A quantitative comparison of automated and manual precipitation gage quality control techniques

Whereas automated quality control (QC) methods must rely on statistical measures and comparisons (e.g., neighbor checks), manual procedures often bring knowledge of physical causes for gage measurement failures (clogged or dripping gages, for example).This knowledge generally produces better judgments. On the other hand, quantitative techniques can reduce human errors and inconsistencies. We pursue an assessment of the types and magnitudes of the differences that can arise between different QC methodologies by comparing the statistics of each during four months of the summer of 2008 over the Lower Mississippi River Forecast Center (LMRFC) domain. The manual QC was performed and tabulated by forecast personnel at the LMRFC, and automated techniques were developed at the Global Systems Division of the Earth System Research Laboratory, and at the National Climatic Data Center. In addition to straight comparison of flagged station lists for each, the error types captured by the separate schemes are illustrated and examined, the impact of the differences in station selection are estimated by appropriate metrics (including neighbor comparisons), and a qualitative assessment of the overall performance of the automated techniques is made by direct comparison with independent radar observations accessed from the Q2 precipitation product of the National Severe Storm Laboratory. The implications of these differences vis-à-vis representativeness errors are discussed.