The Uncertainty of Precipitation-Type Observations and Its Effect on the Validation of Forecast Precipitation Type

Abstract Herein, an evaluation of the uncertainty of precipitation-type observations and its effect on the validation of forecast precipitation type is undertaken. The forms of uncertainty are instrument/observer bias and horizontal/temporal variability. Instrument/observer biases are assessed by comparing observations from the Automated Surface Observing Station (ASOS) and Meteorological Phenomena Identification Near the Ground (mPING) networks. Relative to the augmented ASOS, mPING observations are biased toward ice pellets (PL) and away from rain (RA). However, when mPING is used to validate precipitation-type algorithms, the probabilities of detection (PODs) for both RA and PL are decreased relative to those from the augmented ASOS. The decreased POD for RA is the result of numerous mPING reports of RA in the presence of a surface-subfreezing layer in the nearest observed sounding. Temporal and spatial variability effects are also assessed. The typical lifespan of transitional forms of precipitation is between 10 and 40 min, with many events having two or more forms of precipitation reported in a 1-h time frame. Depending on how one defines a hit for these rapidly evolving events, inherent biases in the forecasts may be dampened or masked altogether. Spatial variability also exerts a strong control on the performance of postprocessing algorithms, as both FZRA and PL often have spatial scales that are too small to be resolved, even by convection-allowing forecast models. However, the degree of variability is not strongly dependent on the distance separating any two observation pairs and, consequently, validation statistics do not change significantly as a model’s grid spacing is increased, all else being equal.