Questions Continued

Jimmie C. says - Awesome plots and thanks for highlighting this issue in the Plains as well. Is the GPS data available online? I saw the GSD website has images but what about actual data files for select stations?

I ask because I wonder how much balloon drift may effect the RRS PW. It is clear from the GPS data that there are rapid small scale changes occurring and I wonder if that has a big enough impact to cause the differences you are seeing. I suspect one could correlate the differences in PW to distance the balloon traveled.

An off the cuff thought was to see how inversion strength or lapse rates might affect the GPS measurement. I certianly don't think the RRS moisture sensor is error free, but the randomness might be explained by multiple factors and then you get at the systematic issues.

An interesting mystery and one worthy of exploring.
---------------------------------------------------------------------
Most of these questions/issues relate to the differences in, and the accuracy of, the two observing systems.

I consider the go-to-expert here to be Seth Gutman at the NOAA labs in Boulder.

seth.i.gutman@noaa.gov

Since the GPS calculations are using data from a number of different GPS satellites, and thereby incorporating a number of "lines of sight", I suspect that the GPS observations may be more "representative" than are the RRS (or any sounding instrument) data taken along a trajectory determined by the local winds and rate of balloon rise.

I do suspect that the first order problems with the RRS PW values relate to sensor problems of several types. Since most of the PW is present in the early part of the flight, the 4-D trajectory may contribute much less to inaccuracies, except in the case of extreme horizontal gradients.
--------------------------------------------------
Several interesting issues relate to your questions. It should be straight forward to examine the data observed in any sounding with respect to: where were the data actually taken in time and space. Of course, it is much easier to assume (falsely, of course) that the observed data are positioned directly above the release point. At some point in the fairly distant past such an assumption was easily justified - however, it is much harder to justify such an assumption given the computing power available today. I am constantly amazed that someone, or some organization, doesn't provide upper-level plots with the data properly located in space. Or, maybe someone does? The time issue is a bit tougher. I suppose that model initialization procedures could (or do?) use the observed data exactly when and where the observation occurred.
-------------------------------------------------
Another issue involves the standardized observing times for upper-air data. The reasons for taking data globally at 00 and 12 UTC were/are quite sound. However, given that 4-D data assimilation seems to drive things today, it would appear that all upper-air data might not need to be taken so rigidly wrt time. A local Arizona example - consider the NWS soundings taken near Flagstaff in the summertime. At 12 UTC in the morning the sounding typically samples an extreme surface-based temperature inversion, often making evaluation of convective potential difficult. Convective storms up in the Flagstaff area tend to occur between 11 am to 3 pm or so in the afternoon. Thus, the 00 UTC sounding is often sampling a local environment that has been contaminated by storms. The obvious question is: why not allow the Flagstaff sounding to be flown at times more in sync with forecasting local convection?