• Andrieu, H., and Creutin J. D. , 1995: Identification of vertical profiles of radar reflectivities for hydrological applications using an inverse method. Part I: Formulation. J. Appl. Meteor., 34 , 225239.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Austin, P. M., and Bemis A. , 1950: A quantitative study of the bright band in radar precipitation echoes. J. Meteor., 7 , 165171.

  • Baeck, M. L., and Smith J. A. , 1998: Rainfall estimation by the WSR-88D for heavy rainfall events. Wea. Forecasting, 13 , 416436.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Benjamin, S. G., Brown J. M. , Brundage K. J. , Schwartz B. , Smirnova T. , and Smith T. L. , 1998: The operational RUC-2. Preprints, 16th Conf. on Weather Analysis and Forecasting, Phoenix, AZ, Amer. Meteor. Soc., 249–252.

    • Search Google Scholar
    • Export Citation
  • Borga, M., Anagnostou E. N. , and Krajewski W. F. , 1997: A simulation approach for validation of a brightband correction method. J. Appl. Meteor., 36 , 15071518.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Cheng, M., and Collier C. G. , 1993: An objective method for recognizing and partially correcting brightband error in radar images. J. Appl. Meteor., 32 , 11421149.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Doviak, R. J., and Zrnic D. S. , 1984: Doppler Radar and Weather Observations. Academic Press, 593 pp.

  • Fabry, F., and Zawadzki I. , 1995: Long-term radar observations of the melting layer of precipitation and their interpretation. J. Atmos. Sci., 52 , 838851.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Fulton, R. A., Breidenbach J. P. , Seo D. J. , Miller D. A. , and O'Bannon T. , 1998: The WSR-88D rainfall algorithm. Wea. Forecasting, 13 , 377395.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gourley, J. J., Maddox R. A. , Burgess D. W. , and Howard K. W. , 2002: An exploratory multisensor technique for quantitative estimation of stratiform rainfall. J. Hydrometeor., 3 , 166180.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Harrold, T. W., and Kitchingman P. G. , 1975: Measurement of surface rainfall using radar when the beam intersects the melting layer. Preprints, 16th Conf. on Radar Meteorology, Houston, TX, Amer. Meteor. Soc., 473–478.

    • Search Google Scholar
    • Export Citation
  • Howard, K. W., Gourley J. J. , and Maddox R. A. , 1997: Uncertainties in WSR-88D measurements and their impacts on monitoring life cycles. Wea. Forecasting, 12 , 166174.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kain, J. S., Goss S. M. , and Baldwin M. E. , 2000: The melting effect as a factor in precipitation-type forecasting. Wea. Forecasting, 15 , 700714.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kitchen, M., and Jackson P. M. , 1993: Weather radar performance at long range—Simulated and observed. J. Appl. Meteor., 32 , 975985.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Maddox, R. A., Zaras D. S. , MacKeen P. L. , Gourley J. J. , Rabin R. , and Howard K. W. , 1999: Echo height measurements with the WSR-88D: Use of data from one versus two radars. Wea. Forecasting, 14 , 9941006.

    • Search Google Scholar
    • Export Citation
  • Maddox, R. A., Zhang J. , Gourley J. J. , and Howard K. W. , 2002: Weather radar coverage over the contiguous United States. Wea. Forecasting, 17 , 927934.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Marwitz, J., and Toth J. , 1993: The Front Range blizzard of 1990. Part I: Synoptic and mesoscale structure. Mon. Wea. Rev., 121 , 402415.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • O'Bannon, T., 1997: Using a “terrain-based” hybrid scan to improve WSR-88D precipitation estimates. Preprints, 28th Conf. on Radar Meteorology, Austin, TX, Amer. Meteor. Soc., 506–507.

    • Search Google Scholar
    • Export Citation
  • OFCM, 1991: Doppler radar meteorological observations. Part C: WSR-88D products and algorithms. Federal Meteorological Handbook, FCH-H11C-1991, Office of the Federal Coordinator for Meteorological Services and Supporting Research, Rockville, MD, 210 pp.

    • Search Google Scholar
    • Export Citation
  • Sanchez-Diezma, R., Zawadzki I. , and Sempere-Torres D. , 2000: Identification of the bright band through the analysis of volumetric radar data. J. Geophys. Res., 105 (D2) 22252236.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Seo, D. J., Breidenbach J. , Fulton R. , and Miller D. , 2000: Real-time adjustment of range-dependent biases in WSR-88D rainfall estimates due to nonuniform vertical profile of reflectivity. J. Hydrometeor., 1 , 222240.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Smith, C. J., 1986: The reduction of errors caused by bright bands in quantitative rainfall measurements made using radar. J. Atmos. Oceanic Technol., 3 , 129141.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Stewart, R. E., 1984: Deep 0°C isothermal layers within precipitation bands over southern Ontario. J. Geophys. Res., 89 , 25672572.

  • Vicente, G. A., Scofield R. A. , and Menzel W. P. , 1998: The operational GOES infrared rainfall estimation technique. Bull. Amer. Meteor. Soc., 79 , 18831898.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Westrick, K. J., Mass C. F. , and Colle B. A. , 1999: The limitations of the WSR-88D radar network for quantitative precipitation measurement over the coastal western United States. Bull. Amer. Meteor. Soc., 80 , 22892298.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • White, A. B., Gottas D. J. , Strem E. T. , Ralph F. M. , and Neiman P. J. , 2002: An automated brightband height detection algorithm for use with Doppler radar spectral moments. J. Atmos. Oceanic Technol., 19 , 687697.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • View in gallery
    Fig. 1.

    Vertical cross section of radar beams in VCP-11. The region with gray horizontal lines represents a theoretical melting layer, with radar-derived brightband top and bottom heights shown in black

  • View in gallery
    Fig. 2.

    (a) Time series of brightband top heights (in black) from the KIWA radar and 0°C heights (in gray) measured independently by the Tucson, AZ, radiosonde from 0804 to 1754 UTC 4 Feb 1998. (b) Same as in (a) but from 1700 UTC 8 Feb to 1200 UTC 9 Feb 1998. (c) Same as in (a) but from 0604 to 1206 UTC 15 Feb 1998. (d) Same as in (a) but from 0317 UTC 6 Mar 2000 to 0006 UTC 7 Mar 2000.

  • View in gallery
    Fig. 3.

    (a) Time series of brightband top heights (in black) from the KLZK radar and 0°C heights (in gray) measured independently by the Little Rock, AR, radiosonde from 0506 to 1200 UTC 8 Jan 2000. (b) Same as (a) for the KDDC radar measured independently by the Dodge City, KS, radiosonde from 1013 to 1924 UTC 2 Mar 2000. (c) Same as (a) for the KTLX radar measured independently by the Norman, OK, radiosonde from 0508 to 1229 UTC 29 Jan 2001. (d) Same as (a) for the KINX radar measured independently by the Springfield, MO, radiosonde (line in gray) and by the Norman, OK, radiosonde (line in black) from 0148 to 1534 UTC 29 Jan 2001

  • View in gallery
    Fig. 4.

    (a) Observed sounding from DDC plotted on a standard skewT–logp diagram, valid 1200 UTC 2 Mar 2000. (b) Same as (a), but valid 0000 UTC 3 Mar 2000.

