The Atmospheric InfraRed Sounder (AIRS) retrievals, the Global Positioning System (GPS) surface based Integrated Precipitable Water (IPW) and the radiosonde water vapor have been used in a three way comparison. Validation of AIRS data requires a accurate ground truth that is readily available. The radiosonde reports we are using are the ones delivered to the forecast centers, and specifically NCEP. Many procedures have been developed for the raw radiosonde reports, but these reports are not widely available in real time. As part of our validation, we have been able to improve the accuracy of the radiosonde water vapor estimates. The first and most obvious use of a three-way match is to identify individual problem observations. But the most useful result is a procedure to adjust the radiosonde results to remove calibration biases. Work at the ARM sites has resulted in several radiosonde correction procedures. One of these uses the unbiased, upward looking microwave IPW to adjust individual radiosonde observations. This can’t be widely applied because few sites have the microwave equipment. But the GPS based IPW has many of the same characteristics and is based on inexpensive equipment that is becoming widely deployed. Our approach has been to take radiosonde stations where a GPS is located at or close by, and use the GPS value to obtain a ratio by comparing the GPS IPW with the radiosnde IPW. This ratio is then used to adjust the radiosonde based layer IPW values which are then compared to the AIRS values. Since the AIRS are independent, the AIRS is used to validate the procedure, then the adjusted values are used for the AIRS validation. Significant improvements in the agreement between AIRS and the radiosonde are obtained, although the amount depends on the radiosonde type. The improvements are greatest in the lower levels where most of the water vapor is located. Although this procedure can adjust for a radiosonde bias, there are other corrections that affect the shape of the vertical profile. The bias adjustment can actually make things slightly worse at the upper levels because many radiosondes with a dry bias at the lower levels actually have a wet bias at the upper levels. We are starting to apply these shape dependent corrections to the operational radiosonde reports as well to resolve the upper levels and see additional improvements in our initial results. Our conclusions are that these procedures have a substantial impact on the ability to assess the true AIRS accuracy. We have also demonstrated that these procedures should be applied to operational reports used for other purposes such as numerical weather forecasting. Given the low cost of the GPS equipments, we recommend that they be deployed at operational radiosonde sites. Of course we will show the resulting AIRS accuracy which agrees with the GPS and radiosonde IPW values to about the same accuracy as they agree with each other.