1. Introduction

The radar melting-layer bright band has fascinated radar meteorologists for more than half a century. Early studies (e.g., Austin and Bemis 1950) explained the fundamental aspects of the phenomenon, but its details are still examined to reveal clues about microphysical processes and finescale dynamics of the transformation of snow to rain in and near this layer (e.g., Drummond et al. 1996; Szyrmer and Zawadzki 1999). Perhaps the most comprehensive radar brightband study was conducted by Fabry and Zawadzki (1995), using over 800 h of high-resolution observations from vertically pointing radars in Montreal, Quebec, Canada.

The bright band also represents a difficult challenge for estimating rainfall rates by radar because its enhanced reflectivities give an exaggerated impression of the rainfall rates occurring beneath it at the ground. Its presence provides definitive evidence of solid (ice) hydrometeors above its altitude and liquid below. Thus, it has recently begun serving as a convenient profiling radar marker for automated determination of the snow–rain transition altitude, knowledge of which is important to numerous applications, including forecasting runoff from winter storms in mountainous terrain (White et al. 2002). Brandes and Ikeda (2004) developed an objective algorithm for determining brightband presence and height from scanning polarimetric radar data.

In stratiform precipitation, the bright band appears as a nearly horizontal narrow layer of sharply enhanced reflectivity on time–height displays of profiling radars or on range–height indicator (RHI) displays of radar elevation scans. Reflectivity usually peaks about 200–400 m below the 0°C isotherm (e.g., Stewart et al. 1984). Corresponding vertical profiles of Doppler velocity reveal an abrupt increase in downward speeds and broadened spectra below the 0°C isotherm, where slowly falling snowflakes melt and become rapidly falling raindrops. In the plan position indicator (PPI) displays of tilted, azimuth scans made by storm surveillance radars, the bright band appears as a circle of enhanced reflectivity centered on the origin at the range corresponding to the melting-layer altitude.

Usually the temperature profile aloft contains a single 0°C level, which produces a single bright band as snow falls through it. As melting proceeds, the cooling of air produced by the diabatic melting of ice produces an isothermal 0°C melting layer that expands downward and may affect airflow kinematics (e.g., Atlas et al. 1969). On occasion, however, the profile is more complicated, such as when warm air arrives aloft before it reaches the surface with an approaching warm front. In these situations there may be two (or more) altitudes of 0°C, with a corresponding bright band beneath each. Fabry and Zawadzki (1995) observed double bright bands 1.4% of the time in their hundreds of hours of radar profiles in stratiform rain. Fabry et al. (1992) showed radar profiles for one double-brightband case, along with “probable” temperature profiles to explain their cause. Ikeda et al. (2005) showed the evolution of a double bright band in Oregon, and conducted a detailed analysis and interpretation of polarimetric parameter measurements from their radar’s RHI scans to infer particle types aloft above, below, and between the bright bands.

In this article we present observations of another double-brightband case, observed with a scanning polarimetric radar. Unlike earlier double-brightband studies, however, the thermal structure of the atmosphere in this case was more precisely documented by serial radiosonde data that were coincident in location and time with the radar measurements. Therefore, we limit our attention to the connection between temperature profiles and the radar reflectivity and vertical Doppler velocity features to illustrate the obvious cause-and-effect relationship. Data from nearby profiling radar systems, which also observed features of the double bright band, are presented to provide additional storm context and to test the profiler melting-layer algorithm of White et al. (2002) in this uncommon situation. The multiple radar observations of the double bright band are another unique aspect of this case study.

2. Instrumentation

The double-brightband observations were obtained on 25 February 2004 in Sonoma County, California, about 100–120 km northwest of San Francisco, as part of the Hydrometeorology Test bed (HMT) project of the National Oceanic and Atmospheric Administration (NOAA). Figure 1 shows the locations of instruments used in this study. These include a 9.34-GHz (X band) polarimetric scanning radar (described by Martner et al. 2001) on the coastline at Fort Ross (FRS), a GPS radiosonde system, also at FRS, a 915-MHz (UHF) wind profiling radar (Carter et al. 1995) on the coastline at Bodega Bay (BBY), and a pair of 2875-GHz (S band) precipitation profiling radars (White et al. 2000) at BBY and in the coastal mountains at 475-m altitude at Cazadero (CZD). Surface meteorological stations, including rain gauges, were located at all three sites.

