You are on page 1of 13

VOLUME 43 JOURNAL OF APPLIED METEOROLOGY NOVEMBER 2004

Freezing-Level Estimation with Polarimetric Radar


EDWARD A. BRANDES AND KYOKO IKEDA
National Center for Atmospheric Research,* Boulder, Colorado

(Manuscript received 19 November 2003, in final form 10 May 2004)

ABSTRACT
A simple empirical procedure for determining freezing levels with polarimetric radar measurements is de-
scribed. The algorithm takes advantage of the strong melting-layer signatures and the redundancy provided by
the suite of polarimetric radar measurements—in particular, radar reflectivity, linear depolarization ratio, and
cross-correlation coefficient. Freezing-level designations can be made with all volumetric scanning strategies.
Application to uniform (stratiform) precipitation within 60 km of the radar and with brightband reflectivity
maxima of greater than 25 dBZ suggests an accuracy of 100–200 m.

1. Introduction ments of radar reflectivity and vertical velocity for de-


termining the 08C level. The increase in hydrometeor
An automated method for estimating freezing-level terminal velocity that accompanies melting is used to
heights would have several benefits.1 Freezing-level confirm designations based on reflectivity and may per-
heights are required for designating potential icing lay- mit freezing-level designations in situations in which
ers in storms and for verifying their prediction with strong reflectivity signatures are not observed.
numerical forecast models. Knowing the height of the Here we examine the utility of polarimetric mea-
freezing level is important for determining whether pre- surements for designating freezing levels. Measure-
cipitation observed by radar is rain or snow and is nec- ments of differential reflectivity, linear depolarization
essary for its quantification. Alerts could be provided ratio, cross-correlation coefficient, and differential prop-
for those engaged in snow removal and aircraft-deicing agation phase are particularly sensitive to the presence
operations. The freezing-level height is also a compo- of large, wetted particles characteristic of melting layers.
nent of polarimetric radar–based algorithms for general In comparison with radar reflectivity, they can give more
hydrometeor classification (e.g., Vivekanandan et al. precise definition of melting layers and may exhibit
1999). melting-layer signatures in convective precipitation.
The long-recognized association between melting hy- Although the problems are closely related, interest
drometeors and bright bands in vertical profiles of radar here is in freezing-level designation rather than melting-
reflectivity measurements can be exploited to estimate layer determination. In many respects the latter problem
08C heights (e.g., Mittermaier and Illingworth 2003; is more difficult. Radar-derived melting-layer charac-
Gourley and Calvert 2003). Retrievals should improve teristics depend on the habits and size of the frozen
if supplemental information is considered. White et al. particles present. Melting-layer upper boundaries are of-
(2002) describe a procedure using profiler measure- ten inferred by a change in curvature of the reflectivity
profile (e.g., Fabry and Zawadzki 1995). However, re-
flectivity increases can also arise from particle growth
1
The freezing level is defined (Glickman 2000) as the lowest level
in the atmosphere at which the temperature, measured by a ther-
and aggregation. It is not unusual for reflectivity in-
mometer exposed to the air, is 08C. It is recognized that cloud droplets creases to begin hundreds of meters above the freezing
may supercool to much lower temperatures before freezing and that level and well before melting-layer signatures occur in
the melting of hydrometeors may begin at a slightly warmer air tem- other polarimetric measurements. Although lower
perature because of the wet-bulb effect. boundaries of melting layers appear to be sharply de-
fined in radar measurements, the boundary is probably
* The National Center for Atmospheric Research is sponsored by indistinct. Large ice particles associated with stratiform
the National Science Foundation. precipitation can fall to temperatures of 58C and higher
before melting is complete (e.g., Willis and Heymsfield
Corresponding author address: Dr. Edward A. Brandes, National
1989). However, hydrometeors in the final stages of
Center for Atmospheric Research, P.O. Box 3000, Boulder, CO 80307. melting are virtually indistinguishable from raindrops.
E-mail: brandes@ncar.ucar.edu Radar depictions of melting layers are strongly influ-