  • View in gallery
    Fig. 5.

    (a) Time series of brightband top heights (in black) from the KCAE radar and 0°C heights (in gray) derived independently from the nearest RUC-2 model grid point from 2100 UTC 6 Feb to 1200 UTC 8 Feb 2002. (b) Same as (a) for the KAKQ radar and derived independently from the nearest RUC-2 model grid point from 1500 UTC 7 Feb to 0000 UTC 8 Feb 2002. (c) Same as (a) for the KLTX radar and derived independently from the nearest RUC-2 model grid point from 0300 UTC 7 Feb to 0900 UTC 8 Feb 2002. (d) Same as (a) for the KRAX radar and derived independently from the nearest RUC-2 model grid point from 0600 UTC 7 Feb 2002 to 0000 UTC 8 Feb 2002

  • View in gallery
    Fig. 6.

    An S-Band Precipitating-Cloud Radar System image and specifications

  • View in gallery
    Fig. 7.

    Time series of brightband top and bottom heights (in black) from the KIWA radar overlaid on reflectivity observations from an independent vertically pointing radar located 141 km away for 24-h period beginning at 0000 UTC 6 Mar 2000

  • View in gallery
    Fig. 8.

    Precipitation rate (mm h−1) from the KIWA radar with brightband identification areas overlaid (in black) at 1724 UTC 6 Mar 2000

  • View in gallery
    Fig. 9.

    The terrain-based hybrid scan lookup table used at the KIWA radar site. The tilt numbers (shaded in grayscale) correspond to the elevation angles (0.5°–3.35°) used by the WSR-88D for precipitation estimation. The maximum range is 230 km

  • View in gallery
    Fig. 10.

    Frequency diagram for the maximum reflectivity values in each grid column (the composite reflectivity) for the calibration dataset

  • View in gallery
    Fig. 11.

    Sensitivity of brightband depth calculations to the reflectivity reduction parameter and minimum reflectivity thresholds ranging from 24 to 36 dBZ

  • View in gallery
    Fig. 12.

    Scatterplot of the percent coverage parameter vs the standard deviation of the brightband top parameter for the calibration dataset. Each square has been shaded (in grayscale) to indicate the number of occurrences of data points falling in that area

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 312 111 19
PDF Downloads 153 70 6

Automated Detection of the Bright Band Using WSR-88D Data

Jonathan J. GourleyCooperative Institute for Mesoscale Meteorological Studies, University of Oklahoma, Norman, Oklahoma

Search for other papers by Jonathan J. Gourley in
Current site
Google Scholar
PubMed
Close
and
Chris M. CalvertCooperative Institute for Mesoscale Meteorological Studies, University of Oklahoma, Norman, Oklahoma

Search for other papers by Chris M. Calvert in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

During stratiform precipitation, hydrometeors within the melting layer increase backscatter to radar. This layer can persist at a nearly constant height for hours and can lead to serious radar-based overestimates in accumulated surface rainfall. Sophisticated precipitation algorithms of the present and near future are beginning to identify regions where there is contaminated reflectivity in order to make corrections to the data. An automated algorithm that operates on full-resolution Weather Surveillance Radar-1988 Doppler (WSR-88D) reflectivity data (i.e., archive level II) to identify the height and depth of the bright band for every volume scan has been developed. Results from the algorithm are compared with 0°C heights from nearby radiosonde observations and from model analyses for three different regions in the United States. In addition, reflectivity observations from an independent, vertically pointing radar situated in complex terrain are compared with results from the brightband algorithm operating on WSR-88D data. The output from the brightband algorithm matches observations well. A case is presented to show how the radar-observed brightband heights can be used to identify regions in precipitation products where radar is sampling within the melting layer and therefore may be subject to overestimation. Improved monitoring of the bright band, because of the comparatively high temporal resolution of the radar observations, results from application of the algorithm. The algorithm output can provide guidance to forecasters who are using radar-based quantitative precipitation estimates to issue advisories and warnings. Moreover, the melting-layer observations can be used with a digital elevation model to map the approximate rain–snow line.

Additional affiliation: NOAA/OAR/National Severe Storms Laboratory, Norman, Oklahoma

Corresponding author address: Jonathan J. Gourley, National Severe Storms Laboratory, 1313 Halley Circle, Norman, OK 73069. Email: gourley@ou.edu

Abstract

During stratiform precipitation, hydrometeors within the melting layer increase backscatter to radar. This layer can persist at a nearly constant height for hours and can lead to serious radar-based overestimates in accumulated surface rainfall. Sophisticated precipitation algorithms of the present and near future are beginning to identify regions where there is contaminated reflectivity in order to make corrections to the data. An automated algorithm that operates on full-resolution Weather Surveillance Radar-1988 Doppler (WSR-88D) reflectivity data (i.e., archive level II) to identify the height and depth of the bright band for every volume scan has been developed. Results from the algorithm are compared with 0°C heights from nearby radiosonde observations and from model analyses for three different regions in the United States. In addition, reflectivity observations from an independent, vertically pointing radar situated in complex terrain are compared with results from the brightband algorithm operating on WSR-88D data. The output from the brightband algorithm matches observations well. A case is presented to show how the radar-observed brightband heights can be used to identify regions in precipitation products where radar is sampling within the melting layer and therefore may be subject to overestimation. Improved monitoring of the bright band, because of the comparatively high temporal resolution of the radar observations, results from application of the algorithm. The algorithm output can provide guidance to forecasters who are using radar-based quantitative precipitation estimates to issue advisories and warnings. Moreover, the melting-layer observations can be used with a digital elevation model to map the approximate rain–snow line.

Additional affiliation: NOAA/OAR/National Severe Storms Laboratory, Norman, Oklahoma

Corresponding author address: Jonathan J. Gourley, National Severe Storms Laboratory, 1313 Halley Circle, Norman, OK 73069. Email: gourley@ou.edu

1. Introduction

Quantitative estimates of rainfall rates using operational Weather Surveillance Radars-1988 Doppler (WSR-88D) data can be inaccurate because of sampling of hydrometeors in the melting layer. This region of artificially high reflectivity, or bright band (Austin and Bemis 1950), is caused by several factors related to the physical characteristics of melting hydrometeors. Specifically, there is an increase in the dielectric constant as hydrometeors melt, larger particle cross sections due to enhanced aggregation, and increased backscatter due to wetting of frozen hydrometeors. Fabry and Zawadzki (1995) suggest two additional processes that may explain the dramatic increase in reflectivity: increased backscatter due to the nonsphericity of hydrometeors undergoing melting (shape effect) and varying water distributions within a melting snowflake (density effect). The bottom of this layer is marked by the level where frozen aggregates melt completely to rain and particle fall velocities increase.

