ABSTRACT In this study, a comprehensive quality control (QC) system for in-situ precipitation data records was developed and applied to Canadian in situ precipitation datasets. The system includes a pair of screening procedures to screen for two types of random errors: one procedure is applied to the untransformed monthly total precipitation data series, which is good at finding erroneous data of unusually large values; another is applied to the log-transformed monthly precipitation data (in mm) series, log(P + 0.1), which is good at identifying erroneous zero or near-zero monthly total precipitation amounts. The system then applies three QC (threshold, kriging, and temporal) tests and a decision-making process to confirm whether the screened suspects are erroneous. There is generally good agreement between all the QC tests, while the decision-making process yields the most accurate results when compared to the manually reviewed results. The QC work on Canadian precipitation data sets revealed that it is necessary to apply a pair of screening procedures to identify both types of random errors. All the monthly values identified to be erroneous are set to missing, and so are the corresponding daily values, while keeping records of the original data.