Limitations on the Radar Data
Compiled by Sharon Lewis

Index for this report.
Processing and Gridding the Radar Data
Radar Geometry Issues by F. Merceret
Distance from Radar
Missing Radar Data
Cone of Silence
Scan Gaps
Attenuation
Radar Boundary Layer 'noise'
Differences in Radars (74C vs 88D) by J. Dye
Reflectivity Averaging: A Monte Carlo Study by F. Merceret
Why we average dBZ rather than Z by S. Lewis
Radar Issues: A Presentation at the Nov 2002 Workshop by F. Merceret

Towards our goal of a radar based parameter to be used in a Lightning Launch Commit Criteria we must keep in mind the limitations of the radar data we are currently using. This report will try to summarize those limitations and provide links to reports done by other investigators.

The bottom line is: It is not recommended that a single 1x1x1 km cube of radar data be compared with other measurements. This is a temptation when you want to "fly" the aircraft through the volume and compare the radar data to the aircraft in situ data. The work for this project has taken averages (e.g. averaging 11x11x16 km).

 

Processing and Gridding the Radar Data
Return to Top of Page

The scan pattern used for the WSR74C during this program was an iterleved sweep pattern. The radar would make a single sweep (one complete rotation of the radar antenna) at a fixed elevation. The antenna elevation was then increased and another sweep was made. When the antenna reached the maximum elevation it made sweeps as the elevation was decreased. These downward elevations were in between the elevations made going up. The synchronous time steps for the elevation angles were:
0.401 2.000 3.999 7.498 13.002 20.001 25.999 16.002 9.998 4.999 2.999 1.000
This sequence of sweeps (from elevation 0.401 to 25.999 and back down to 1.000) is considered a "volume" of radar data.

The processing used for the WSR74C for this project consisted of transferring the raw (SIGMET format) radar sweep data to universal format (which will be referred to as "uf data") using the TRMM-RSL library. The 3D gridded radar data has been produced by using the MMM-SPRINT program. Since a "volume" as defined earlier (a complete sequence of sweeps) is required for the gridded format, it is customary to also refer to the gridded format as a "volume". (SPRINT was used to grid both the WSR74C and the NEXRAD radar data.)

The error associated with processing the radar data from raw sweeps to gridded format should be considered. F. Merceret investigated the standard deviation for the reflectivity, Z, for the WSR74C to be about 1.1 dB without averaging the bins. He also noted that the 74C averages 4 bins radially. (He did not pursue the result of the averaging.) The exact number is not as important here as the recognition of its existence. This error is compounded by the fact that it is necessary to do all of the processing using a single operating system. Different systems use different algorithms for floating point round off. The error is typically on the order of 2dBZ (LJ Miller personal communication).

Fig 1: This figure shows one example of a single volume that was processed using different machines. These differences are not large, but there is a large tail with a very small fraction having differences as large as 8 to 10 dBZ.

It should not be surprising to find that differences have been found when the processing was done on different machines using different operating systems. Therefore, in the summer of 2003 it was agreed upon to regrid all of the WSR74C data to make sure it was all done on the same machine. At that time we also created additional sets of data with new grid coordinates.

When a radar rule is applied to future work the following uncertainties must be considered:

  • Techniques and software used to process and grid the radar data.
  • Operating system (and its floating point round off algorithm) used.
    • For these reasons it is not recommended to work with a "single grid cube" of radar data (and a single 1x1 column would also be in question). This project is proposing an averaging process (11x11x[top of cloud - 5km] or 21x21x[topof cloud - 5km]). These averages from different machines have shown differences on the order of 0.5 dBZ with a standard deviation of 1 dBZ.

     

    Radar Geometry Issues
    Return to Top of Page

    Radar Geometry Issues (including scan gaps) by F. Merceret

    This paper looks at four issues related to the effects of ground-based radar beam geometry on the measurement of the position and intensity of radar echoes. The issues are:

    Summary from the report:
    Four potential sources of position or intensity error in the radar analysis of weather radar returns were examined. None of them pose a threat of intensity errors exceeding 1 dB except for partial beam filling, which is discussed in detail by Merceret and Ward (2000). Two of them, beam gaps and refractive effects, have the potential to introduce position errors greater than 1 Km. The refractive error may often be the larger of the two and the most difficult to analyze in a particular case. It may be necessary to allow for these errors in the design of lightning launch constraints since eliminating the refractive errors is probably technically impractical and eliminating the scan gaps may be operationally impractical.

     

    Distance from Radar
    Return to Top of Page

    As the system being studied is further away from the radar, the beamwidth and gate spacing increase until the radar values exceed the 1km grid spacing used. Valuable information can still be obtained under these conditions by averaging over larger volumes, but it must be recognized that a "single 1km cube" may yield misleading results.

