At ESRL/GSD we have been evaluating the TAMDAR data in a number of ways, including the overall quality of the data, its potential use for meteorological applications such as severe weather forecasting, and its possible impact on short-term numerical weather prediction (NWP) model forecasts. Parallel versions of the Rapid Update Cycle (RUC) model have been running for over two years on a continuous basis, one without TAMDAR data and one with. From these runs we have generated long-term statistics of RUC analyses and forecasts compared to RAOB soundings, using the entire CONUS, an eastern subset of the CONUS, and a smaller subset over the Great Lakes region where most of the TAMDAR is concentrated at this time. Specific case studies using the model forecasts have also been done, often demonstrating a very notable positive impact on the 6-h forecasts of precipitation and other fields. The long-term RUC statistics have generally shown positive results for relative humidity, although at times the improvement is quite small. There are also times when there has been a negative impact from the TAMDAR data for relative humidity, and this study seeks to gain some insight into these negative impact days. In particular, using the statistics, days with a negative impact for TAMDAR on short-term (3 and 6-h) forecasts valid at 0000 UTC are isolated, and individual forecast soundings from the two RUC runs are compared to the observed RAOB. Using this methodology, we hope to determine whether the negative impacts seen in the statistics are true poor forecasts, or if they may be the result of issues such as an unrepresentative sounding, very sharp moisture discontinuities in the vertical, or other issues. Suggestions for improving how we measure forecast accuracy against the RAOBs will be made where appropriate, hopefully leading to a better overall statistical measure of the impact of TAMDAR on NWP.