Calibration Proxy for RICO Ka

S-PolKa RICO, Codrington, Barbuda
Nov 2004 - Jan 2005

Introduction       [Tower Plots]

The S-Pol Ka-band calibration for RICO is uncertain, but it is fairly certain that the calibration varies with time. Noise power has been examined previously to establish variations in receiver gain over time (spot checks several times per operation day), but no two-way (transmit/receive) calibration information is readily available. A decision was therefore made to examine power return from a nearby free-standing radio tower, and determine if such signal would serve as a calibration proxy.

Results from the tower proxy are presented. The tower seems to serve as some sort of indicator of Ka calibration stability, but results indicate that the Ka tower return varies erractically over time periods as short as three minutes. Excursions in the tower return can be as small as +/- 1 dB for multi-hour periods, but can be greater at times (on the order of +/- 2 to 3 dB), and often show sudden, large negative excurions that may last for a few data points, or only a single point. The excursions in the horizontal power do not often match excursions in vertical power, and expected changes in Ka Zdr during periods of excursions are not always evident (ZDR_VH_K rarely shows aggregate changes of 3 dB during periods of noisy tower cals; see data of 11-Jan-05, near 13:20 for a case of changing ZDR_VH_K). The issue to resolve is whether the tower returns serve as a good indicatior of moment-to-moment Ka calibration.

At a minimum, it seems that a steady run of maximum returned tower power can be used to track Ka calibration stability on a day-to-day, or perhaps an hour-to-hour basis. Additionally, if the S-band reflectivity is considered stable throughout the project, and we can determine Ka/S-band reflectivity offset for a few stable periods of tower return, we can then directly relate the changing tower return to a modified Ka/S-band reflectivity offset for large periods of the project, without doing daily cloud echo comparisions. [Note: proper consideration must be given to discontinuities in the system, such as the magnetron change out of 14-Dec, and the code change of 26-Dec that attempt to re-adjust the Ka HH/VV gate match.]

Methodology

There was a single, highly visible tower located on Barbuda at a distance of about 2.40 km, azimuth 14.4° from the radar. Since there is some blockage by shrubs at the lowest routine scan elevation, it was decided to use either the 1.3° or 1.5° elevation scan (these scans are mixed in the plots, with no indication of when which angle was used). The tower was routinely visible at these tilts when the S-PolKa PPI sector was centered to the north, a condition that existed for about 2/3 of the RICO project.

There are obvious issues associated with ensuring that estimates of tower return were consistent with time. Since S-Pol does not do beam "indexing", and beam spacing was on the order of 0.8°, it might seem that a high degree of variability would be introduced in the tower return estimate. This is not the case: the Ka system is extremely insensitive to ground clutter, with Ka clutter about 40 dB below that seen at S-band; the tower signal did not saturate the Ka receiver (actually, any single pulse may have saturated, but we do not have time series data available to analyze this!), and antenna sidelobes were not a concern; additionally, the Ka receiver performs pulse averaging in linear (as opposed to logarithmic) space. Under these circumstances, as long as all the energy returned from the tower is properly and completely summed, results should be consistent. [A further note is available.]

Meteorological phenomena will still introduce inconsistencies, as will changes in refractive index, alterations to the tower, or, perhaps, small inconsistencies in elevation angle. An important consideration is whether the Ka tx/rx dish is wet.

Summing of energy for the six or seven radar beams surrounding the tower was done for every available 1.3°/1.5° scan for all periods of Ka operations. Special care was taken to center the set of six beams so the energy of the end beams was at least 20 dB below the peak energy, and preferably 30 dB below. The 30dB requirement was met most of the time; the 20dB condition seemed sufficient to remove any appreciable variation in the estimate, and points were not excluded based on the less restrictive condition. Typically, tower scans were available every three minutes.

Tower Data Plots

Results are presented in graphical form (tables will be made available at a later time).

Please especially note that plots are provided separately for the single gate at 2.25 km, and for the gate at 2.40 km. Be sure and read the legends on the plots!