The X-band radar conducted a repetitive sequence of PPI and RHI volume scans for 19 h at FRS as the landfalling storm approached and crossed the coast. The NOAA crew launched a series of radiosondes at FRS at approximately 2-h intervals, starting at about the same time the X-band radar began observing the double bright band and continuing well afterward. The profiling radars recorded data autonomously and continuously. The wind profiler at BBY (26 km southeast of FRS) continuously provided hourly averaged vertical profiles of horizontal wind velocity between ∼0.1- and 4.0-km height above ground level (AGL) with 100-m height resolution in clear, cloudy, and precipitating conditions. It also recorded signal-to-noise ratio (SNR) and radial velocity data from its vertical and oblique beams with 6-min temporal resolution. The S-band profilers (S-PROF) at CZD (11 km north of FRS) and at BBY measured vertical profiles of reflectivity and vertical radial velocity in precipitation and cloud with 60-m gate spacing and 8.5-km maximum range every ∼30 s.

3. Meteorological conditions

On 25 February 2004, a landfalling storm impacted northern California. A warm-frontal occlusion and trailing cold front pivoted around a strong extratropical cyclone, while the cyclone’s ∼980-mb center remained offshore of Oregon. Storm totals of 37 and 38 mm of rain fell at the coast at FRS and BBY, respectively, while the orographically favored mountain site at CZD received 91 mm. Precipitation was more intense farther south, where the storm produced flash flooding in San Francisco (Blier et al. 2005) and deluged portions of the Santa Lucia Mountains with as much as 250 mm of rain. The double bright band in Sonoma County, however, was observed in the leading portion of the storm as it first came onshore (approximately 0600–0900 UTC), during which time only 0.76 mm of rain fell at FRS, and 5.33 mm fell at CZD. After about 0930 UTC the rain intensified significantly at all three sites and maintained a nearly steady rate of 14 mm h⁻¹ over the next 5 h at CZD before tapering off.

The continuous data from the wind profiler at BBY (Fig. 2) reveal the progression of conditions aloft for the entire 24-h storm period. The profiler’s hourly averaged wind data clearly depict the passage of a warm-frontal occlusion. Specifically, a warm front and its associated layer of concentrated zonal vertical wind shear descended from 3.8 km MSL at 0600 UTC to the surface at 1500 UTC (Fig. 2a). This layer contained an axis of maximum clockwise turning of the wind vector with height from southerly to southwesterly, indicative of focused warm advection in a geostrophic framework (e.g., Neiman and Shapiro 1989). The occluded cold front aloft is more easily discerned in the meridional isotach analysis (Fig. 2b). The front was preceded immediately by a strong (>35 m s⁻¹) low-level jet centered at ∼800 m MSL, and it was followed by an ascending layer of enhanced meridional vertical wind shear extending upward from the warm-frontal surface at 1430 UTC to ∼4 km MSL at 2000 UTC. Modest thermal wind-derived cold advection was maximized in the frontal shear zone in the slowly weakening southwesterly flow.

Heavy black dots in Fig. 2 indicate the altitude of the bright band’s peak reflectivity within the melting layer determined by the algorithm of White et al. (2002), which is routinely applied to the HMT profiler data in real time. The melting level (0°C isotherm height) is typically a few hundred meters above this height. The accuracy of this height determination is approximately ±100 m. The bright band attained its highest altitude on this day in the warm air immediately ahead of the cold front at 1400 UTC and then descended 330 m during cold-frontal passage. Several hours earlier, the descending warm front caused an abrupt 1.1-km rise in the melting level between 0730 and 0830 UTC. Although the realtime algorithm is not designed to identify more than one melting-layer brightband signature height in each profile, the time surrounding the abrupt rise coincides with the time the double bright band was observed by the X-band radar at FRS.