q 2004 American Meteorological Society 1541

Unauthenticated | Downloaded 07/06/21 01:11 PM UTC


1542 JOURNAL OF APPLIED METEOROLOGY VOLUME 43

enced by beam smoothing (Sánchez-Diezma et al. pending on crystal type, ZDR can be 2–5 dB. Aggregates
2000), and their precise determination with operational have large axis ratios and low bulk density; ZDR for these
radars would require a deconvolution of the spatially particles is small (,0.5 dB).
distributed measurements. Hydrometeors whose principal axes are not aligned
Here the problem of freezing-level detection is re- with the electrical field of the transmitted radiation cause
duced to finding the height at which signatures for the a small amount of the energy to be depolarized and to
various polarimetric parameters are maximized and then appear in the orthogonal direction. The depolarized
using predetermined statistical relationships between (cross polar) signal stems primarily from nonspheroidal
those heights and the freezing level. Retrieval is facil- particles that wobble or tumble as they fall, creating a
itated by redundancy among polarimetric measurements distribution of orientations (canting angles). Signals are
and their sensitivity to mixed-phase hydrometeors. enhanced for wetted and melting particles. LDR (dB)
There is a potential problem. Upper portions of melting is defined as the logarithm of the ratio of cross-polar
layers may exhibit isothermal layers with dry-bulb tem- and copolar signals; that is,
peratures close to 08C (Stewart et al. 1984; Willis and
LDR 5 10 3 log(ZVH /ZH ),
Heymsfield 1989). The total distance over which melt-
ing occurs and the statistical relation between radar sig- where ZVH is the signal received at vertical polarization
natures and the freezing level could be affected. This (cross-polar return) for a transmitted horizontally po-
situation was not observed in the data examined here. larized wave. LDR is small (on the order of 234 to
We begin by describing the properties of polarimetric 225 dB, depending on antenna isolation) for rain and
radar measurements and examine typical profiles. Model dry snow. Wet and melting snow can have an LDR from
parameter profiles are then presented and used to esti- 215 to 220 dB.
mate freezing levels for several examples. Estimates are The cross-correlation coefficient rHV is computed
verified with soundings and aircraft observations. from reflectivity at horizontal and vertical polarization.
This parameter is sensitive to the distribution of particle
sizes, axis ratios, and shapes. Theoretical values are
2. Polarimetric measurements
;0.99 for raindrops and ice crystals. For melting ag-
Polarimetric measurements with strong melting-layer gregates, rHV can be less than 0.90. Because rHV and
signatures include radar reflectivity Z, differential re- LDR are both sensitive to the presence of large wetted
flectivity ZDR , linear depolarization ratio (LDR), cross- particles, their melting-layer responses are similar—but
correlation coefficient rHV , and different propagation of opposite sign.
phase FDP . [For detailed descriptions of these param- The above polarimetric parameters are derived from
eters, their usage, and typical values for different hy- signal measurements that depend upon backscattering
drometeor types, see Doviak and Zrnić (1993, chapter properties of illuminated particles. Radar waves are also
8)]. Radar reflectivity at horizontal polarization (H) and subject to attenuation and phase shifts. For an aniso-
vertical polarization (V) for a unit volume (mm 6 m 23 ) tropic medium like raindrops or pristine ice crystals,
are defined as propagation constants for horizontally and vertically po-

Z H,V 5
l4
p |K| 2
5 E
0
Dmax

sH,V (D)N(D) dD,


larized waves differ. Horizontally polarized waves
‘‘see’’ a larger particle cross section and consequently
propagate more slowly than do vertically polarized
where l is the radar wavelength, K is the dielectric factor waves. Signals returned to the receiver for the two po-
of water, sH,V (D) are the radar cross sections of scat- larizations exhibit different accumulative phase (time)
terers at horizontal and vertical polarization, N(D) is shifts depending on hydrometeor size, shape, orienta-
the size distribution, and D is the particle equivalent tion, quantity, and distance from the radar. The differ-
diameter. Reflectivity is most often expressed in ‘‘dBZ’’ ential phase shift FDP (8) actually has two components—
(10 3 logZH,V ). a propagative component related to the difference in
Differential reflectivity (dB) is computed from forward-scattering amplitudes and a component related
to a backscatter differential phase. The backscatter phase
ZDR 5 10 3 log(ZH /ZV ). shift can become significant when Mie scatterers (e.g.,
Differential reflectivity is positive (negative) for parti- large melting aggregates) are present. As a consequence,
cles whose major axes are close to horizontal (vertical). FDP measurements can also be used to detect melting
Differential reflectivity is zero for particles that are layers.
spherical or for aspherical particles with a random dis- Radar measurements utilized in this study were ob-
tribution of orientations. Raindrops tend to flatten and tained with the National Center for Atmospheric Re-
orient themselves with their major axes close to hori- search’s S-band dual-polarization radar (S-Pol). To
zontal. For rain, ZDR ranges from 0.3 to 3 dB. Pristine make the LDR measurement, the radar alternately trans-
ice crystals have small axis ratios (vertical dimension mits horizontally and vertically polarized electromag-
divided by horizontal dimension) and high bulk density, netic waves and simultaneously receives scattered sig-
and they fall with their major axes near horizontal. De- nals at both polarizations (Randall et al. 1997). Oper-

Unauthenticated | Downloaded 07/06/21 01:11 PM UTC


NOVEMBER 2004 BRANDES AND IKEDA 1543

FIG. 1. Profiles of polarimetric measurements (ZH , LDR, rHV , ZDR , and FDP ). The estimated 08C level (2.47 km) is shown by a horizontal
line. Heights are above mean sea level.

