From 1997 to 1999 the National Weather Service (NWS) evaluated an acoustic sensor as a means of detecting ice pellets and hail on ASOS (Automated Surface Observing System). The acoustic sensor consisted of a microphone located beneath an inverted metal bowl approximately 20 cm in diameter. The bowl was heated to keep snow and ice from accumulating on its surface. The acoustic sensor was interfaced to an enhanced version of the Light-Emitting Diode Weather Identifier (LEDWI), manufactured by ScTI (Scientific Technology, Inc.), now Optical Scientific, Inc.

The theory behind the acoustic sensor was that ice pellets and hail would be detectable by the increased sound that they would make compared to rain and snow. The NWS evaluated the sensor over 2 winters at their research facilities at Sterling, VA and Johnstown, PA, and concluded that the enhanced LEDWI (ELEDWI) with the acoustic sensor "showed no significant improvement" in precipitation identification over the LEDWI sensor that is currently being used operationally at nearly 900 ASOS stations. Thus, the acoustic sensor is no longer being considered as an upgrade to ASOS stations, and the NWS is continuing to search for an enhanced precipitation identification sensor that is capable of detecting ice pellets and hail.

A reanalysis of the data collected during this period has shown that the acoustic sensor, in fact, works quite well at detecting ice pellets and hail, and can even distinguish snow pellets from snow. The statistics provided by the NWS in their reports summarizing the performance of the enhanced LEDWIs show that the 8 sensors tested correctly identified ice pellets 82% of the time on an average, with one sensor correctly identifying ice pellets 96% of the time. While these numbers are below the 97% acceptance criteria established by the NWS, they are considerably better than the 0% identified by the current LEDWIs. In fact the ELEDWI ice pellet identification is sufficiently good that one might wonder about the circumstances surrounding the misidentifications that occurred. Might a slight modification to the identification algorithm improve the statistics? Were the human observations of precipitation type correct every minute? How does one handle the problem whereby only one precipitation type can be reported when mixed types are occurring? For example, if only 20% of the precipitation is in the form of ice pellets (IP) and 80% is freezing rain (ZR), the observer might report mixed ZR and IP, but the ELEDWI algorithm might report rain. Is this a case of erroneous identification?

This paper will discuss these issues using examples from ELEDWI data collected at Sterling and Johnstown during 1997-1998. The paper will argue that, with minor modifications to the precipitation identification algorithm, the addition of an acoustic sensor will allow ASOS to report ice pellets, snow pellets, and hail with an accuracy equal to or better than the human observer.