4. Double-brightband and temperature profile observations

The double-brightband structure was most clearly seen in the X-band radar data at FRS, where the radiosondes were also launched. Examples of the X-band radar’s images of effective radar reflectivity factor (hereafter, reflectivity) are shown in Fig. 3. An RHI scan toward the south at 0830 UTC is shown in Fig. 3a. The upper bright band at ∼2 km AGL extends to at least 15 km in range from the radar. The lower bright band near 1.2 km AGL is only discernible to about a range of 5 km. The bright bands appear as two distinct, concentric circles in the PPI scan of 0752 UTC at an elevation angle of 19° shown in Fig. 3b. The upper ring is broken at azimuths east of the radar where the adjacent coastal mountains may have locally disrupted conditions. Particle fall streaks of moderate reflectivity at various azimuths complicate the double-ring pattern somewhat.

The NOAA crew at the X-band radar site at FRS launched a series of radiosondes at approximately 2-h intervals during the storm, starting at 0604 UTC, when only virga was present, and ending at 1657 UTC, by which time the rainfall accumulation had reached 37 mm and the rain was ending. Temperature data from the radiosondes are plotted in Fig. 4 for the first four launches. Overrunning warm air created a strong elevated temperature inversion. Its downward progress can be tracked in Fig. 4 by the gradual lowering of the top of the inversion, centered at 2.3 km MSL in the first sounding and at 1.8 km in the fourth sounding 6 h later. Sounding 2 most closely corresponds to the times of the radar images shown in Fig. 3.

Temperature profiles from the first two soundings (Fig. 4) clearly contained two melting levels (T = 0°C) aloft and an intervening layer of air with T < 0°C. In sounding 1, the maximum temperature in the elevated warm layer was 0.5°C and the thickness of the air with T > 0°C was 215 m. The subfreezing layer beneath it was 738 m thick and reached a minimum temperature of −5.0°C. This profile suggests, as will be shown later, that probably only partial melting of falling snowflakes occurred in the thin, elevated warm layer and substantial or complete refreezing may have occurred in the deeper cold air beneath it. The refrozen hydrometeors would finally melt completely after reaching the lower 0°C level near 1.5 km MSL. By the time of sounding 2, about 2 h later, the elevated warm layer had warmed to 1.1°C and the minimum temperature in the underlying subfreezing layer had also increased to −2.0°C. By the time of sounding 3, the overrunning warm air had advanced and deepened sufficiently to almost eliminate the intervening layer of subfreezing air, and by sounding 4 it was completely gone, leaving only the higher 0°C altitude. Rainfall had intensified sharply by the time of sounding 3.

Figure 5 shows the temperature and wind profiles of sounding 2, along with the nearly concurrent reflectivity and vertical Doppler velocity profiles from the zenith beam of the X-band radar’s RHI scan at 0757 UTC. The upper-brightband reflectivity maximum was located about 400 m below the top of the elevated melting layer, and the lower-brightband peak was about 250 m below top of the lower melting layer. These vertical separations between the 0°C isotherms and the reflectivity maxima are typical of those reported in numerous articles about single bright bands, such as Stewart et al. (1984), Willis and Heymsfield (1989), and Martner et al. (1993).

The radar’s measurement of mean Doppler vertical velocity is the combined effect of updraft speed minus particle terminal fall speed. Figure 5 shows that in the “snow” region, at altitudes above 2.6 km, the measured downward speeds were fairly steady at about 2 m s⁻¹. The observed fall speed is greater than that of most snowflakes (0.5–1.5 m s⁻¹), but not uncommon for graupel (Locatelli and Hobbs 1974). Grauple seems unlikely in this case, however, because of the modest reflectivities and stratiform nature of the storm. Therefore, it is more likely that the observed 2 m s⁻¹ speeds were produced by large snowflakes embedded in a modest downdraft of about 0.5–1.0 m s⁻¹. Within the shallow, elevated melting layer the hydrometeors’ downward speeds increased only slightly to about 3 m s⁻¹. This very minor acceleration indicates only partial melting occurred in this layer. In the refreezing layer the measured speeds varied between 2 and 4 m s⁻¹, perhaps affected by modest vertical air motions of <1 m s⁻¹. Complete melting finally occurred beneath the lower bright band where fall speeds reached values of ∼6 m s⁻¹, typical for small raindrops. The radiosonde’s wind data show that strong speed shear (0.012 s⁻¹) and veering (44°) occurred through the 537-m-thick warm-frontal temperature inversion layer that was created by the overrunning warm air. A jet of 28 m s⁻¹ coincided with the uppermost 0°C altitude near 2.4 km MSL.