When reflectivity measurements are contaminated by brightband effects, surface rainfall overestimations can be as large as a factor of 10 (Smith 1986). A wealth of research has shown how precipitation estimates by radar are dependent upon the range (and thus the height) at which reflectivity was measured (e.g., Baeck and Smith 1998). Overestimates occur from brightband-contaminated reflectivity at close to midrange, while Kitchen and Jackson (1993) reveal that radar typically underestimates surface precipitation where beams overshoot the melting layer at longer ranges. Because of this, several attempts have been made to identify the brightband and account for its impact on surface rainfall estimates (e.g., Harrold and Kitchingman 1975; Smith 1986; Cheng and Collier 1993; Andrieu and Creutin 1995; Borga et al. 1997; Sanchez-Diezma et al. 2000; White et al. 2002). Recent studies (e.g., Seo et al. 2000) utilize a function that describes the vertical variability between reflectivity measured at some elevation angle aloft to that measured at the surface. This vertical profile of reflectivity (VPR) function is used to make corrections to surface rainfall estimates where reflectivity measurements are made in and above the melting layer. A VPR technique assumes the profile(s) measured at close range represent the reflectivity profiles under the entire radar domain. This has been noted as a potentially serious limitation of this technique (Borga et al. 1997). Questions also remain regarding the correlation between reflectivity measured at longer ranges, and thus higher heights, and surface rainfall rates.

In addition to affecting precipitation estimates by radar, proper identification of melting-layer heights can be useful for forecasting precipitation types (Marwitz and Toth 1993). Kain et al. (2000) demonstrate how the “melting effect” can produce a downward-propagating isothermal layer. They emphasize that this process can be dominant when local temperature changes from processes such as horizontal and vertical temperature advection, condensation, and evaporation are small, and they present a case where melting-layer heights descend and result in an unanticipated changeover from rain to heavy snow. It is believed that automated monitoring of the melting layer offers the potential to anticipate such changes in surface precipitation types.

The study presented here does not attempt to make corrections to precipitation estimates, rather a simple technique is presented to identify regions where precipitation estimates from WSR-88Ds are derived from or unaffected by measurements within the brightband layer. It is hoped that unambiguous identification of uncontaminated reflectivity data will be useful for future applications that evaluate the relative calibration differences between adjacent WSR-88Ds and those that use reflectivity data to calibrate other sensors such as satellite data for multisensor quantitative precipitation estimation (QPE; e.g., Vicente et al. 1998; Gourley et al. 2002). As will be shown in section 2, the developed algorithm identifies the top and bottom of the melting layer, which permits evaluation of its performance by comparing heights of the top of the melting layer with 0°C heights derived from radiosonde observations, model analyses, and reflectivity profiles from a vertically pointing radar (shown in section 3). The algorithm results will not be compared with output from VPR correction algorithms that operate on WSR-88D data already reported in the literature (e.g., Seo et al. 2000) because of the difficulty in extracting the actual melting-layer heights from the profile. Fabry and Zawadzki (1995) reveal an algorithm for retrieving the brightband top and bottom heights from VPR data collected using an X-band vertically pointing radar. The height resolution of this radar is approximately 15 m, which is at least an order of magnitude smaller than the WSR-88D and makes application of algorithms developed on such high-resolution data difficult to extend to operational WSR-88D platforms. The extraction algorithm relies on significant curvature in the profile near the top and bottom of the melting layer that may not be as recognizable with coarser WSR-88D data. Section 4 shows how regions in radar-derived precipitation maps suspected of having brightband contamination can be identified. The sensitivity of each adaptable parameter used in the brightband identification algorithm is tested in section 5, and conclusions are presented in section 6.

2. Description of brightband identification technique

a. Data ingest

The brightband identification algorithm (BBID hereinafter) ingests a full volume of high-resolution reflectivity data from WSR-88Ds every 5–6 min [dictated by radar scanning strategy; see OFCM (1991)]. The algorithm operates on reflectivity data (with data bins having an approximate beamwidth of 1° in the azimuthal and elevational directions by 1 km in range) from all tilts [dictated by volume coverage pattern (VCP); see OFCM (1991)] between ranges of 10 and 30 km from the radar. Within 10 km, significant gaps exist above the highest elevation angle (19.5°; see Fig. 1). The bright band may reside above the highest tilt at close range, thus remaining invisible to the radar. Beyond 30 km, beam depths exceed 500 m. The radar must have sufficient vertical resolution to accurately portray a brightband feature that has been observed to be only a few hundred meters thick (Smith 1986). This “beam spreading” at ranges longer than 30 km leads to echo height uncertainties exceeding typical depths of brightband features. See Howard et al. (1997) for a more complete discussion of beam height uncertainties. The vertical resolution of WSR-88D reflectivity data is suitable for accurately detecting a brightband signature within 10–30 km from the radar.

Many radars in the western United States are situated in complex terrain (Westrick et al. 1999; Maddox et al. 2002). Considerations must be made for terrain-induced beam blockages. Terrain obstacles can cause many low-elevation bins to be devoid of representative reflectivity samples. BBID thus utilizes the terrain-based hybrid scan (O'Bannon 1997) for any given radar as the lowest radar data bin in the volume scan. All reflectivity bins below the hybrid scan are discarded at this point. The upper radar data bin is set to the maximum elevation angle of 19.5° (also dictated by VCP). Reflectivity from data bins between the hybrid scan and the maximum tilt are used in BBID calculations with the following constraint. If a particular bin in the hybrid scan is composed of reflectivity measured from the third elevation angle (2.4°) or greater (suggesting that no reflectivity data from the lowest two tilts are available because of terrain blockages), then reflectivity data from all tilts above that particular grid point are discarded and not used in BBID calculations. At a range of 30 km, reflectivity from the third elevation angle comes from a beam center height of approximately 1.3 km above the radar elevation (Fig. 1). For some cool-season events, this height may be greater than the brightband feature being searched and thus limits the utility of data measured at these heights and higher. Next, BBID begins the searching procedure following the described selection of grid cells between 10 and 30 km that are devoid of significant terrain blockages.

b. The searching procedure

Following data ingest, BBID searches above each valid grid cell from the lowest bound (dictated by the hybrid scan) to the uppermost elevation angle (19.5°; dictated by radar VCP) for the maximum reflectivity value. This maximum value must exceed a minimum threshold of 30 dBZ (this threshold can be adapted for various meteorological scenarios and its sensitivity is tested in section 5). Data from grid columns, that is, bins at a fixed range and azimuth but at different elevations, that do not have any reflectivity values greater than 30 dBZ are discarded from this point forward. These grid points are assumed to be devoid of brightband-contaminated reflectivity. All reflectivity values in valid grid columns are normalized by the maximum reflectivity value found in that column. Then, BBID searches in elevation angles above and below the maximum reflectivity value for a reduction in reflectivity of 20% or greater. An increase (decrease) in the reflectivity reduction parameter (which can be adapted accordingly and is tested in section 5) results in thicker (thinner) calculated brightband depths. In the event of multiple melting layers (those layers having a maximum reflectivity greater than 30 dBZ), the algorithm will identify the layer closest to the ground because it searches for the column maximum from the bottom of the column up. The heights of the brightband top and bottom are then computed using the range and elevation angle, assuming the gradient of atmospheric refractive index is constant (Doviak and Zrnic 1984). There are instances in which the brightband top may be identified, but the bottom may be near or below ground level, making it difficult or impossible to identify the lower boundary. The BBID thus ingests 0°C heights from the Rapid Update Cycle-2 (RUC-2) model (Benjamin et al. 1998) on an hourly basis and ensures that this height is at least 1 km above the radar elevation. Otherwise, the algorithm will not report a brightband height for that volume scan.

