An efficient detection of outliers and a good knowledge of the accuracy of GPS ZTD estimates are important for optimal use of these data in meteorology and climatology. The accuracy of GPS ZTD estimates depends on many factors such as the reliability of the equipment, the number of observations, the magnitude of noise and interferences in the measurements, the quality of the troposphere model used for the data processing. As these factors may change over time, an adaptive screening and quality control is required. Here we propose a method to detect and eliminate ZTD estimates that are less accurate and/or may be classified as outliers. The method depends on parameters available at the end of the data processing such as the number of fixed ambiguities, the position and ZTD estimates and their formal errors. It is therefore totally independent of external data. This method is first applied to the HyMeX post-processed GPS dataset produced with GIPSY-OASIS II 6.2 software. It is evaluated against a method with uses an atmospheric model analysis as a reference. The method is intended to be generalized to GPS solutions produced with other GPS software and especially to IGS and EUREF products. The methods can also be applied to other space geodetic techniques like DORIS and VLBI.