Assessing the accuracy and precision of radiosonde data has historically been a challenge due to changes in instrumentation, temporal and spatial differences in datasets, and a general lack in transfer standards. This paper will examine the quality of 10-year (1996-2005) operational radiosonde data from two U.S. National Weather Service (NWS) stations (Norman, Oklahoma and Barrow, Alaska) by comparing them with research-quality radiosonde and microwave radiometer data collected by the Atmospheric Radiation Measurement (ARM) program located in Purcell, Oklahoma and Barrow, Alaska. For the radiosonde comparisons in Oklahoma, four different types of radiosonde, VIZ-B , VIZ-B2, RS80-H and RS90 were launched by the two sites between 1996 and 2002. These soundings were matched into 490 pairs, launched within a half-hour of each other. The comparisons show both known and unknown errors in radiosonde data. Additionally, comparisons of Vaisala RS80-H and RS90 data consistently show unexplained, significantly drier (~5% in RH) RS90 data in the UT. Similar comparisons will be made between temporally matched radiosondes launched at the neighboring operational NWS site and ARM research facility near Barrow, Alaska. These matched pairs will also be compared with microwave radiometer data in order to gauge the performance of the radiosonde sensors. The Barrow sites were chosen because of their proximity, the availability of long term datasets spanning eight years (1998-2005), and the climate in Barrow allows for the comparison of different radiosonde types in a cold environment, where sensors typically do not perform as well.