The elevation angle corresponding to the top of the beam is used to calculate the brightband top height instead of using the beam center. A discussion about radar beam height uncertainties is relevant in this study because beam depths are nearly 500 m at a range of 30 km. Furthermore, similar to gradients near storm top, vertical reflectivity gradients in the vicinity of brightband layers are significant. Maddox et al. (1999) and Howard et al. (1997) found the radar consistently underestimates storm top by reporting the height calculated using the elevation angle corresponding to the beam center. Figure 1 shows a cross section of a theoretical bright-band layer (gray horizontal lines) superimposed on radar beams for a WSR-88D scanning in VCP-11. If we assume that the algorithm does not detect a bright band when more than 50% of the bin volume is above the layer, then the saw-toothed black line indicates where BBID would report the top of the bright band. Notice that the average of the BBID top heights in the radial direction straddle the theoretical brightband top. Use of the beam center elevations to compute the brightband top would result in an underestimated brightband top height. Similarly, the bottom of the bright band is calculated as shown in Fig. 1 using the elevation angle corresponding to the beam bottom.

All brightband top and bottom heights are averaged in the radial direction so that there are up to 360 radial-averaged brightband heights for a given volume scan. The radial-averaged brightband heights are used quantitatively in lieu of the individual gridpoint values from this point forward because the radial-averaging process smoothes out the range dependency of heights calculated from WSR-88D VCPs. The next section demonstrates how the BBID algorithm determines if brightband heights for a given volume scan shall be reported.

c. Brightband existence

BBID is designed to report results only when radar identifies a horizontally homogeneous layer of relatively high reflectivity from 10 to 30 km from the radar. Two criteria are used to determine if there is a spatially consistent bright band near the radar, and the sensitivities of both parameters are examined in section 5. The first criterion examines the number of grid columns that have at least one bin at any elevation angle with a reflectivity value in excess of 30 dBZ (adaptable threshold). This count is divided by the maximum number of potential grid columns from 10 to 30 km (up to 7200 assuming no beam blockages in the third elevation angle or greater). If this ratio is larger than 6% (adaptable threshold; corresponds to 432 grid columns), then the first criterion is met and the horizontal homogeneity of the detected brightband layer is tested further as described below.

BBID assesses the spatial consistency of the brightband layer by examining the variability of the radial-averaged brightband heights. The second existence criterion is met if the standard deviation of the radial-averaged brightband top heights is less than 500 m (adaptable threshold). The brightband top height is preferred over the brightband bottom height because it has been observed to exhibit less variability during stratiform precipitation events. If the standard deviation of the brightband top height is less than 500 m, then there is adequate horizontal homogeneity of a high reflectivity layer common in stratiform rain. Otherwise, the BBID will not report a calculated brightband top or bottom height for that particular volume scan.

d. Spatial and temporal averaging

If a bright band is found to exist according to the aforementioned criteria, then all radial-averaged brightband heights are averaged together to produce a single brightband top and bottom height for each volume scan. Otherwise, no brightband height results are reported. The final step of BBID involves temporally averaging the previous 30 min of brightband top and bottom heights that were reported by the algorithm. Initial studies (not shown) on occasion have indicated aberrant behavior of the BBID results from volume scan to volume scan requiring the heights to be smoothed throughout a 30-min time window. If a given radar is operating in VCP-11, then the maximum number of brightband heights that can be reported and thus averaged is six. The spatial averaging step only utilizes brightband heights that were found to exist according to the criteria outlined in section 2c. Results are output to a time series file. Some of these results are shown in the next section. As will be demonstrated in section 4, the heights are also used to generate a graphical product showing where radar may be sampling reflectivity that is contaminated by melting hydrometeors.

3. Evaluation of BBID results

a. Comparison with nearby radiosonde data

The accuracy of BBID results is assessed using the 0°C isotherm derived from nearby radiosonde observations. The height of the 0°C isotherm indicates where melting begins for falling hydrometeors in stratiform precipitation; thus, its height should correspond to the top of the brightband layer. Time series plots of computed brightband top heights and radiosonde-derived 0°C heights are shown for several stratiform rainfall cases in Arizona and in the southern plains. BBID results from four of these events are evaluated at the Phoenix, Arizona, WSR-88D radar site (KIWA). The remaining BBID results are examined at the Little Rock, Arkansas; Dodge City, Kansas; Oklahoma City, Oklahoma; and Tulsa, Oklahoma WSR-88D radars (KLZK, KDDC, KTLX, and KINX, respectively). Table 1 lists event start and end times for all cases as well as the radiosonde site used for verification.

The BBID computes brightband top and bottom heights using reflectivity within 10–30 km of the WSR-88D. These heights can be extrapolated beyond these ranges (or interpolated between adjacent radars) when the melting layer possesses horizontal homogeneity. Results from the Arizona cool-season, stratiform precipitation events are presented in Fig. 2. The nearest radiosonde observation taken below the melting layer was at the Tucson sounding location, which is approximately 140 km to the southeast of the KIWA radar site. These cases test the ranges at which brightband observations from a nearby radar can be extrapolated in space. The brightband top heights diverge from the radiosonde-observed melting levels as time increases beyond the synoptic observation times (1200 and 0000 UTC). These times are noted by the discontinuities in the radiosonde 0°C heights (lines in gray) that have been interpolated between sounding observation times. Figures 2a,b,d reveal melting-level heights derived from recently observed 0°C heights and from the BBID differ by less than 250 m. Figure 2c shows the BBID algorithm does not detect a bright band at the same time in which a radiosonde observation is made. Several horizontal gaps in reported BBID heights are noted in each event. At least one criterion in brightband existence (see section 2c) is not met in these situations, and no brightband top heights are reported. Relaxation of parameters in brightband existence would lead to more brightband height reports, although their vertical extents would be more uncertain.

Figures 2a,c,d (and to a lesser degree, Fig. 2b) show dropping melting levels well in advance of the times in which a radiosonde observation was made at Tucson. While liquid precipitation fell at the KIWA radar site during all four events, frozen precipitation fell at higher elevations within a 230-km radius of the radar. The dropping melting-layer heights suggest the altitude of the rain–snow line above the radar was decreasing. These heights, when used in conjunction with a digital elevation model, can be used to map the approximate rain–snow line near the radar. Observations of the approximate rain–snow line will be useful for transportation purposes such as snow removal operations, hydrologic modeling, and snowpack monitoring for watershed management and avalanche forecasting purposes. It must be acknowledged, however, that the melting layer may exhibit spatial variability in complex terrain or near the location of frontal features. Thus, the BBID results are most reliable near radar locations.

