Report on Polarisation Weather Radar by A. R. Holt & D. H. O. Bebbington
|
|  Yüklə 464 Kb.
|
Experience in the PADRE and DARTH projectsThe PADRE and DARTH projects had polarisation data available from three C-band radars. These were the German DLR radar situated at Oberfpaffenhofen, near Munich (north of the Alps), and the Italian CSIM and SMR radars. These latter are both situated in the Po valley (south of the Alps), and are about 80km apart. The CSIM radar is situated at Teolo, nr Padova, and the SMR radar at St Pietro Capofiume, nr Bologna. The CSIM radar is elevated, being on the crest of a ridge, suffers some obscuration at lower elevations, and has the Alps to the north, whereas the SMR radar is in the plain of the PO, with the Apennines to the south. These three radars were from different manufacturers, and offered different capabilities. The DLR radar was built by EEC in the USA, and was designed to measure a wide range of polarisation parameters, as well as having full Doppler capability. It could measure in any polarisation basis, but for the purposes of the two projects the bases chosen were linear (vertical and horizontal), slant linear (± 45º), and circular (LHC and RHC). There is always a trade-off between the amount of data collected, and the time taken to collect. This was particularly true for the DLR radar, since it recorded data onto a hard disk, and thus data intensive measurement modes meant that the disk was rapidly filled. The data then had to be downloaded onto tape before more data could be collected, and this placed a significant restriction on the data that could be collected. This was particularly the case when both polarisation and Doppler data was being collected. Since the measurement of differential phase when linear polarisation is being used requires Doppler data, it was particularly relevant. It has flexibility in its PRF, and in its range length, and does not have a radome. Its beamwidth is 1.0º. The CSIM radar was an Ericsson radar, from Sweden. It had Doppler capability, and normally used horizontal polarisation. It had, however, the capability also to use vertical polarisation, and hence to provide information on ZDR, as well as on ZH, Doppler velocity, and Doppler Spectral Width. Unfortunately the software was programmed to include an attenuation correction algorithm, which had an in-built upper limit to the correction. Since the limit is reached at close radial range when horizontal polarisation is being used, than when vertical polarisation is used (since horizontal attenuates more than vertical), this causes real problems with ZDR. It is fair to say that that the ZDR images from this radar were consequently not satisfactory. The ZDR must be obtained before any attenuation correction is applied, and this was not possible from a hardware point of view. The authors do not believe that automatic application of any attenuation correction algorithm is possible, since such correction is bound to be dependent on the type of precipitation encountered. The range bins on the radar were 1km. The radar was fitted with a radome, and it had a beam-width of 0.86º The SMR radar, situated near to Bologna, is an Italian radar built by SMI, and installed in the early 1990s. It has Doppler and ZDR capability, but does not have the processing software to extract differential phase. It operates without a radome, with range bins of 250m in Doppler mode. Its beam-width is 0.9º .It had a normal recording cycle over every 15mins of a long range non-Doppler (ZH only) scan, a 4 elevation Doppler (no ZDR) scan, and a 20 elevation Doppler (with ZDR) scan. The 4 elevation Doppler scan does not have the restrictive velocity ambiguity that the 20 elevation scan has. The great advantage that the proximity of the SMR and CSIM radars had for the project was that it enabled some storms to be looked at from different angles simultaneously. This not only gave visual information on possible attenuation, but also other factors such as “double-trip echo”, in which echo from beyond the maximum radar range is recorded as being close to the radar in the next scan. However, the difference in the range bin lengths and beam-widths was a complicating factor. Nevertheless, it was possible to obtain the horizontal wind field from the two operational radars, - their scan patterns were not programmed for the task (ref. 31). T  o illustrate the effects of attenuation at C-band, we give three examples from the PADRE and DARTH