We used 14 years of data from 12 GPS sites in Sweden and Finland to estimate trends in the atmospheric integrated water vapor (IWV) for 8 different elevation cutoff angles, from 5° to 40°, for the observations used in the analyses. These trends were compared to the corresponding trends obtained from radiosonde data at 7 nearby (<;120 km) sites. The results show a variation in the correlation of the trends between the two techniques for different elevation cutoff angles. The highest correlation coefficient of 0.88 is obtained for the 25^ solution, whereas the smallest root-mean-square (RMS) differences between the IWV estimates themselves are obtained mainly for elevation cutoff angles of 10° and 15°. The results show that elevation-angle-dependent systematic errors vary with time. Therefore the elevation cutoff angle giving the best agreement between radiosonde and GPS for individual IWV estimates is not necessarily the optimum when estimating linear trends. The correlation between the trends from the two completely independent techniques is strong evidence that the two techniques provide information on the IWV trends although the true individual values are too small to be uniquely detected. In addition, we found that the choice of mapping functions is not critical for the IWV trend estimation.