Results from the BBID for stratiform rainfall cases in the southern plains are presented in Fig. 3. Table 1 shows the WSR-88D sites supplying data to the BBID, event start and end times, and the radiosonde sites that are used for verification. The BBID heights measured near synoptic observation times in Fig. 3 show good agreement between brightband top heights and nearby radiosonde 0°C heights. The BBID time series in Fig. 3b for the KDDC radar reveals much more variability in the temporal evolution of melting-layer heights than seen in Fig. 2. The sounding data from Dodge City, Kansas (DDC), for this case are shown in Fig. 4. The 1200 UTC 2 March 2000 radiosonde observation (Fig. 4a) indicates a 1-km-deep isothermal layer at 0°C extending from approximately 1500 to 2500 m. This is a typical depth of an isothermal layer (Stewart 1984), and the depth is directly proportional to the precipitation rate when atmospheric temperature tendencies from other processes (vertical and horizontal temperature advection, condensation, evaporation, etc.) are relatively small (Kain et al. 2000). The isothermal layer is no longer evident at 0000 UTC 3 March 2000 (Fig. 4b) and the melting layer is just above the surface at an altitude of 1 km. The three periods of precipitation seen in Fig. 3b are believed to be at least partially responsible for the significant cooling that occurs throughout the period. When diabatic heating effects from condensation and evaporation are small, melting-induced cooling can have a significant effect, such as a rapid changeover from rain to snow at the surface (e.g., Kain et al. 2000). The results in Fig. 2b reveal how individual precipitation bands as detected by KDDC and analyzed by the BBID algorithm start out with melting-layer heights near 2500 m (at the top of the isothermal layer) that descend very quickly with time as frozen hydrometeors fall, melt, and diabatically cool the atmosphere. The melting-induced cooling is merely a component of the overall atmospheric cooling, otherwise the brightband heights would have shown decreases with time without any “recovery” of melting-layer heights to higher altitudes.

Figure 3d shows the KINX brightband top heights along with the 0°C heights from both the Norman, Oklahoma (line in black), and Springfield, Missouri (line in gray), radiosondes. While the Springfield upper-air observation site is closer in distance to KINX, the synoptic weather conditions are considerably different. In this case, the Norman radiosonde observation is more representative of the synoptic conditions present near the KINX radar, although comparison with BBID results from KTLX in Fig. 3c suggests the atmospheric cooling that is denoted by the decrease in the brightband top heights to an approximate height of 1500 m precedes the cooling seen in Fig. 3d by approximately 3 h. The timing of the advanced atmospheric cooling is consistent with the west to east storm motion associated with this event and suggests melting is a significant latent heating effect as discussed in Kain et al. (2000). Nonetheless, the Springfield radiosonde observation indicates no melting-induced cooling.

b. Comparison with RUC-2 0°C isotherm heights

Model analyses from the RUC-2 (Benjamin et al. 1998) were obtained for comparison with output from the BBID algorithm. The height of the 0°C isotherm at the model grid point nearest to the radar location is compared to brightband top heights reported from the BBID. A stratiform precipitation event is examined at four radar locations in the coastal Carolinas region. The case start and end times as well as the radar sites used in the comparisons are reported in Table 1. The temporal resolution of the model is hourly, while the BBID output variables are produced every 5–6 min.

The time–height cross sections in Fig. 5 reveal overall good agreement between trends in RUC-derived melting-layer heights and those derived from the BBID. In comparison with the BBID results in Arizona and in the plains (Figs. 2, 3), there is a propensity for melting-layer heights to increase with time from this limited dataset in the Carolinas region. Figures 5b and 5c show BBID trends that have good correlations with the RUC-2 time series, but with some noted scatter. It is believed that more stringent criteria in the brightband existence function (see sections 2c and 5) could have been used in this situation to eliminate some of these points with relatively high height uncertainties. Figure 5a shows that radar-derived melting-layer heights increase near 0300 UTC 7 February 2002 prior to the increase indicated by the model. Similarly, the rapid decrease in melting-layer heights is observed shortly after 0000 UTC 8 February 2002 using WSR-88D data, several hours prior to the decrease denoted in the RUC-2 trends. The high temporal resolution of the WSR-88D measurements can detect atmospheric thermodynamic changes before they are reflected in numerical prediction model analyses. Figure 5d reveals a sudden 1.5-km jump in RUC-2 0°C heights at 1200 UTC 7 February 2002. The BBID trends indicate a much more gradual increase in heights during this time period. It is believed that the RUC-2 heights are responding to observations at these synoptic times and that the trend had not been captured in the earlier model analyses. It has been suggested that radar-derived 0°C heights may be assimilated in numerical models by Sanchez-Diezma et al. (2000). It is believed that model analyses can now be improved following the assimilation of radar-observed melting-layer heights.

c. Comparison with a vertically pointing radar

During the winter months of 2000 and 2001, the National Severe Storms Laboratory employed a vertically pointing 10-cm wavelength radar in complex terrain in Arizona. The profiling radar collected a full spectrum of Doppler radar data at 2-min intervals. Operating parameters were set to measure reflectivity aloft at range gate spacings of 43 m. A description of the operating characteristics of this mobile radar is reported in Fig. 6. The operation of the vertically pointing radar provides for the comparison of high-resolution reflectivity observations from an independent sensor with BBID algorithm results using WSR-88D reflectivity data. The mobile radar was located 141 km to the north of the Phoenix WSR-88D and at an elevation approximately 500 m higher. The situation of the vertically pointing radar in a prominent valley invites the investigation of the horizontal homogeneity of the melting layer in complex terrain. Questions regarding the accuracy and usefulness of BBID calculations from a distant WSR-88D in complex terrain can begin to be addressed.

Figure 7 illustrates a reproduction of BBID algorithm brightband top and bottom heights overlaid on a time–height trend of reflectivity from the independent radar. The event start and end times are reported in Table 1. A visual inspection shows the BBID-calculated heights nearly encompass the melting layer observed by the vertically pointing radar. Reflectivity from the profiling radar indicates the brightband depth is thinner than indicated by the BBID algorithm operating on WSR-88D data. The two sensors have very different vertical resolutions where the vertically pointing radar observes more detail in the reflectivity profiles. The results from this brief analysis indicate that the reflectivity reduction parameter used in the BBID algorithm (see section 2b) is set at a value (20%) that is too large and depicts the melting layer as being too thick as compared to measurements from the vertically pointing radar. The sensitivity of the minimum reduction parameter is explored in detail in section 5.

More importantly, the BBID shows the trend of decreasing brightband heights at the same time the melting layer drops in height. The decrease in altitude of the melting layer indicates a lowering of the rain–snow line. Examination of more cases like these will determine the spatial extent at which a rain–snow line determined from the BBID operating on KIWA radar data apply to surrounding regions.

4. Identification of contaminated reflectivity in radar precipitation maps

The following section demonstrates how the bright band may be automatically identified with real-time radar reflectivity data and used to identify areas on precipitation rate maps that may have artificially high reflectivity due to brightband contamination. Data collected on the hybrid scan for a given WSR-88D are used to formulate precipitation estimates (Fulton et al. 1998). The hybrid scan lookup table is typically an array of prescribed elevation angles to be used for QPE at each range and azimuth bin. These elevation angles and ranges are converted to beam center heights for the BBID assuming the vertical gradient of atmospheric refractive index is constant [4/3 assumption; see Doviak and Zrnic (1984)]. These beam heights are then interrogated in BBID to see if they are lower than the brightband top and higher than the brightband bottom. If a particular bin is determined to fall within the bright band, then it is flagged as contaminated for that volume scan. A graphical product is output that shows the spatial distribution of bins that have been identified as being potentially contaminated by artificially high reflectivity in the melting layer.