projects. The first is a set of four Low level PPIs from the SMR radar at Bologna on 18th June 1997, spaced at 15min intervals. Each figure gives the PPI of both horizontal reflectivity, and differential reflectivity. In figure 1, at 1418hrs, we see two storms, one at 40km distance and approx azimuth 20deg, and the other at 70km and azimuth 285deg. The ZDR image shows a region of negative ZDR at about 75km distance, on the far side of the storm from the radar. It is not clear whether or not this is a hail region. Whereas the ZDR values are negative, they are radially on the far side of a high reflectivity region from the radar. Figure 1. PPI of ZDR and ZH from SMR Bologna radar on 16th June 1997 at 1418hrs. The background colours (pink for ZH and green for ZDR) are for regions of no echo. The distances are in km. Fifteen minute later, the ZDR image shows a strong region of negative ZDR radially between azimuths 20 and 35deg. This is typical of the effects of attenuation, and it is clear that the reflectivity image (labelled ‘Corrected Zxx’) is degraded in this region through the effects of the storm distant 25km from the radar around azimuth 25deg. Echo in the given range is missing beyond 35km range (cf figure 2)
Figure 2. As figure 1, but at 1434hrs A further 15mins later, as seen in figure 3, the image of the storm to the NE of the radar is clearly massively affected by attenuation between azimuths 35 and 75 approximately. This can be seen from the wide swath of negative ZDR bordering this region. It is impossible to tell, what, if any, reflectivity echo should be seen in this region. The region of negative ZDR in the storm to the North-West of the radar has increased in size, but one cannot tell whether it is caused by attenuation or by h  ail. A gust front, or perhaps a convergence line, can be seen in both parameters, as a thin line emanating from the southern end of the NE storm, passing through the clutter close to the radar, and finishing just south of the western end of the storm to the NW. Figure 3. As figure 1, but at 1449hrs. Background colours are in white Finally, at 1504hrs, the storm to the North-west of the radar is displaying a much larger region of negative ZDR, and this appears to have more the appearance of the effects of attenuation, rather than being a large region of hail. The region of strong attenuation to the East, is still very evident, as is the gust front. Note that this image (figure 4) is taken at a higher elevation than is figures 1-3. Subsequently, the storm to the west passed almost over the radar, with the result that the storm to the east appears to have almost dissipated….except that it appears again 15mins later when the other storm is no longer positioned over the radar.  Figure 4: As figure 3, but 1504hrs, and at 1.4deg elevation. The effects of attenuation may not be limited to low elevation scans. It all depends on how far the storm producing the attenuation is from the radar. To illustrate this, we show in figures 5 and 6 scans of ZDR and ZH from the Bologna radar for 23rd June 1992, at 0.44deg and 5.19deg elevation respectively. It will be seen that there is a region of severe reduction in ZDR in both elevations. If anything the effect is greater at the higher elevation angle. It will also be seen in the reflectivity scans that there is a region radially beyond the storm in which the echo appears to have disappeared. In fact the echo here is typically –20dBZ, and the display treats it as if there is no echo. The ZDR values are typically –4dB or lower in this region. There is currently no assured method of reconstructing the true values of ZH in this region. F F To illustrate that this problem does not just occur with one radar, or in one location, we show in figure 7 a section of a PPI from the DLR radar at Oberpfaffenhofen, near Munich. The scans shown are reflectivity (marked Zyy), ZDR, and LDR (Linear Depolarisation Ratio). The Zyy scan shows a region of high reflectivity to the north-east of the radar, with apparently a gap in echo around azimuth 35deg (approx) and beyond 60km. On consulting the ZDR scan, it will be seen that the ZDR values in this region are <-6dB, and that there are negative ZDR values over a wide azimuthal region beyond the storm. Propagation effects are also clearly seen in the LDR scan, where the effect is to increase LDR from the value of around –20dB to –5dB. F  igure 7. Section of a PPI of ZH, ZDR and LDR for the DLR radar on 10th August 1994 at 2157hrs.
The estimation of rainrate from radar reflectivity data is no easy process. The first problem is to differentiate precipitation echoes from those due to other scatterers (such as ground clutter, anaprop, insects or birds). The second is to distinguish between different types of precipitation, such as rain, hail, ice crystals, and melting particles. A third problem is radar calibration. An improperly calibrated radar will give rise to incorrect estimates of rainfall. In addition, if a wavelength is being used which can suffer significant attenuation, such as C-band, the rainrates deduced from regions radially on the far side of a storm from the radar, will be too low. To all this must be added the point that the radar resolution volume increases as the square of its distance from the radar, and its mean height increases with distance from the radar. Consequently it will be sampling at some distance above the ground, and the accuracy of the rain estimates is likely to be better close to the radar than at greater distances. Polarisation radars have the potential to improve rainfall estimation in that in principle they can identify clutter, and anomalous propagation. Further, if they measure reflectivity, differential reflectivity and differential propagation phase, it is possible to calibrate the reflectivity (refs. 53, 54). Several papers have been written on the correction of the effects of attenuation at C- and X-band (eg refs 55,56). They all require information on the three parameters ZH, ZDR and KDP. One of the problems at C-band is that the parameters are quite sensitive to the unknown parameters such as dropsize distribution, drop shape, and temperature. Hence although schemes have been devised to write rainfall rate as a product of powers of these parameters, it is not clear that such schemes could ever be universal, since these unknown parameters will inevitable vary from time to time, and from location to location. Nevertheless, polarisation should certainly lead to an improvement of rainfall measurement, since it should help remove clutter and anaprop. The use of differential propagation phase is probably not reliable at C-band for rainrates less than about 25mm h-1. This is because the phase measurement will have a definite uncertainty associated with it. However, if one is interested in area rainfall, then the integrated phase over a path may well give a useful measure of the rainfall over the area covered by the path. One method of correcting for the effects of attenuation on reflectivity is the ZPHI method (ref. 57) which requires the knowledge of differential propagation phase along the path.
It is clear that at C-band in situations of heavy rain, attenuation can be a major problem. The authors of this report are sceptical that formulae based on combinations of variables will be able to correct readings in real-time. Our belief is that correction algorithms will be dependent on the type of precipitation (including the presence or otherwise of hail). It is hard enough to take past data and to try to correct it to bring agreement with ground-based measurements. It is a very different thing to produce an algorithm which should work in all, or at least in all the extreme, events. Two major goals must be:-
The main operational use of polarisation at present is “dual-linear”, in which the transmitted radiation is switched between linear horizontal and linear vertical polarisation. The received signal is measured in the same polarisation as the transmitted signal. This gives both horizontal reflectivity and differential reflectivity. If differential phase is also needed, then this requires Doppler processing in addition. Though this is simple in concept, it has major drawbacks. There is a hardware requirement of a polarisation switch, which is expensive, and has maintenance implications. An alternative possibility is “slant-linear” in which ± 45º linear polarisation is sent, and the H and V components of the return signal, and their complex correlation, are measured. In fact, only one polarisation needs to be transmitted (either + 45º or – 45º), and hence this removes the need for a polarisation switch. Moreover, Doppler processing is not required to obtain differential phase, which is derived from a single measurement, rather than from a pair of alternate measurements. Though both measurement schemes, “dual-linear” and “slant-linear” are subject to the effects of propagation, it is likely that the “slant-linear” measurement scheme does suffer from extra propagation effects. However, it is being developed in the United States as a possible implementation of polarisation for NEXRAD, and it is being considered in Canada for implementation on their research C-band radar. We believe that CARPE DIEM provides an excellent opportunity for this to be investigated in Europe, on the DLR radar. We suggest that a careful comparison study of the merits of the two systems should be undertaken.
This work has drawn heavily on our experience during the PADRE (EV5V-CT92-0181) and DARTH ( ENV4-CT96-0261) projects. In particular the contributions from our partners Madhu Chandra (DLR), Sandro Nanni and Pier Paolo Alberoni (SMR), and Marco Monai (CSIM) are gratefully acknowledged, as is the contributions to those projects of members of Essex university. |
