January 2002.