Missing Winds from RRS Soundings - An Extreme Example

On Monday afternoon, August 13, 2007, Rob Dale and Eric Hanson alerted me to the strange characteristics of the special 1800 UTC sounding from MPX (Chanhassen, MN, an RRS/Sippican sonde site).

The unbelievably smooth hodograph is what had caught their attention. The sounding, as analyzed by software analysis programs at the Storm Prediction Center (SPC) is shown in Fig. 1. Note the perfectly smooth hodograph and the winds that back continuously from the surface to 150 mb. This seems a very strange sounding indeed!

However, examination of the sounding at the FSL RAOB site indicated what the problem was. The MPX sounding at FSL (Fig. 2) indicates that the Sippican sonde apparently lost GPS tracking immediately off the surface and did not recover the signals until 150 mb - i.e., the winds were actually missing from the surface to 150 mb.

So where did the "data" plotted at SPC come from? It turns out that many sounding analysis programs use, as part of their code, GEMPAK. The GEMPAK procedure interpolates data into layers where observations are missing - assuming probably that such layers are never very deep.

The sounding plot from the Univ. of Wyoming site (Fig. 3) shows interpolated winds that are similar to those on the SPC plot. However, Lance Bosart alerted me to look at the same sounding at the UCAR/RAP upper-air site (Fig. 4). Their plot shows a completely different wind profile through the troposphere! Apparently, RAP software uses a different interpolation procedure. The winds in their "answer" for the deep layer of missing observations essentially keep the direction at 150 mb constant, while decreasing the speed smoothly until near the surface, where a shallow layer of backing winds is shown.

While I'm not sure what impact this comedy of errors might have had on forecasts during the afternoon of August 13th, it does vividly illustrate how serious the Sippican missing winds problem can be. I have been alerted by a number of people around the country about the missing winds problem, and I will document it further at a later time. However, this is a particularly insidious problem for non-NWS users of the data. If one suspects that a sounding may have missing winds, then a time-consuming examination of the sounding at different sites has to be made for accurate evaluation.