Figure 8 shows the instantaneous precipitation rate from the KIWA radar at 1724 UTC 6 March 2000. The regions encompassed by the overlaid black lines indicate where the hybrid scan samples reflectivity within the BBID brightband layer. There is a positive correlation between the areas of relatively high accumulations and regions with potential brightband contamination as computed from the BBID. However, it should be noted that the depth of the bright band was overestimated in this case as discussed in section 3c and depicted in Fig. 7; thus, the areal extent that has been identified as being contaminated by the bright band is overestimated as well.

The ringed appearance of high echo returns around the radar suggests the melting layer has been sampled at a nearly constant height (and thus range). The inner, concentric ring is a consequence of the hybrid scan utilizing data from the 1.45° elevation angle and, thus, sampling the bright band at close range before encountering it again at farther range with the 0.5° elevation angle. The terrain-based hybrid scan (O'Bannon 1997) utilizes reflectivity measured closest to the ground as long as the bottom of the radar beam clears the underlying terrain by at least 150 m and is 50% or less blocked from intervening terrain features (see discussion in Maddox et al. 2002). The hybrid scan lookup table for KIWA is shown in Fig. 9. At ranges near the radar, the former criterion is not met, which forces the hybrid scan to utilize reflectivity from higher tilts (Fig. 9). Thus, the bright band resides at a near-constant height but is sampled by the second tilt at a close range and then again by the first tilt at a farther range. Figure 8 demonstrates how the BBID can be used as a real-time application to identify and map regions where radar-derived precipitation accumulations may be in error.

5. Sensitivity of parameters

Several adaptable parameters influence the accuracy and probability of detection capabilities of the BBID. The following section demonstrates how the algorithm is optimized at a particular radar site using archived WSR-88D data. The parameter sensitivity analysis will also show the ranges the parameters may be set to in order to yield acceptable results and, thus, elucidate how far the optimized parameter settings deviate from their default values described in section 2. The BBID utilizes adaptable parameters to determine the brightband top and bottom heights as well as the confidence in each brightband detection by analysis of the spatial coverage and the degree of homogeneity of the brightband top height. The parameter that dictates the thickness of the bright band is the percent reduction in reflectivity above and below the column maximum, which was set to 20% for the results shown previously. A brightband height for the particular volume scan is reported only in the event that at least 6% of the valid grid columns (those that do not have blockages in the third tilt or higher) have maximum reflectivity values greater than 30 dBZ, and the standard deviation of the brightband top must be less than 500 m. While the default values for these parameters were optimized subjectively for the cases shown, a detailed parameter sensitivity test was performed for the KIWA radar site.

Over 1200 volume scans of reflectivity collected during cool season, stratiform rainfall events at the KIWA WSR-88D site were used to evaluate the sensitivity of the following parameters: the 30-dBZ minimum reflectivity threshold, the 20% minimum reduction threshold in reflectivity above and below the column maximum, the 6% minimum threshold of valid grid columns exceeding the minimum reflectivity threshold, and the 500-m maximum threshold of the standard deviation of the brightband top heights. Using the entire calibration dataset, each grid column was searched for its maximum reflectivity value between ranges of 10 and 30 km. Figure 10 shows the frequency of these composite reflectivity values. There is a notable peak in the frequencies between 24 and 36 dBZ suggesting the predominance of contamination from hydrometeors sampled in the melting layer. The magnitude of the minimum reflectivity parameter coupled with the value used in the percent reduction in reflectivity above and below the column maximum determines the brightband heights and thus depths.

Figure 11 shows the average brightband depths computed from the BBID versus different percent reduction parameter settings for several minimum reflectivity thresholds between 24 and 36 dBZ. The computed depths are relatively insensitive to the differing minimum reflectivity settings, but quite sensitive to the percent reduction threshold. Thinner (thicker) brightband depths will result when lower (higher) settings of the reflectivity reduction parameter are used. The steepest slope of the curves and thus greatest sensitivity lies between percent reduction thresholds of approximately 15% and 30%. Four years of melting-layer observations reported by Fabry and Zawadzki (1995) indicate typical bright bands in stratiform rain are approximately 500 m thick. Figure 11 indicates the BBID is capable of detecting brightband depths that better simulate observations by using a percent reduction threshold of 15% in lieu of the 20% default value. This parameter adjustment will result in thinner brightband depths leading to better agreement with observations from the vertically pointing radar (Fig. 7). The accuracy of the brightband heights is relatively insensitive to the minimum reflectivity threshold but should be set between 24 and 36 dBZ (Fig. 10).

The following parameter optimization study uses a fixed percent reduction threshold of 15% and the minimum reflectivity threshold set to 30 dBZ for the 1200 volume scans of WSR-88D data and evaluates the relationship between the two criteria that control the brightband existence (see section 2c): the percent of valid grid columns that have a composite reflectivity value greater than 30 dBZ (i.e., the spatial coverage of significant reflectivity) and the standard deviation of the brightband top height (i.e., the spatial homogeneity of the bright band). The two parameters must be considered in tandem, as the former criterion alone can be met in the presence of widespread convective echo lacking a true brightband feature. The latter criterion ensures that the echo being sampled has spatial homogeneity, or is layered, which is common in stratiform precipitation with brightband contamination.

Figure 12 shows a scatterplot of the percent coverage parameter versus the standard deviation of the brightband top height for the calibration dataset. A high (low) threshold of the standard deviation parameter coupled with a low (high) percent coverage threshold will lead to a high (low) probability of detection of the bright band but a high (low) false alarm rate. A compromise in the two parameter settings is thus sought out in order to maximize the skill of the brightband detection capabilities of the algorithm. There are two clusters of data occurrences worth noting (Fig. 12). The first grouping has a high percent coverage (>80%) combined with a relatively low standard deviation (<250 m). These data points correspond to widespread brightband occurrences that cover most of the monitoring region around the radar from 10 to 30 km. The other grouping occurs at low percent coverage values (<10%) with standard deviations ranging from 0 to 1000 m. The high density of data in this parameter space corresponds to the majority of the volume scans that are composed of little reflectivity and are devoid of a bright band. The two parameters are optimized to screen out data from this region of the parameter space with a minimum percent coverage parameter threshold of 10% and a maximum standard deviation of the brightband top threshold set to 500 m.

6. Conclusions

A method of detecting the heights and depths of the melting layer using full-resolution WSR-88D reflectivity data has been presented. Results indicate the automated technique captures meteorologically realistic evolutions of the bright band at several locations in the United States. Cases are presented for diverse geographic regions including the complex terrain of Arizona, the southern plains, and the coastal Carolinas region. Time–height cross sections of BBID output are compared to 0°C heights from nearby radiosonde data as well as collocated RUC-2 grid points. The trends show good agreement with in situ observations and model analyses. In fact, it is suggested that the BBID algorithm can offer improvements in derived melting-layer heights as compared to the verification sources. Time series of radar-derived brightband heights capture increases and decreases of the melting-layer height well before these changes are represented in RUC-2 model analyses or radiosonde observations in some of the cases shown. Most of these improvements can be attributed to the relatively high temporal resolution associated with the collection of data from the WSR-88D. This study suggests the BBID algorithm can provide accurate, real-time melting-layer heights that could be potentially used in many ways, for example, assimilated in atmospheric numerical prediction models. This information may also be used in conjunction with a digital elevation model to map regions receiving frozen versus liquid precipitation. Real-time identification of precipitation phase may be useful for transportation purposes as well as for hydrologic applications.

