|
Mike,
The analysis file contains an estimate of the current conditions based on both current observations and a previous forecast (1h, in the case of the RUC) that itself contains information from previous observations.
So, this is an estimation problem. We don't have observations at every grid point, and even if we did, the observations themselves have errors. So, the process of 'data assimilation' produces that gridded analysis that you are looking at. It takes into account the expected error of both the observations and the forecast background field. It also takes into account the expected spatial correlations in forecast error fields, based on previous statistics. It ends being a sophisticated interpolation or minimization problem taking into account all of these and other factors to produce the 'analysis grids'.
There is more information on data assimilation at this web site:
http://meted.ucar.edu/nwp/pcu1/ic6/frameset.htm
So, the short answer is that the analysis file is based on observations but should not be considered equivalent to the observations themselves.
Stan
|