ational considerations dictate that the Weather Surveil- be an issue with operational radars. Vertical cross sec-
lance Radar-1988 Doppler (WSR-88D) modified for tions and vertical scans are not permitted with the WSR-
polarimetric measurements will transmit at 458 (slant) 88D, which currently makes measurements in a series
polarization and simultaneously receive returned signals of 3608 scans consisting of 9 or 14 elevation angles
at horizontal and vertical polarization (Doviak et al. from 0.58 to 19.58. Sánchez-Diezma et al. (2000) ex-
2000). The planned configuration precludes the LDR amined the dependence of reconstructed reflectivity pro-
measurement. files on the distribution of antenna elevation scans typ-
The radar observations were obtained during field ical of operational radars. Melting-layer signature max-
programs conducted in various geographical areas. The ima can be displaced upward or downward, depending
freezing-level detection algorithm to be described (sec- on the distribution of antenna elevation angles relative
tion 4) has been tested with both warm-season and cold- to the actual freezing level. However, averages of height
season precipitation. Experience shows that bright bands estimates should be unbiased. An alternate approach is
can often be detected to radar ranges of 40–60 km and to estimate the freezing level from the distribution of
more, depending on precipitation intensity. With in- measurements along radar rays at antenna elevation an-
creasing distance, beam smoothing causes the thickness gles that pass through the melting layer.
of the melting layer to be increasingly overestimated
and the magnitude of melting-layer signature extremes
3. Polarimetric melting-layer signatures
to be underestimated (e.g., Sánchez-Diezma et al. 2000).
Freezing-level detections may be prevented altogether Profiles of radar reflectivity, linear depolarization ra-
by signal loss in weak precipitation. tio, cross-correlation coefficient, differential reflectivity,
Freezing-level designations are possible with all vol- and differential propagation phase are shown in Fig. 1
umetric scanning strategies. Vertical cross-sectional for a winter storm observed at 1631 UTC 28 November
scans at constant azimuthal angle are well suited for 2001 as part of the Improvement of Microphysical Pa-
diagnosis because detailed height information is ob- rameterization through Observational Verification Ex-
tained along the various rays that make up the scan. periment (IMPROVE) field program conducted in the
Designations are also readily made from vertically Oregon Cascade mountains (Stoelinga et al. 2003). [Pre-
pointing data, but vertical resolution (gate length) can vious descriptions of polarimetric measurements in

Unauthenticated | Downloaded 07/06/21 01:11 PM UTC


1544 JOURNAL OF APPLIED METEOROLOGY VOLUME 43

melting layers include Bader et al. (1987), Hall et al. is not due to noise is supported by the cross-correlation
(1984), Herzegh and Jameson (1992), and Zrnić et al. coefficient, which does not decrease significantly with
(1993).] The profiles were computed at a radar range height. Differential propagation phase increases above
of 10 km. The data, from orographically forced strati- the melting layer, particularly above approximately 5
form precipitation, were collected in vertical cross sec- km. The increase is also an indication of particle align-
tions with 18 azimuthal and 0.68 vertical spacing. The ment.
measurements were averaged over a radius of 2 km in Melting and resultant changes in dielectric factor, par-
the horizontal direction and over 0.2 km in the vertical ticle fall speed, particle shape and size, and aggregation
direction. The 3-dB beamwidth (;0.2 km at 10 km) are factors that cause reflectivity bright bands (Battan
also contributes to the smoothing. Smoothing artifically 1973; Fabry and Zawadzki 1995). The total increase in
increases minima, reduces maxima, and deepens the ra- radar reflectivity from the maximum curvature point in
dar-indicated melting layer. Advection in nonuniform the reflectivity profile (23 dBZ) to the brightband max-
precipitation can cause radar reflectivity profiles to dif- imum (40 dBZ, 2.3 km) is 17 dB. Part of the increase
fer considerably from reflectivity along particle trajec- comes from changes in the density and dielectric factor
tories (Fabry and Zawadzki 1995, their Fig. 2). How- of the hydrometeors (6.5 dB, Fabry and Zawadzki
ever, the height at which the strongest melting-layer 1995). Enhanced aggregation is also believed to be a
signatures appear should be relatively unaffected. Pro- significant contributor to the increase in this case.
files of LDR and rHV are insensitive to advection. Melting also affects the other polarimetric measure-
Radar reflectivity in the upper ‘‘ice only’’ precipita- ments. LDR, rHV, and FDP values just above the melting
tion layer (3–7 km) increases toward the ground at a layer are 228 dB, 0.99, and 9.98, respectively. Rapid
rate of more than 4 dB km 21 (in situ hydrometeor ob- changes occur as melting ensues, with LDR increasing
servations are not available for this event). Hydrometeor to a maximum of 218 dB, rHV decreasing to a minimum
growth in this layer is thought to be due to vapor de- of 0.91 (a melting-layer signature maximum), and FDP
position and aggregation of small crystals. A rapid in- increasing to 10.78. Melting signatures for these three
crease in radar reflectivity begins at 3.2 km—well above variables peak at 2.1 km. The decrease in rHV is likely
the melting level of 2.47 km determined from three a consequence of a greater variety of shapes and axis
soundings released within 60 km of the radar and having ratios associated with partly melted particles and the
an average release time of 1612 UTC.2 The break point introduction of raindrops. Large LDR implies a distri-
in the reflectivity profile is considerably higher than that bution of mixed-phase particles with a broad distribution
for LDR and rHV (2.7 and 2.6 km, respectively).3 Dif- of canting angles, perhaps a result of increased hydro-
ferences from the measured freezing level respond in meteor fluttering and spinning. The increase in FDP may
part to radar beam and analysis smoothing. However, be due primarily to melting and particle growth through
in this case the higher altitude of the upper break point a backscatter differential phase shift. The indicated low-
in the reflectivity trace is thought to be due primarily er boundary of the melting layer is 1.7 km for reflec-
to a rapid increase in aggregation. Hobbs (1973) pre- tivity, 1.6 km for LDR and FDP , and 1.5 km for rHV .
sents data showing accelerated ice particle growth by For this event, the melting-layer signature for FDP close-
deposition and aggregation as ambient temperatures rise ly matches that for ZH , LDR, and rHV . In general, how-
above 258C and particle stickiness increases. Hence, ever, the measurement is noisy and responds to local
the profile is illustrative of a potential problem when gradients of reflectivity (Ryzhkov and Zrnić 1998), and
enhanced reflectivity is used as a melting indicator. the distribution of measurements through the melting
Differential reflectivity in the ice layer is small and layer may exhibit large fluctuations (e.g., Zrnić et al.
declines from 0.35 dB at 5.7 km to a minimum of 0.13 1993). S-Pol measurements on occasion show two max-
dB at 2.8 km (well above the 08C level determined from ima in the melting layer with a minimum between them
soundings). Hydrometeors in the uppermost regions of at the height of the reflectivity maximum. Because FDP
the ice layer either had a greater degree of orientation depends on the total pathlength, the global maximum
with respect to the horizon or a greater eccentricity than is usually at the storm top or at far distances for mea-
the particles near the melting layer. At other times (not surements at low antenna elevation angles.
shown) ZDR was less than 0.15 dB throughout the ice As temperatures rise and melting proceeds, the hy-
layer. Small ZDR values are due in part to the lower drometeors eventually collapse into raindrops. Particle
dielectric factor for ice particles and their low bulk den- terminal velocities increase, and faster-falling particles
sity. LDR in the ice layer increases toward the storm are removed from the region. The reflectivity decrease
top. It is possible that particles at higher elevations had may also respond to the breakup of partly melted hy-
a broader distribution of canting angles. That the signal drometeors. Reflectivity at the top of the rain layer is
;28 dBZ. LDR and rHV take on values close to those
in the ice layer.
2
All heights are above mean sea level (MSL).
3
Break points correspond to changes in the profile curvature as Differential reflectivity maxima associated with melt-
used, for example, by Fabry and Zawadzki (1995) to delimit reflec- ing usually occur at lower levels than do reflectivity
tivity bright bands. maxima. This fact suggests that the maximum eccen-