     

    Missing Radar Data
    Return to Top of Page

    There are two ways a gap could occur in the radar data, both are due to the scan pattern used. They are:

     

    Cone of Silence (No Data Cone)
    Return to Top of Page
    Return to Missing Radar Data

    The following scan pattern plots were taken from F. Merceret's report. ( Radar Geometry Issues ). Note in particular that there is a cone of silence at the zenith of the radar.


    Scan Pattern
    ( Return to Scan Gaps )
    Fig. i : Current operational scan strategy for the PAFB WSR74C (Short, 2000).
    Fig. i +1: WSR-88D VCP 11 (Wheeler, 1997)

    The resulting CAPPIs thereby contain holes or "data gaps". This is a particular concern when the system being studied is immediately over the radar.

    ( Return to Scan Gaps )
    A list of times when the aircraft flew through the cone of silence during the ABFM project have been compiled and are listed in the following reports:
    WSR74C         NEXRAD

     

    Scan Gaps -- This is covered more thoroughly in F. Merceret's report: Radar Geometry Issues
    Return to Top of Page
    Return to Missing Radar Data

    Look again at the plots of the scan patterns (Scan Pattern ) and notice that the further away from the radar the storm is the higher the probability is that a region of the storm will not be scanned. This is particularly evident in the CAPPI shown above by the rings at 10 km. These rings of no data are what we call scan gaps.

    Another example occurred on 010604 when the storm was further away from the radar than most of the storms in the project. In Fig. ii, notice the banded pattern in what should show an extended anvil. In order to make a cross-section of this volume along the anvil it was necessary to rotate the image 25 degrees (Fig. ii+1). The 'X' in Fig. ii marks the point of rotation for Fig. ii+1. The line through 'X' in Fig. ii is the x' = 0 line in the rotated plane of Fig. ii+1.

    The line to the north of the 'X' (Fig. ii), and the line at x'=25 in Fig. ii+1 is where the cross sectional cut was taken (Fig. ii+2).

    Fig.ii+2: Cross section of radar data at x'=25 (fig.ii+1). The line is at z = 11 km to coincide with the CAPPI level shown.

    The cross section shows the rings observed in the CAPPI (Fig ii and Fig ii+1), but also shows the rings that appear (CAPPIs not shown here) at lower levels.

     

    Attenuation
    Return to Top of Page

    Attenuation
     

    Radar Boundary Layer 'noise'
    Return to Top of Page

    Radar Boundary Layer 'noise'

    This is a brief description of a radar return in the boundary layer that appears close to the radar.

     

    Differences in Radars (74C vs 88D)
    Return to Top of Page

    Differences in Radars (74C vs 88D) by J. Dye

    This is documentation of earlier analysis errors in comparing the processing and plotting of the WSR74C vs. the NEXRAD radar data. These have been corrected.

    Differences in plots between the two radars that still exist are due to:

     

    Reflectivity Averaging: A Monte Carlo Study
    Return to Top of Page

    Reflectivity Averaging: A Monte Carlo Study by F. Merceret

    This report presents a comparison of three methods for computing average radar reflectivity in a box using Monte Carlo simulations. The 3 methods are:

     

    Why we average dBZ rather than Z
    Return to Top of Page

    Why we average dBZ rather than Z by S. Lewis

    This discusses (with examples) calculation of an average of dBZ vs and average of Z. It concludes that by using the Z there are more occurrences of "false positive".

     

    Radar Issues: A Presentation at the Nov 2002 Workshop
    Return to Top of Page

    Radar Issues: A Presentation at the Nov 2002 Workshop by F. Merceret

    A review of the radar issues as of the November, 2002, Workshop.

     

    References:

    Merceret, F.J., J.G> Ward (2002): Attenuation of Weather Radar Signals Due to Wetting of the Radome by Rainwater or Incomplete Filling of the Beam Volume, NASA/TM-2002-211171, NASA/YA-D, Kennedy Space Center, FL, 32899, 16 pp.

    Short, D.A.(2000): Final Report on IRIS Product Recommendations, NASA Contractor Report CR-2000-208572, Applied Meteorology Unit, ENSCO, Inc., 1980 N. Atlantic Ave, Cocoa Beach, FL 32931, 26 pp.

    Wheeler, M.M.(1997): Report on the Radar/PIREP Cloud Top Discrepancy Study, NASA Contractor Report CR-204381, Applied Meteorology Unit, ENSCO, Inc., 1980 N. Atlantic Ave, Cocoa Beach, FL 32931, 18 pp.