A new quality control system is presented which combines climate mapping methods and probability statistics into a single, powerful technology. Traditional quality control systems are deterministic. They typically employ a series of quality checks that an observation must pass if it is to be considered valid. The outcomes of these checks are of a "yes" or "no" nature, and the observation is either tossed or kept based on these outcomes. However, more and more climate observations are made electronically, where the observations are subject to a host of continuous errors, such as instrument drift. In addition, there is a need for quantitative confidence estimates for both the observation and an estimated observation, if used. This calls for a probabilistic system that estimates a continuous quantitative confidence probability for each observation and estimated value. The first generation of such a system as developed for USDA-NRCS Snotel temperature data is presented. Issues unique to the use of spatial methods in data QC are discussed, such as the problem of using station data of imperfect quality to assess the quality of other nearby station data.