Now here's a first peek inside the analysis. Overall, the analysis dewpoints are much closer to the radiosonde dewpoints than were those of the first guess, and the analysis temperatures also are in slightly better agreement with the radiosonde observations except in the 750 to 550 hPa layer. This layer in the analysis is striking both in its deviation from an apparently good first guess and as a saturated layer with a dry-adiabatic lapse rate.
The analysis can make a dry-adiabatic saturated layer because temperature and moisture are analyzed independently. Temperature and moisture are of course physically linked in the first guess, since it comes from a 3-hour forecast during which all the model physics are integrated. And, since the July 2001 model upgrade, observed precipitation estimates are used to nudge the moisture fields and diabatic heating rates in the EDAS cycle which produces that first guess. But when the analysis makes changes to the first guess, no physics (condensation, microphysics, radiation, turbulence, friction) are involved.