Parameters included in the plots are:

• S-dBZ S-band HH reflectivity, provided as a "control"
• ZG0_HH_K Ka HH reflectivity, adjusted using gate-zero transmit power estimate
• Z_HH_K Ka HH reflectivity, using assumed, project-constant transmit power
• P_HH_K Ka HH received power
• P_VV_K Ka VV received power

Plots are presented for the following combination of variables:

• All (S-dBZ, ZG0_HH_K, Z_HH_K, P_HH_K, P_VV_K)
• Ka power fields
• Ka reflectivity estimates (sometimes with the S-band reflectivity)

Note that S-dBZ is included as a "control", and serves to indicate the general stability we should expect from the tower power/reflectivity estimators. S-dBZ is not equivalent to the Ka parameters, however, in that the S-dBZ is highly affected by sidelobes, and there are pulses included in the integrations that have likely saturated the S-band receiver (receiver saturation would tend to stabilize, or limit, an estimate).

Results

As usual, all results are preliminary.

• The tower calibration proxy technique seems to have a repeatability/reliability of better than +/- 1 dB for all quantities under "good" circumstances. Any variability greater than this is considered to be an indicator of potentially suspect data (implying either that the Ka calibration drifted, or the tower calibration itself is in error).
• When tower power makes sudden negative excursions, inspection of the individual beam/gate data shows that this is most often due to weak return in the single beam that is most centered on the tower. This is due on occasion to the Ka dish being wet, or to precipitation echo between the Ka and the tower. But all sudden negative excursions are not easily explained.
• Excursions in the Ka HH and Ka VV quantities are largely uncorrelated, except for the upward excursions that are due to precip echoes overlaying the tower. In these upward excursions, the P_VV_K and P_HH_K quantities tend to match to about 1-3 dB, and there is a tendency for any noisy excursions in either polarization to disappear (note: logarithmic scale on axis may also create this impression).
• Excurions in the Ka HH or Ka VV tower return are not regularly matched by expected changes in cloud echo Zdr_VH_K measurements (however, clouds are not present during all times, nor are they conveniently present just beyond the tower; see data of 11-Jan-05, near 13:20 for a case of changing ZDR).
• Ka test pulse values (averaged in the same way as the tower return) do not track the excursions in the Ka HH or VV tower power [Note: there can be many reasons for this, given that the test pulse is injected into the radar system, and is not handled in exactly the same way as a target return; the test pulse signal statistics are also unlike real-world returns, from either clouds or hard targets.]
• The variance of the Ka HH quantities is arguably less than the Ka VV quantites until about 26-Dec. (when processor code was changed).
• After 26-Dec, the Ka HH variability seems to be about two to three times greater than the Ka VV quantity variability.
• Prior to the processor code change, the tower appeared essentially as a single gate in P_VV_K. After the code change the tower position in P_VV more closely matched the two-gate tower in P_HH_K. Note that the 2.25 km gate info for the P_VV_K tower is low-power and not really usable (there's no tower there!) prior to 26-Dec.
• The ZG0 quantity shows an unexplained change on 10-Jan. The return from the tower is very constant, but the ZG0 quantity changes by 4 or 5 dB. Overall, the ZG0 quantitiy is not considered reliable, and should not be used.
• Some of the variance in the determined tower powers is undoubtedly due to variations in antenna rotation rate. During RICO, scanning was highly standardized, but some small variations did occur. Variations in scan rate should affect both P_HH_K and P_VV_K, perhaps equally.
• An issue of concern is Ka radar processor "miraging" (see "Known Problems"). Apparently, an error in radar processor code resulted in some energy from different pulses/beams being assigned to erroneous locations in a rotating data/beam buffer within the processor. The net result is a loss of energy in a real echo, and the creation of "mirage" echo at some offset in azimuth from the real echo. It is not possible to correct for any errors caused by this miraging effect. Mirage echoes of the tower are clearly visible in a large fraction of the scans for certain particular days (but not all days). The impact of the mirage effect on the tower calibration is unknown, with results pending further investigation. However, mirage echo strength tends to be at least 15 dB below the strength of the tower return, making this phenomenom unlikely to explain all the variance in the tower power.
• It is unlikely that the Ka radar is mis-triggering and dropping transmit pulses. If this were happening, the Ka HH and VV should track each other. Inspection of Ka housekeeping also shows that there is essentially no variance in the number of pulses (or "hits") sampled to make up a received beam; the radar virtually always uses 50 hits to compute each beam average.
• We really do not know why there are many of the large negative excursions in the tower power return. We do know that there are likely more miscellaneous (and sometimes random) reasons for negative excursions from the mean than there would be for positive excurions (the single reason for positive excursions seems to be rainfall at the tower).

  Two possibilities for large negative excursions in tower return power:

  • gate shift at one polarization or the other, eliminating tower energy from the gate under consideration (this effect could be operative on only a single polarization)
  • wetting of the bush that is directly between the radar and the tower; or, perhaps, other effects due to the bush.

• There is limited suggestion of an anti-correlation between changes in P_HH_K and P_VV_K. There is no obvious explanation for such an effect (e.g., see plots for 050111).
• There are many more conclusions, but this much should be useful as a start.

There is relatively little information in the literature concerning the use of hard targets, and particularly random towers, for monitoring of radar calibration. Very little of the available information suggests that a tower calibration proxy has more than a few dB of stability. Rinehart (1978) suggests that hard target observations might have a day-to-day repeatability of around 1 dB (when only the maximum signal returns on a given day are considered); Fabry (2004) mentions concerns with phase cancelling of power, and states that the most significant use of hard targets is in monitoring phase for refractive index studies, as opposed to echo strength; Pellarin et. al. (1999) provide an 80% confidence estimate of about 2.5 to 3.0 dB for the inter-monthly repeatability of reflectivities from non-vegetated mountain sides. No literature was found to support the idea that consistent, short term scans of a nearby tower would have the degree of repeatability seen during the "good" portions of Ka operations. On the contrary, there are issues with refractive index changes, scan repeatability (i.e., exact elevation and scan rate), and signal statistics that would argue for fairly noisy short-term observations.

It should be noted that previous studies were for longer wavelength radars, and not for a Ka system that is extremely insensitive to clutter sidelobe effects. Furthermore, the tower used for the RICO study is perhaps the closest thing to an ideal tower: free-standing (no guy-wires and narrow in azimuthal extent), in the radar far-field, but still relatively close to the radar, visible at a reasonable elevation angle, with the radar beam path to the tower covered by vegetation (likely limiting potential beam multi-path problems).

Recommendations for Use of Tower Calibrations

There's lots of information in the tower data. The best use of the information would be in selecting periods of stable Ka operation (either horizontal or vertical polarization). There are often obvious ramp-up periods shortly after system turn-on when Ka power (particularly HH power) has not reached stable operating conditions; those periods should be avoided for most purposes, or Z_HH_K should be compensated appropriately.

Pending further analyses, it seems safe to not be alarmed by periods of large negative excurions in Ka tower power. If in doubt about a specific period, review the parameter ZDR_VH_K for general stability (bad periods should show sudden, large changes in the gross field of ZDR_VH_K). [Note: ZDR_VH_K is a statistically noisy field, due partly to the unfortunate fact that the range for P_VV_K is offset by about one-half gate from the range of P_HH_K; this situation was only marginally improved with the software change of 26-Dec.]

Maximum tower power return can likely be used to adjust Ka relfectivity on a day-to-day basis. Examples of such adjustments will be developed, although there is no current intention to apply daily adjustments to the distributed data set.

Further Work

Further analysis is required/desired. Particularly, it is desired to determine if variations in background noise power correlate with changes in returned power from the tower. Secondarily, any relation to "mirage echoes" should be determined. [Note: both of these issues have been addressed in a preliminary fashion, and insight was incorporated into the "Results" section, above; more formalized work-up is pending.]

In a practical way, it would be good to select a stable period of Ka operation, then relate the Ka reflectivities to the S-band reflectivity through cloud echo analyses. One could then determine an emperical "absolute reflectivity" of the tower. A correction could then be applied to Ka data that would also compensate for any variations in Ka magnetron transmit power.

For comparitive purposes, work can be done with data from other S-Pol projects.

This study is continuing

References

Fabry, F., 2004: Meteorological value of ground target measurements by radar. J. Atmos. Oceanic Technol., 21, 560-573.

Pellarin, T., G. Delrieu, and J.D. Creutin, 1999: Variability of mountain returns during dry-weather conditions: Applications to radar calibration control. Preprints, 29th Conf. on Radar Meteorology, Montreal, QC, Canada, Amer. Meteor. Soc., 772-775.

Rinehart, R.E., 1978: On the use of ground return targets for radar reflectivity factor calibration checks. J. Appl. Meteor.,17, 1342-1350.