The rainfall, aided by orographic forcing, was more intense at CZD, where the S-band profiler monitored conditions aloft. Figure 6 contains 8-h time–height displays of the velocity and reflectivity data from this radar. The S-PROF data suggest that a double-brightband structure existed intermittently between 0630 and 0900 UTC, as indicated by locally enhanced vertical gradients of reflectivity and velocity. Between 0630 and 0830 UTC minor oscillations of the lower-brightband height appear to be inversely correlated with the downward speeds of the frozen hydrometeors just above it. Assuming the 0°C isotherm height was steady during this time, the observed undulation is consistent with downward growth of the melting-layer thickness when larger, more rapidly falling ice particles are present, as described by Fabry and Zawadzki (1995). Except for a brief moment at about 0640 UTC, the upper bright band is very faint or absent until about 0800 UTC, which was the time of sounding 2 at FRS. This absence again suggests that the elevated layer of warmer-than-freezing air was too thin to cause substantial melting. The snowflakes may have melted enough for their surfaces to get wet but there was not enough melting to increase their fall speeds and decrease their reflectivity noticeably.

After approximately 1000 UTC, Fig. 6 shows that only the upper bright band continued to exist, which is in agreement with the continued warming of the temperature profile (Fig. 4, soundings 3 and 4). This was also approximately concurrent with a sharp intensification of the rainfall at all three sites. Rain rates measured by the gauge at the CZD S-band profiler site reached 14 mm h⁻¹ and maintained that rate for 5 h, according to the rain gauge data. Reflectivites observed by S-PROF during this period in the rain altitudes were about 40 dBZ and fall speeds were about 7.5 m s⁻¹. According to the plots of Atlas et al. (1973), based on simple theory and observations, these reflectivity, fall speed, and rain rate values are very reasonable for the raindrop size distributions and Z–R relation of Marshall and Palmer (1948).

The operational profiler melting-layer algorithm of White et al. (2002) is not designed to find more than one melting layer aloft. It examines vertical gradients of reflectivity and velocity upward from the surface until it detects the presence of a bright band, but does not continue searching higher. In this case, when two bright bands are apparent from visual inspection of the profiler’s data, the algorithm only specifies the height of the reflectivity peak within the lower melting layer. However, in a special modification that continues the upward search for this study, the algorithm does specify a double-melting-layer structure over the CZD site, as indicated in Fig. 6, at 0630 UTC and more consistently from 0800 to 0845 UTC at approximately 1.2 and 2.3 km MSL. The heights agree well with the brightband heights observed from the shoreline at FRS by the X-band radar at that time. At BBY, however, the vertical velocity gradient beneath the upper bright band before 0800 UTC was too weak to meet signature criteria, and even the modified algorithm failed to specify a double-melting-level configuration (Fig. 2), although the profiler’s SNR plots (not shown) suggest two reflectivity bright bands existed there as well.

When double bright bands exist, hydrological applications are more impacted by the height of the lower melting layer, whereas, aviation may be interested in both layers. Thus, it may be prudent to employ the revised melting-layer algorithm routinely with the profilers.

5. Summary and conclusions

Temperature inversions formed by frontal surfaces aloft can produce multiple melting layers with associated radar bright bands. The double-brightband configuration is fairly uncommon and has been described in only a few studies. The landfalling storm in northern California on 25 February 2004 contained a well-defined double bright band that was observed with scanning and profiling radars. This case is particularly interesting because, unlike earlier studies, serial radiosonde data, coincident in time and location with the radar observations, directly documented the thermal structure of the air in which the two bright bands existed. The cause (two melting layers aloft) and effect (two bright bands) are clearly illustrated. The case was also a unique opportunity to test the melting-layer algorithm of White et al. (2002), which is routinely applied to the data of wind profiling and precipitation profiling radars. As expected, the operational algorithm only indicated the presence of the lower melting level, but in modified form it detected both at times.