Unauthenticated | Downloaded 07/06/21 01:11 PM UTC


NOVEMBER 2004 BRANDES AND IKEDA 1545

FIG. 2. Model profiles (ZH , LDR, and rHV ). The assumed freezing level (4.5 km) is shown.

tricity of melting hydrometeors occurs at a lower level crystals and aggregates contribute equally to reflectivity.
in the atmosphere than does their maximum size, as- Assumed axis ratios were 0.2 and 0.8, respectively.
suming that the largest particles are located close to the However, observed profiles obtained from low-elevation
reflectivity maximum. The height depression of the ZDR and vertically pointing scans show little difference, an
maximum is a function of reflectivity magnitude. Mean indication that the observed particles were more spher-
depressions were typically 0.2 km for bright bands with ical on average. Although ZDR and FDP may provide
a reflectivity maximum of 25 dBZ and 0.5–0.6 km for useful information in some situations, the task of iden-
a maximum of 45 dBZ. In comparison, depressions for tifying those events is made unnecessary by robust sig-
LDR and rHV are smaller and more narrowly distributed natures for ZH , LDR, and rHV . As a consequence, only
(more discussion in section 4). S-Pol beams for hori- the latter parameters are used for designating freezing
zontal and vertical polarization have a small mismatch levels in this study.
that can cause spurious ZDR values in regions of strong
reflectivity gradient. The minimum in the ZDR trace at
4. Model and implementation
2.8 km is thought to be a manifestation of this problem.
Differential reflectivity spikes, like that at 1.9 km, ap- Model freezing-level profiles for reflectivity, linear
pear in several datasets. At times, as in Fig. 1, the spike depolarization ratio, and cross-correlation coefficient
is close to the bottom of the melting layer just when are shown in Fig. 2. The profiles are based on statistics
other parameters are losing their melting-layer signa- gathered from several field programs conducted in dif-
tures. It is not clear whether the spike is an artifact or ferent climatic regimes and are representative of average
representative of the melting process. On occasion, the profiles obtained with operational radars at ranges of
uppermost regions of storms exhibit ZDR values that are less than 60 km. The profiles are used to determine if
greater than 2 dB, indicative of pristine ice crystals. the basic melting-layer patterns described in section 3
Such measurements are often transient in our datasets exist. The procedure is insensitive to the profile details;
and, like FDP measurements in the melting layer, are not rather, it is the overall correspondence between the ob-
easily modeled. Also, there is no ZDR or FDP signature served and model profiles and the presence of melting-
at zenith. (This mode is used for ‘‘calibrating’’ the ZDR layer signature maxima that is important. For illustration
measurement and determining the error level in FDP .) purposes, a freezing-level height of 4.5 km is assumed.
Reflectivity is essentially independent of the antenna For reflectivity, a constant rain-layer value is also as-
elevation angle. Simulations conducted by Vivekanan- sumed (27 dBZ). Within the lower portion of the bright
dan et al. (1993) show that LDR and rHV have an ele- band, reflectivity increases with height to 35 dBZ over
vation-angle dependence for precipitation in which ice a depth of 0.7 km and then decreases to 23 dBZ over

