Combined use of weather radar and limited area model for wintertime precipitation type discrimination

Abstract

Snowstorms affect a variety of human activities such as transportation management in urban areas, highways and airports, commerce, energy and communications (Rasmussen et al. 2003). For these reasons, monitoring snowfalls in real-time with the maximum time-space resolution available is critical, as well as locating areas affected by these phenomena and nowcasting their evolution. In wintertime, the problem is even more evident, especially in regions like the Piedmont, in Northwestern Italy, where the complex orography favors abundant but irregular snowfalls in densely populated areas. The knowledge of the exact location of the rain-snow boundary is also necessary to evaluate precipitation amounts for hydrological purposes. When radar reflectivity measurements are available, the precipitation estimation process implies different Z-R or Z-S relationships, depending on rain-snow precipitation type (Smith 1984). A better identification of dry snow, wet snow and rain brings to more accurate snow water equivalent accumulations that could be also obtained using snow density fields varying in time and space, according to the interpolated field of temperature derived from ground stations (La Chapelle 1961; Hedstrom and Pomeroy 1998). The contour line of zero degree air temperature is one of the most simple and common indicator in order to classify the limit between snow and rain. Nevertheless, to identify operationally wintertime precipitation type at ground is not a simple task: several microphysical processes are involved in precipitation growth and in temperature profile evolution, making the snow-rain boundary strongly influenced by local scale processes. As a matter of fact, the precipitation type depends on lower-tropospheric air temperature and humidity profile, which are affected by horizontal and vertical advection, deep moist convection, vertical mixing/surface fluxes, atmospheric radiation and different latent heating (Olsen 2003). When solid precipitation passes through the freezing level, before reaching the ground, the latent heating generated by the melting, causes a negative tendency in temperature, thus creating, in conditions of weak advection, a zero degree isothermal layer and propagating downward the solid precipitation (Kain et al. 2000). Immediately when precipitation reaches sub-saturated air, evaporation begins. The heat required for transformation from water to vapor, proportional to both intensity of precipitation and relative humidity, is taken from the environmental air, hence causing cooling. Cooling by evaporation is one order greater than the melting one. Relative humidity must be considered if temperature at surface is several degrees above zero when wet-bulb temperature is near freezing (Matsuo and Sasyo 1981). The wet-bulb temperature profile is thus a key factor to find out the precipitation type at ground (Baumgardt 1999). Computation of Mitra et al. (1990) showed that inside clouds of 100% relative humidity and a lapse rate of 0.6 C/100 m, 99% of ice mass of 10 mm snow flake melts within a fall distance of 450 m. This fall distance is about 100 m longer if the relative humidity is only 90%. These results are also consistent with radar observations which show that typically the bright band extends between 0 and 5 C and encompasses several hundred meters (Pruppacher and Klett 1998). Therefore, melting re-freezing and evaporation processes contribute to modify both in time and space the rain-snow boundary, making arduous to identify the precipitation type at ground. A relatively new approach in precipitation type discrimination involves the use of dual-polarization radar data. Several studies have shown the utility of polarimetric radar observables for discriminating hydrometeor particle types (Straka and Zrnic 1993; Ryzhkov and Zrnic 1998). Due to the polarimetric signature overlap for different particle types, the fuzzy logic is the most widely used method to face the problem. Membership functions are usually defined for all available polarimetric observations (Z, Zdr, Kdp, Ldr, ρHV) and for the vertical temperature profile which plays a key role in the classification process. A hydrometeor type is then assigned to each single radar cell. Unfortunately, radar measurements always come from a certain altitude above ground, due to the Earths curvature and to the complex orography, making dubious to assess precipitation type at ground without further assumptions or observations. Moreover, both the beam blockage and the rain attenuation at C-band or X-band can significantly affect differential reflectivity measurements, producing artifacts, whose consequence is a misleading classification of hydrometeor. Several algorithms for discriminating precipitation type are currently available in literature: most of them use observed thermodynamic vertical profiles (Ramer 1993; Baldwin et al. 1994; Bourgouin 2000), others use the average virtual temperature calculated by geopotential heights of two pressure surfaces (Zerr 1997). A complete review of those algorithms can be found in Cortinas and Baldwin (1999). In this study, we make a comparison between three algorithms, aimed at distinguishing between solid (snow, ice), mixed (wet snow, sleet) and liquid (rain) precipitation at ground over the Piedmont. These algorithms are based on reflectivity data, measured by operational Cband polarimetric radar, 2 m air temperature and wet-bulb temperature, derived from ground network observations and limited area numerical model (LAM) short-term forecasts. The algorithms verification is carried out by comparing each algorithms output for several snow events occurred during 2005/2006 winter season with data collected by seven present weather Vaisala FD12P sensors, located in the Po valley. Section 18.2 will provide an explanation of the data sources and a full description of the three algorithms tested for discriminating the precipitation type. In Sect. 18.3, the algorithms verification method will be presented; some case studies and the results are in Sect. 18.4, followed by some concluding remarks in Sect. 18.5.