The brightband identification algorithm operates on full-resolution WSR-88D reflectivity data gathered within 10–30 km of the radar. The melting layer may exhibit substantial temporal and spatial variability in complex terrain and in the presence of frontal features. This variability may pose a significant limit on the range from radar at which the BBID results can be applied. To begin addressing the consistency of the melting layer in complex terrain, BBID results are compared with reflectivity observations from an independent, vertically pointing radar located 141 km away in a prominent valley. Results from this particular case show good agreement between the evolution of the bright band with time. This brief assessment shows promise for BBID results to be used for domain-wide rain–snow mapping. To help to accomplish this goal, the BBID output may need to be combined with spatially variable 0°C heights from a numerical model. Many more cases like this one are being examined and invite future research.

Results from a case study presented in section 4 show the BBID can be used to graphically display where radar-derived rainfall estimates may be in error due to brightband contamination. It is suggested that sophisticated precipitation algorithms of the present and near future utilize such information in their estimation schemes. A parameter sensitivity study was performed on the four main parameters used in the BBID in section 5. Over 1200 volume scans of reflectivity data from the Phoenix, Arizona, WSR-88D were used to optimize the parameters. Although the sensitivity study revealed parameter settings were optimized at values very close to the subjectively chosen default ones, optimization strategies may need to be performed to improve the algorithm performance when operating on different radars and in different geographic locations.

Future improvements to the BBID will involve retaining the azimuthal variability of the bright band. In its current configuration, the BBID computes brightband heights using reflectivity profiles at a maximum of 7200 grid points but averages each height in the radial and azimuthal directions. Sloping bright bands such as those that are common near fronts or in complex terrain may be resolvable by utilizing brightband heights at different locations around the radar. The BBID algorithm has been designed to reduce its false-alarm rate; however, there are instances in which the algorithm can mistakenly trigger on precipitation that has horizontal homogeneity but lacks a true brightband signature. Future improvements will include continued assimilation of RUC-2 0°C heights to impose meteorologically realistic bounds on BBID output. We also will explore the presentation of BBID output in a probabilistic sense so that the uncertainty in the brightband detections and the range of brightband height estimates are conveyed to the users.

Acknowledgments

The authors thank and recognize Ken Howard, Bob Maddox, and Jian Zhang for their useful suggestions and recommendations provided throughout this study. Jack Kain provided reviews of this article that greatly enhanced its readability. The S-Band Precipitating-Cloud Radar was provided by Herb Winston from the URS/Radian Corporation (now Vaisala) and was operated by Richard Becker of the Verde Natural Resource Conservation District. Funding for this research was provided under NOAA-OU Cooperative Agreement NA17RJ1227; the Radar Operations Center in Norman, Oklahoma; as well as the Salt River Project in Phoenix, Arizona. Comments from three anonymous reviewers improved the quality of this manuscript.

REFERENCES

  • Andrieu, H., and Creutin J. D. , 1995: Identification of vertical profiles of radar reflectivities for hydrological applications using an inverse method. Part I: Formulation. J. Appl. Meteor., 34 , 225239.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Austin, P. M., and Bemis A. , 1950: A quantitative study of the bright band in radar precipitation echoes. J. Meteor., 7 , 165171.

  • Baeck, M. L., and Smith J. A. , 1998: Rainfall estimation by the WSR-88D for heavy rainfall events. Wea. Forecasting, 13 , 416436.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Benjamin, S. G., Brown J. M. , Brundage K. J. , Schwartz B. , Smirnova T. , and Smith T. L. , 1998: The operational RUC-2. Preprints, 16th Conf. on Weather Analysis and Forecasting, Phoenix, AZ, Amer. Meteor. Soc., 249–252.

    • Search Google Scholar
    • Export Citation
  • Borga, M., Anagnostou E. N. , and Krajewski W. F. , 1997: A simulation approach for validation of a brightband correction method. J. Appl. Meteor., 36 , 15071518.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Cheng, M., and Collier C. G. , 1993: An objective method for recognizing and partially correcting brightband error in radar images. J. Appl. Meteor., 32 , 11421149.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Doviak, R. J., and Zrnic D. S. , 1984: Doppler Radar and Weather Observations. Academic Press, 593 pp.

  • Fabry, F., and Zawadzki I. , 1995: Long-term radar observations of the melting layer of precipitation and their interpretation. J. Atmos. Sci., 52 , 838851.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Fulton, R. A., Breidenbach J. P. , Seo D. J. , Miller D. A. , and O'Bannon T. , 1998: The WSR-88D rainfall algorithm. Wea. Forecasting, 13 , 377395.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gourley, J. J., Maddox R. A. , Burgess D. W. , and Howard K. W. , 2002: An exploratory multisensor technique for quantitative estimation of stratiform rainfall. J. Hydrometeor., 3 , 166180.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Harrold, T. W., and Kitchingman P. G. , 1975: Measurement of surface rainfall using radar when the beam intersects the melting layer. Preprints, 16th Conf. on Radar Meteorology, Houston, TX, Amer. Meteor. Soc., 473–478.

    • Search Google Scholar
    • Export Citation
  • Howard, K. W., Gourley J. J. , and Maddox R. A. , 1997: Uncertainties in WSR-88D measurements and their impacts on monitoring life cycles. Wea. Forecasting, 12 , 166174.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kain, J. S., Goss S. M. , and Baldwin M. E. , 2000: The melting effect as a factor in precipitation-type forecasting. Wea. Forecasting, 15 , 700714.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kitchen, M., and Jackson P. M. , 1993: Weather radar performance at long range—Simulated and observed. J. Appl. Meteor., 32 , 975985.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Maddox, R. A., Zaras D. S. , MacKeen P. L. , Gourley J. J. , Rabin R. , and Howard K. W. , 1999: Echo height measurements with the WSR-88D: Use of data from one versus two radars. Wea. Forecasting, 14 , 9941006.

    • Search Google Scholar
    • Export Citation
  • Maddox, R. A., Zhang J. , Gourley J. J. , and Howard K. W. , 2002: Weather radar coverage over the contiguous United States. Wea. Forecasting, 17 , 927934.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Marwitz, J., and Toth J. , 1993: The Front Range blizzard of 1990. Part I: Synoptic and mesoscale structure. Mon. Wea. Rev., 121 , 402415.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • O'Bannon, T., 1997: Using a “terrain-based” hybrid scan to improve WSR-88D precipitation estimates. Preprints, 28th Conf. on Radar Meteorology, Austin, TX, Amer. Meteor. Soc., 506–507.

    • Search Google Scholar
    • Export Citation
  • OFCM, 1991: Doppler radar meteorological observations. Part C: WSR-88D products and algorithms. Federal Meteorological Handbook, FCH-H11C-1991, Office of the Federal Coordinator for Meteorological Services and Supporting Research, Rockville, MD, 210 pp.