Unauthenticated | Downloaded 07/06/21 01:11 PM UTC


1546 JOURNAL OF APPLIED METEOROLOGY VOLUME 43

the next 0.5 km. Above the upper break point (4.7 km), versus maximum brightband reflectivity in Fig. 3a. A
the reflectivity decreases at a rate of 6 dB km 21 . The conservative fit to the depressions is
upper break point is taken to be above the 08C level to
d1 5 0.0886 2 0.000 400ZH 1 0.000 112Z H2 , (1)
allow for aggregation that begins before melting. The
reflectivity maximum is shown to be 0.3 km below the where d1 (km) is the depression of the reflectivity max-
08C level. Although a depression of 0.3 km is a good imum from the estimated 08C level and ZH is in dBZ.
mean value, the actual offset depends on precipitation Also plotted in Fig. 3a is a sampling of depressions
intensity. Profiles presented by Fabry and Zawadzki from field experiments conducted in Florida and Brazil
(1995, their Fig. 10) disclose that the brightband re- for which sounding information was available. The de-
flectivity maximum lowers and the melting layer thick- pressions are based on measurements from constant an-
ens as the reflectivity increases in magnitude. The pro- tenna-elevation-angle and constant azimuthal-angle
files were constructed from 350 h of measurements ob- scans in a ratio of 6:7. The depressions, which are shown
tained with a vertically pointing radar at a resolution of regardless of correlation with the model, agree nicely
15 m. They assume that the upper break point in the in the mean with the measurements of Fabry and Za-
reflectivity profile coincides with the 08C level. De- wadzki (1995), an indication that aggregation above the
pressions for their profiles were determined and plotted 08C level does not dominate the relation.

FIG. 3. Height depressions of (a) maximum brightband ZH from the 08C level and (b) maximum
melting-layer signatures for LDR and rHV from the brightband reflectivity maximum.

Unauthenticated | Downloaded 07/06/21 01:11 PM UTC


NOVEMBER 2004 BRANDES AND IKEDA 1547

FIG. 4. Retrieved 08C heights. The grid spacing is 5 km. The upper (boldface) number is the
consensus freezing-level height (km MSL), the middle number is s h (km), and the lower number
is the number of parameters with r . 0.7. The reflectivity pattern at 1.58 elevation is underlaid.

Modeled values of LDR and rHV are taken as constant was applied to the measurements (Fig. 3b), and the result
and equal in the respective ice-only and rain-only layers. was added to (1) to obtain
The assumed values are 225 dB and 0.993. Model melt-
d 2,3 5 0.121 1 0.000 445ZH 1 0.0002Z H2 , (2)
ing-layer signatures for LDR and rHV are confined to
the 3.5–4.5-km layer. Melting-layer signature extremes where d 2 and d 3 are the depressions for the LDR and
(216 dB for LDR and 0.92 for rHV ) are assumed to rHV melting-layer signature maxima from the 08C level.
reside at the middle of the layer or 0.5 km below the For measurements collected as a stepped series of
08C level on average and 0.2 km below the reflectivity constant-elevation scans or as vertical cross sections,
brightband maximum. the algorithm constructs measurement profiles at 0.1 km
Melting-layer depressions for LDR maxima and rHV resolution. This vertical spacing readily reproduces
minima from the 08C level are also functions of pre- prominent profile features (e.g., Fig. 1). In general, all
cipitation intensity. Height depressions tend to be larger measurements within a 3-km horizontal radius of the
than for radar reflectivity. This is illustrated in Fig. 3b, desired profile location are averaged. The vertical av-
where the depression of LDR and rHV melting-layer ex- eraging interval is 0.2 km. For vertically pointing data,
tremes from the brightband reflectivity maximum are the range spacing of the radar (usually 0.15 km) is used,
shown. Mean heights for LDR and rHV melting-layer and the radar data are averaged in time over one antenna
extremes from constant-elevation scans are usually rotation of 3608.
within 0.1 km; hence, the depressions were assumed to Operational radars scan by rotating the antenna
be equal. The relationship between the depression of through full 3608 sweeps, incrementing the elevation
the LDR and rHV melting-layer extremes from the re- angle through a tilt series. Vertical resolution may be
flectivity maximum is not quite linear. A polynomial fit too coarse to construct a profile of radar parameters

Unauthenticated | Downloaded 07/06/21 01:11 PM UTC


1548 JOURNAL OF APPLIED METEOROLOGY VOLUME 43

FIG. 5. Time series of freezing-level designations for radar


reflectivity, linear depolarization ratio, and cross-correlation
coefficient. Observed freezing levels deduced from sound-
ings and aircraft are superimposed.

