Vertical cross-sections of a) reflectivity factor and b) result of hydrometeor classification for the June 6, 1996 storm in Oklahoma. The class designation is as follows: LR - light rain < 3 mm/h, MR - moderate rain between 5 and 30 mm/h, HR - heavy rain > 30 mm/h, LD - large drops, R/H - rain /hail mixture, GSH - graupel small hail, HA - hail aloft, DS - dry snow, WS - wet snow, IH - ice crystals horizontally oriented, IV - ice crystals vertically oriented.

Exposing the multiple personalities of precipitation

by Dusan Zrnic and Alexander Ryzhkov

Wind, temperature and precipitation are the weather conditions that affect our daily activities. The "Jekyll and Hyde" of these is precipitation. Some precipitation is indispensable, too much is disastrous. And because it reaches ground in a variety of forms, it produces effects from beneficial (needed rain or snow) to detrimental (hail and floods). Remote classification and measurement of precipitation was an evasive goal of radar meteorologists until the advent of polarimetric radar.

Polarimetry entails probing of precipitation with orthogonal electric fields. For example, horizontally and vertically polarized electric fields interact differently with contrasting hydrometeors, such as rain or hail, to produce tell-tale signatures in the fields of polarimetric variables. Several polarimetric variables can be measured; among the useful ones are differential reflectivity, differential phase, correlation between the orthogonally polarized returns, and linear depolarization ratio.

NSSL began accumulating a gamut of polarimetric radar data in 1992. Soon thereafter, in collaboration with the University of Oklahoma, we began to develop an automatic procedure for discrimination of hydrometeors. Within the past year, scientists from National Center for Atmospheric Research (NCAR) have joined this venture. With partial sponsorship from the Federal Aviation Administration (FAA), the automatic procedure has further evolved into a real-time algorithm that was implemented on the NCARŐs S-Pol radar, and testing began last summer in Florida.

For a given precipitation type, the polarimetric variables cluster in a specific region. The crux of the classification process is to separate overlapping clusters so that the probability of correct classification is high, while the probability of mis-classification is low. Classification is based upon weights assigned to the various multiparameter variables. The choice of weights is founded on previous measurements, physical reasoning, modeling, and sometimes gut feeling.

The example in the figure at the top of the page is well-suited to demonstrate the potential of polarimetry for the classification of hydrometeors. This, we believe, is a fundamental condition for accurate determination of precipitation amounts. First, a correct classification needs to be made, and then, appropriate semi-empirical relations should be applied to each class to estimate the corresponding amounts. This is quite different from the current practice (with reflectivity in the operational world), whereby the choice is between a few relations, and the operators decide if precipitation is frozen or liquid.

The results presented here and elsewhere are very promising, yet much testing and comparisons with in-situ measurements are required to evolve the algorithm into a useful tool.

For more information, contact Dusan Zrnic at: zrnic@nssl.noaa.gov

For a given precipitation type, the polarimetric variables cluster in a specific region