    • Search Google Scholar
    • Export Citation
  • Sanchez-Diezma, R., Zawadzki I. , and Sempere-Torres D. , 2000: Identification of the bright band through the analysis of volumetric radar data. J. Geophys. Res., 105 (D2) 22252236.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Seo, D. J., Breidenbach J. , Fulton R. , and Miller D. , 2000: Real-time adjustment of range-dependent biases in WSR-88D rainfall estimates due to nonuniform vertical profile of reflectivity. J. Hydrometeor., 1 , 222240.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Smith, C. J., 1986: The reduction of errors caused by bright bands in quantitative rainfall measurements made using radar. J. Atmos. Oceanic Technol., 3 , 129141.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Stewart, R. E., 1984: Deep 0°C isothermal layers within precipitation bands over southern Ontario. J. Geophys. Res., 89 , 25672572.

  • Vicente, G. A., Scofield R. A. , and Menzel W. P. , 1998: The operational GOES infrared rainfall estimation technique. Bull. Amer. Meteor. Soc., 79 , 18831898.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Westrick, K. J., Mass C. F. , and Colle B. A. , 1999: The limitations of the WSR-88D radar network for quantitative precipitation measurement over the coastal western United States. Bull. Amer. Meteor. Soc., 80 , 22892298.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • White, A. B., Gottas D. J. , Strem E. T. , Ralph F. M. , and Neiman P. J. , 2002: An automated brightband height detection algorithm for use with Doppler radar spectral moments. J. Atmos. Oceanic Technol., 19 , 687697.

    • Crossref
    • Search Google Scholar
    • Export Citation

Fig. 1.
Fig. 1.

Vertical cross section of radar beams in VCP-11. The region with gray horizontal lines represents a theoretical melting layer, with radar-derived brightband top and bottom heights shown in black

Citation: Weather and Forecasting 18, 4; 10.1175/1520-0434(2003)018<0585:ADOTBB>2.0.CO;2

Fig. 2.
Fig. 2.

(a) Time series of brightband top heights (in black) from the KIWA radar and 0°C heights (in gray) measured independently by the Tucson, AZ, radiosonde from 0804 to 1754 UTC 4 Feb 1998. (b) Same as in (a) but from 1700 UTC 8 Feb to 1200 UTC 9 Feb 1998. (c) Same as in (a) but from 0604 to 1206 UTC 15 Feb 1998. (d) Same as in (a) but from 0317 UTC 6 Mar 2000 to 0006 UTC 7 Mar 2000.

Citation: Weather and Forecasting 18, 4; 10.1175/1520-0434(2003)018<0585:ADOTBB>2.0.CO;2

Fig. 3.
Fig. 3.

(a) Time series of brightband top heights (in black) from the KLZK radar and 0°C heights (in gray) measured independently by the Little Rock, AR, radiosonde from 0506 to 1200 UTC 8 Jan 2000. (b) Same as (a) for the KDDC radar measured independently by the Dodge City, KS, radiosonde from 1013 to 1924 UTC 2 Mar 2000. (c) Same as (a) for the KTLX radar measured independently by the Norman, OK, radiosonde from 0508 to 1229 UTC 29 Jan 2001. (d) Same as (a) for the KINX radar measured independently by the Springfield, MO, radiosonde (line in gray) and by the Norman, OK, radiosonde (line in black) from 0148 to 1534 UTC 29 Jan 2001

Citation: Weather and Forecasting 18, 4; 10.1175/1520-0434(2003)018<0585:ADOTBB>2.0.CO;2

Fig. 4.
Fig. 4.

(a) Observed sounding from DDC plotted on a standard skewT–logp diagram, valid 1200 UTC 2 Mar 2000. (b) Same as (a), but valid 0000 UTC 3 Mar 2000.

Citation: Weather and Forecasting 18, 4; 10.1175/1520-0434(2003)018<0585:ADOTBB>2.0.CO;2

Fig. 5.
Fig. 5.

(a) Time series of brightband top heights (in black) from the KCAE radar and 0°C heights (in gray) derived independently from the nearest RUC-2 model grid point from 2100 UTC 6 Feb to 1200 UTC 8 Feb 2002. (b) Same as (a) for the KAKQ radar and derived independently from the nearest RUC-2 model grid point from 1500 UTC 7 Feb to 0000 UTC 8 Feb 2002. (c) Same as (a) for the KLTX radar and derived independently from the nearest RUC-2 model grid point from 0300 UTC 7 Feb to 0900 UTC 8 Feb 2002. (d) Same as (a) for the KRAX radar and derived independently from the nearest RUC-2 model grid point from 0600 UTC 7 Feb 2002 to 0000 UTC 8 Feb 2002

Citation: Weather and Forecasting 18, 4; 10.1175/1520-0434(2003)018<0585:ADOTBB>2.0.CO;2

Fig. 6.
Fig. 6.

An S-Band Precipitating-Cloud Radar System image and specifications

Citation: Weather and Forecasting 18, 4; 10.1175/1520-0434(2003)018<0585:ADOTBB>2.0.CO;2

Fig. 7.
Fig. 7.

Time series of brightband top and bottom heights (in black) from the KIWA radar overlaid on reflectivity observations from an independent vertically pointing radar located 141 km away for 24-h period beginning at 0000 UTC 6 Mar 2000

Citation: Weather and Forecasting 18, 4; 10.1175/1520-0434(2003)018<0585:ADOTBB>2.0.CO;2

Fig. 8.
Fig. 8.

Precipitation rate (mm h−1) from the KIWA radar with brightband identification areas overlaid (in black) at 1724 UTC 6 Mar 2000

Citation: Weather and Forecasting 18, 4; 10.1175/1520-0434(2003)018<0585:ADOTBB>2.0.CO;2

Fig. 9.
Fig. 9.

The terrain-based hybrid scan lookup table used at the KIWA radar site. The tilt numbers (shaded in grayscale) correspond to the elevation angles (0.5°–3.35°) used by the WSR-88D for precipitation estimation. The maximum range is 230 km

Citation: Weather and Forecasting 18, 4; 10.1175/1520-0434(2003)018<0585:ADOTBB>2.0.CO;2

Fig. 10.
Fig. 10.

Frequency diagram for the maximum reflectivity values in each grid column (the composite reflectivity) for the calibration dataset

Citation: Weather and Forecasting 18, 4; 10.1175/1520-0434(2003)018<0585:ADOTBB>2.0.CO;2

Fig. 11.
Fig. 11.

Sensitivity of brightband depth calculations to the reflectivity reduction parameter and minimum reflectivity thresholds ranging from 24 to 36 dBZ

Citation: Weather and Forecasting 18, 4; 10.1175/1520-0434(2003)018<0585:ADOTBB>2.0.CO;2

Fig. 12.
Fig. 12.

Scatterplot of the percent coverage parameter vs the standard deviation of the brightband top parameter for the calibration dataset. Each square has been shaded (in grayscale) to indicate the number of occurrences of data points falling in that area

Citation: Weather and Forecasting 18, 4; 10.1175/1520-0434(2003)018<0585:ADOTBB>2.0.CO;2

Table 1.

WSR-88D data utilized in the brightband identification algorithm and sources used for verification

Table 1.
Save