accurately. However, radar measurements from beams reflectivity maximum is determined and the model pro-
that pass through the melting level at high elevation file overlaid such that the height of the observed re-
angles in stratiform precipitation often show well-de- flectivity maximum coincides with the model reflectivity
fined bright bands. Vertical profiles with resolutions of maximum. A data file is then generated that consists of
a few tens of meters can be constructed from these data paired observed and model reflectivity values at heights
by plotting the measurements as a function of height for which measurements are available. The linear cor-
rather than distance. relation coefficient r1 is computed from the paired re-
For ZH and LDR all averaging is done in the loga- flectivities as a measure of how closely the observed
rithmic domain to reduce the impact of measurements profile matches the model. The procedure is repeated
with high reflectivity. Averaging for rHV is in linear for LDR and rHV by again matching the heights of melt-
space. Several thresholds are applied. Only those por- ing-layer signature extremes. For Fig. 1, r1 (ZH ) 5 0.91,
tions of profiles with radar reflectivity of more than 10
r 2 (LDR) 5 0.95, and r 3 (rHV ) 5 0.97. Observed profiles
dBZ and only those profiles with a reflectivity maximum
with a correlation of more than 0.7, that is, profiles with
of more than 25 dBZ are retained for analysis. These
constraints remove measurements associated with weak at least 50% of their variance explained by the model,
precipitation and those potentially influenced by system are retained. The estimated freezing-level height is
noise. Also, searches for freezing levels are conducted simply
only over atmospheric depths at which they are cli-
matologically likely. For winter storms in the Oregon h i 5 htmax i 1 d i ,
Cascade mountains, the limits were 1–4 km. (A limit
lower than 1 km was preferred but was prevented by
an extended ground-clutter problem.) For summer where i is the parameter index, htmax i is the height at
storms in Florida, the height limits were 3–6 km. which the signature maximum appears, and d i is the
The model profiles are used to verify that the expected estimated depression from the freezing level. For Fig.
melting-layer signatures are present through pattern 1, h1 5 2.55 km and h 2 5 h 3 5 2.57 km. A consensus
matching. For example, the height of the radar-observed (weighted) height is determined with

Unauthenticated | Downloaded 07/06/21 01:11 PM UTC


NOVEMBER 2004 BRANDES AND IKEDA 1549

FIG. 6. Retrieved freezing-level heights, as in Fig. 4, but for a summer stratiform rain event.
The grid spacing is 10 km; the reflectivity pattern at 1.28 elevation is underlaid. An aircraft
measurement of the freezing-level height and its location (dot) are shown.

Ohr . of the signature maxima. By keying on profile extremes,


n
2
i i the need to retrieve the details of the melting-layer sig-
5
Or
h fzlv n natures by a deconvolution of the smoothed radar mea-
2
i surements is avoided.
The summation is for the n parameters that meet the
correlation criterion. Profiles with higher correlations 5. Results
have greater weight, and the final precision of the con-
sensus freezing-level estimate will, in general, be better a. Examples
than 100 m. For Fig. 1, hfzlv 5 2.56 km. As a measure
of scatter, the standard deviation sh is computed from Freezing-level retrievals on a 5-km grid determined
from vertical cross-sectional scans for the Cascades

! O (h 2 h
1 n storm at 1538 UTC are shown in Fig. 4. The upper
sh 5 i fzlv )2. number to the right of the grid point is the consensus
n
height estimate. The central number is the standard de-
Usually, s h is nonzero, and the parameter serves as a viation of the estimates, and the lower number is the
confidence factor for the retrieval. number of parameter profiles from which the consensus
Note that the correlation is independent of systematic height is estimated. Freezing-level designations are
differences between the observed and model profiles. made to a range of 30 km. The spread in consensus
Further, correlations are relatively insensitive to profile designations is narrow. The largest standard deviations
details, in particular for LDR and rHV , because the cor- occur in regions with ground clutter and weak precip-
relation magnitude is dictated largely by the matching itation. Inspection of the estimated heights reveals that

Unauthenticated | Downloaded 07/06/21 01:11 PM UTC


1550 JOURNAL OF APPLIED METEOROLOGY VOLUME 43

FIG. 7. Measurements of ZH , LDR, and rHV at 68 antenna elevation for the dataset in Fig. 6. The lower-right panel shows retrieved
freezing-level heights, as in Fig. 6, except computed from measurements at 4.88, 6.08, 7.28, and 9.68 antenna elevation.

the highest freezing levels are close to the radar and at any particular time is about 0.1 km. The series for
that the freezing level lowers to the east (right). radar reflectivity shows a number of spurious desig-
Trends in the designated freezing levels for the three nations at heights of 1.4–2.0 km. These arise from clut-
radar parameters are presented in Fig. 5. Individual data ter-contaminated measurements and can easily be re-
points represent either median or weighted-mean heights moved with a consistency check.
determined for the domain shown in Fig. 4. Weights Freezing-level heights determined from soundings re-
were assigned as in section 4. The decrease in heights leased at Salem, Oregon, (60 km north-northwest of the
beginning at 1800 UTC coincides with passage of a cold radar) and at a special sounding site (55 km to the south-
front. west) and from aircraft observations (25 km west of the
Examination of the scatter about the mean trends radar) are also shown in Fig. 5. The comparison is ex-
shows that for this storm the most repeatable designa- cellent, suggesting that detailed information regarding
tions are from the LDR parameter. The range in heights the 08C level can be retrieved and that errors are small.

