6B.2 Determining Precipitation Type from Maximum Temperature in the Lower Atmosphere

Tuesday, 30 June 2015: 10:45 AM
Salon A-5 (Hilton Chicago)
Marc Chenard, NOAA/NWSFO, Sioux Falls, SD; and P. N. Schumacher and H. D. Reeves

This is the first of two novel techniques for discriminating the surface precipitation type that will be presented at this meeting. This product yields the probability of three different types of precipitation: snow, (SN), sleet (PL) and freezing rain/rain (FZRA/RA). The probability distributions are based on a set of over 1700 soundings associated with precipitation types at the surface. Several potential metrics to discriminate between the different forms were considered, but the most meaningful was the maximum temperature between the surface and 600 hPa (Tmax). For SN, the probabilities are greatest for Tmax < 0⁰C and are less than 10% for Tmax = 1⁰C. The probability of PL increases rapidly once Tmax > 0⁰C, reaching a peak between 0.5⁰ and 1⁰C. While rare, PL can occur when Tmax > 5⁰C. FZRA/RA also occur for any Tmax above 0⁰C, but are the most likely precipitation types for Tmax > 2⁰C. The effects of model uncertainty on this technique were also considered by perturbing the observed soundings in accordance with the uncertainty ranges typical for a mesoscale modeling system. As one would expect, the probability distributions broaden when the uncertainty effects are accounted for. This has only marginal effects on the efficacy of the system to provide meaningful probabilities for SN, however PL and FZRA/RA are more strongly impacted. Both have nonzero probabilities for Tmax as low as 2⁰C, with PL the most likely precipitation type from Tmax of -0.5 to 4.5⁰C. The longer the lead time considered, the broader the probability distributions, especially for PL which has probabilities as high as 10% at Tmax = 10⁰C (for the range of uncertainty typical for a 12-h lead time). These results highlight the great difficulty in providing an accurate prediction of precipitation type using deterministic methods, even for short-range forecasts, and justify the need for more probabilistic approaches and the inclusion of more information in the algorithm (such as microphysical controls, as discussed in the second talk).
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner