Re: Several Questions

From John D. - Bob, I have read your posts about the dry bias in this eqipment for the past two Monsoon seasons here in AZ. I know the impact bad sounding data can have on modeling the daily forecast during the monsoon. Has this bad data had a negative impact on the GFS model itself on a much larger scale?
------------------------------------------------------------------------
I don't know if the various problems with data from the RRS sondes are impacting the GFS and NAM forecasts. I would think that folks at NCEF would be evaluating these issues. At times it is obvious that bad data have gotten into the standard level analyses and you'd think that, in such situations, there would be some impact on the forecasts. Remember that the NWS official position seems to be that the data problems with RRS sondes do not impact their forecasts. However, I know that short-term forecasts of convection can be adversely affected, and there are a number of posts here over the last two summers illustrating such situations.

At U of A Atmospheric Sciences, Mike Leuthold runs a high-resolution version of the WRF model routinely. He has found that, especially in the warm season, the forecasts of convection within the model are strongly influenced by the initial precipitable water field. Since these are often inaccurate wrt to GPS precipitable water observations, he routinely uses GPS PW to "correct" the inital analyses of several NWS models.
----------------------------------------------------------------------
This winter alone it seemed as if ECMWF has consistently outperformed the GFS.(Maybe this is just my perception) I would think bad data going into a complex mathematical formula would exponetially increase errors forward in time..especially if the bad data was so random. On a side note, any thoughts on why the EC model did so well this past winter vs the GFS?
---------------------------------------------------------------------
I'm not a modeler, but even a fairly casual user of models has likely noticed that the ECMWF usually does better than the NWS GFS model (not always though, which keeps things interesting). I suspect that there are at least two things at play. The ECMWF operational time lines are not as stringent as those of NWS, and they spend more time and effort in the initialization process. I also think that the resolution of the ECMWF is superior to that of the GFS. Many folks on the Albany MAPS list could probably provide you a much better answer than I can.