Unauthenticated | Downloaded 07/06/21 01:11 PM UTC


NOVEMBER 2004 BRANDES AND IKEDA 1551

FIG. 8. Histograms of freezing-level designations based on individual radar parameters and the
consensus estimate.

For individual parameter designations on 28 November,


the percentage of profiles that did not meet the corre-
lation criterion was 37% for reflectivity, 9% for linear
depolarization ratio, and 54% for cross-correlation co-
efficient.
A retrieval for a summer stratiform rain event that
formed off the east coast of Florida during the ‘‘PRE-
CIP98’’ field program (Brandes et al. 2002) is presented
in Fig. 6. The radar was operated in sector mode, and
the antenna was stepped through a series of elevation
scans. The storm was somewhat stronger than that in
Fig. 4, permitting freezing-level designations to dis-
tances of 60 km. The actual freezing level, as observed
by a research aircraft, is shown. In Fig. 6, two LDR
profiles and four rHV profiles did not meet the correlation
criterion.
An example of melting-layer signatures for ZH , LDR,
and rHV at 6.08 elevation for the dataset in Fig. 6 is
presented in Fig. 7. The lower-right panel in Fig. 7
shows retrievals for constant elevation scans at 4.88,
6.08, 7.28, and 9.68. The data were averaged over azi-
muthal sectors of 108 at each range bin. Plotted data
points are centered on the sector midpoint and radar
range corresponding to the designated freezing-level
FIG. 9. Algorithm-retrieved freezing levels for winter (IMPROVE) and height. For this analysis, one reflectivity and four linear
summer (PRECIP98) storms plotted against in situ measurements. depolarization ratio profiles were rejected. The melting-

Unauthenticated | Downloaded 07/06/21 01:11 PM UTC


1552 JOURNAL OF APPLIED METEOROLOGY VOLUME 43

level pattern, with heights decreasing on average to the 0.2 km at radar ranges of less than 60 km. Potential
east, is similar to that in Fig. 6. There are some differ- problems could arise with events characterized by deep
ences that stem from the way the data were processed. isothermal 08C layers. However, application to the re-
The input data in each case are the same. However, in flectivity profiles from isothermal events described by
Fig. 6 the data are interpolated to a rectangular grid to Stewart et al. (1984) and Willis and Heymsfield (1989)
produce a column of measurements whose vertical res- suggests that the procedure should yield reasonable es-
olution is dictated in large part by the spacing between timates for the base of the 08C layer. The response of
successive elevation scans (1.28). At a range of 40 km, LDR and rHV in deep isothermal 08C layers is yet to be
the vertical distance between scans is about 0.8 km. determined. The current algorithm was not designed to
Melting-layer signature maxima can be displaced up- determine multiple freezing levels routinely. However,
ward or downward, depending on the distribution of designations can be made if there is sufficient separation
antenna elevation angles relative to the actual freezing between freezing levels and algorithm searches are re-
level (Sánchez-Diezma et al. 2000). stricted to heights that contain only one freezing level
The large number of designations that is possible for (e.g., Ikeda et al. 2004, manuscript submitted to J. At-
a precipitation area can be used for evaluation, much mos. Sci.).
like the consistency among individual parameters at a
particular location. Figure 8 shows a histogram of freez- Acknowledgments. We thank Robert Rilling and Jean
ing-level designations for the individual radar param- Hurst for preparing the numerous radar data tapes used
eters and the consensus estimates for yet another storm in the analysis and Drs. Marcia Politovich and Guifu
system. Estimated freezing-level heights are for the grid Zhang for constructive comments on the manuscript.
shown in Fig. 4 and include all designations within a This research is in response to requirements of and fund-
30-min window. Some of the variation is undoubtedly ing from the Federal Aviation Administration (FAA).
caused by meteorological factors. However, the number The views expressed are those of the authors and do
counts are concentrated in the 2.1-km-height category not necessarily represent the official policy or position
for each parameter. The dispersion in Fig. 8 can be used of the FAA.
to eliminate suspected retrievals. The most consistent
freezing-level designations in this case (after removing
outliers) are with radar reflectivity. REFERENCES

