A quality control (QC) procedure is applied to the Challenging Minisatellite Payload (CHAMP) level-2 Global Position System (GPS) radio occultation (RO) data provided by the University Corporation for Atmospheric Research (UCAR) Constellation Observing System for Meteorology, Ionosphere and Climate (COSMIC) Data Analysis and Archival Center (CDAAC). It consists of a range check removing negative values, a biweight check removing data which deviate from the biweight mean by more than four times the biweight standard deviation, another biweight check removing data with large deviations to the National Center for Environmental Prediction (NCEP) analysis, and finally a symmetric check removing the negative bias of GPS observations in the low troposphere below 4 km. These four QC checks are applied sequentially to identify outliers in GPS bending angle and refractivity data at each vertical level using data primarily in the month of March 2004. Having removed 5.5% outlier data, the GPS RO observations compared much more favorably with the NCEP analyses than the original data without QC, resulting in an improved spatial consistency, a more symmetric probability distribution, significantly reduced error variances, and a nearly diagonal vertical error correlation matrix. The effectiveness of the proposed QC procedure is further confirmed by showing that most RO profiles removed by the QC procedure proposed in this study have those characteristic physical parameter values indicating poor quality of RO data. Similar results are obtained when the same QC procedure is applied to CHAMP data in July 2002, which shows the robustness of the proposed GPS QC procedure.