Bader, M. J., S. A. Clough, and G. P. Cox, 1987: Aircraft and dual


b. Verification polarization radar observations of hydrometeors in light strati-
form precipitation. Quart. J. Roy. Meteor. Soc., 113, 491–515.
A comparison of algorithm freezing-level designa- Battan, L. J., 1973: Radar Observation of the Atmosphere. University
tions and in situ measurements made by research aircraft of Chicago Press, 279 pp.
and local temperature soundings is presented in Fig. 9. Brandes, E. A., G. Zhang, and J. Vivekanandan, 2002: Experiments
The available verification data for IMPROVE are from in rainfall estimation with a polarimetric radar in a subtropical
soundings located 25–60 km from the radar and not in environment. J. Appl. Meteor., 41, 647–685.
Doviak, R. J., and D. S. Zrnić, 1993: Doppler Radar and Weather
the region of volumetric sampling. Comparisons labeled Observations. Academic Press. 562 pp.
as PRECIP98 are based on coincident aircraft and radar ——, V. Bringi, A. Ryzhkov, A. Zahrai, and D. Zrnić, 2000: Con-
measurements. The dataset is small, but overall the siderations for polarimetric upgrades to operational WSR-88D
agreement is good, with little or no bias. The mean radars. J. Atmos. Oceanic Technol., 17, 257–278.
Fabry, F., and I. Zawadzki, 1995: Long-term observations of the
absolute error (the departure from the 1:1 line) is 0.14
melting layer of precipitation and their interpretation. J. Atmos.
km, and the root-mean-square error is 0.17 km 2 . Sci., 52, 838–851.
Glickman, T., Ed., 2000: Glossary of Meteorology. 2d ed. Amer.
Meteor. Soc., 855 pp.
6. Summary and conclusions Gourley, J. J., and C. M. Calvert, 2003: Automated detection of the
bright band using WSR-88D data. Wea. Forecasting, 18, 585–
A simple empirical algorithm for estimating freezing- 598.
level heights from polarimetric radar measurements of Hall, M. P. M., J. W. F. Goddard, and S. M. Cherry, 1984: Identifi-
reflectivity, linear depolarization ratio, and cross-cor- cation of hydrometeors and other targets by dual-polarization
relation coefficient is described. Designations are based radar. Radio Sci., 19, 132–140.
on melting-layer signatures and are determined by Herzegh, P. H., and A. R. Jameson, 1992: Observing precipitation
through dual-polarization radar measurements. Bull. Amer. Me-
matching observed profiles of the polarimetric variables teor. Soc., 73, 1365–1374.
with idealized profiles. Consensus freezing-level des- Hobbs, P. V., 1973: Ice in the atmosphere: A review of the present
ignations are made by weighing the goodness of fit for position. Physics and Chemistry of Ice, E. Whalley, S. J. Jones,
the individual parameters. The algorithm was designed and L. W. Gold, Eds., Royal Society of Canada, 308–319.
for stratiform precipitation with brightband reflectivity Mittermaier, M. P., and A. J. Illingworth, 2003: Comparison of model-
derived and radar-observed freezing-level heights: Implications
of greater than 25 dBZ. Testing on a limited dataset for vertical reflectivity profile-correction schemes. Quart. J. Roy.
reveals that the method allows for retrieval of fine detail Meteor. Soc., 129, 83–95.
in the freezing-level surface with an accuracy of 0.1– Randall, M., J. Lutz, and J. Fox, 1997: S-Pol’s high isolation me-

Unauthenticated | Downloaded 07/06/21 01:11 PM UTC


NOVEMBER 2004 BRANDES AND IKEDA 1553

chanical polarization switch. Preprints. 28th Conf. On Radar radar modeling of mixtures of precipitation particles. IEEE
Meteorology, Austin, TX, Amer. Meteor. Soc., 252–253. Trans. Geosci. Remote Sens., 31, 1017–1030.
Ryzhkov, A., and D. Zrnić, 1998: Beamwidth effects on the differ- ——, D. S. Zrnić, S. M. Ellis, R. Oye, A. V. Ryzhkov, and J. Straka,
ential phase measurements of rain. J. Atmos. Oceanic Technol., 1999: Cloud microphysics retrieval using S-band dual-polari-
15, 624–634. zation measurements. Bull. Amer. Meteor. Soc., 80, 381–388.
Sánchez-Diezma, R., I. Zawadzki, and D. Sempere-Torres, 2000: White, A. B., D. J. Gottas, E. T. Strem, F. M. Ralph, and P. J. Neiman,
Identification of the bright band through the analysis of volu- 2002: An automated brightband height detection algorithm for
use with Doppler radar spectral moments. J. Atmos. Oceanic
metric radar data. J. Geophys. Res., 105, 2225–2236.
Technol., 19, 687–697.
Stewart, R. E., J. D. Marwitz, J. C. Page, and R. E. Carbone, 1984: Willis, P. T., and A. J. Heymsfield, 1989: Structure of the melting
Characteristics through the melting layer of stratiform clouds. layer in mesoscale convective system stratiform precipitation. J.
J. Atmos. Sci., 41, 3227–3237. Atmos. Sci., 46, 2008–2025.
Stoelinga, M. T., and Coauthors, 2003: Improvement of Microphysical Zrnić, D. S., N. Balakrishnan, C. L. Ziegler, V. N. Bringi, K. Aydin,
Parameterization through Observational Verification Experi- and T. Matejka, 1993: Polarimetric signatures in the stratiform
ment. Bull. Amer. Meteor. Soc., 84, 1807–1826. region of a mesoscale convective system. J. Appl. Meteor., 32,
Vivekanandan, J., R. Raghavan, and V. N. Bringi, 1993: Polarimetric 678–693.

Unauthenticated | Downloaded 07/06/21 01:11 PM UTC

You might also like