Winter precipitation fields in the Southeastern Mediterranean area as seen by the Ku-band spaceborne weather radar and two C-band ground-based radars

Atmospheric Research 119 (2013) 120–130

Contents lists available at ScienceDirect

Atmospheric Research j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / a t m o s

Winter precipitation fields in the Southeastern Mediterranean area as seen by the Ku-band spaceborne weather radar and two C-band ground-based radars M. Gabella a, b,⁎, E. Morin c, R. Notarpietro a, S. Michaelides d a b c d

Dipartimento di Elettronica, Politecnico di Torino, Torino, Italy MeteoSwiss, Locarno Monti, Switzerland Department of Geography, Hebrew University of Jerusalem, Jerusalem, Israel Meteorological Service, Nicosia, Cyprus

a r t i c l e

i n f o

Article history: Received 24 January 2011 Received in revised form 1 May 2011 Accepted 1 June 2011 Keywords: TRMM Precipitation Radar Range-adjustment Weather radar Cyprus Israel

a b s t r a c t The spaceborne weather radar onboard the Tropical Rainfall Measuring Mission (TRMM) satellite can be used to adjust Ground-based Radar (GR) echoes, as a function of the range from the GR site. The adjustment is based on the average linear radar reflectivity in circular rings around the GR site, for both the GR and attenuation-corrected NearSurfZ TRMM Precipitation Radar (TPR) images. In previous studies, it was found that in winter, for the lowest elevation of the Cyprus C-band radar, the GR/TPR equivalent rain rate ratio was decreasing, on average, of approximately 8 dB per decade. In this paper, the same analysis has been applied to another C-band radar in the southeastern Mediterranean area. For the lowest elevation of the “Shacham” radar in Israel, the GR/TPR equivalent rain rate ratio is found to decrease of approximately 6 dB per decade. The average departure at the “reference”, intermediate range is related to the calibration of the GR. The negative slope of the range dependence is considered to be mainly caused by an overshooting problem (increasing sampling volume of the GR with range combined with non-homogeneous beam filling and, on average, a decreasing vertical profile of radar reflectivity). To check this hypothesis, we have compared the same NearSurfZ TPR images versus GR data acquired using the second elevation. We expected these data to be affected more by overshooting, especially at distant ranges: the negative slope of the range dependence was in fact found to be more evident than in the case of the lowest GR elevation for both the Cypriot and Israeli radar. © 2011 Elsevier B.V. All rights reserved.

1. Introduction All ground-based radars have to measure rain from close to long distances from the sensor itself. The radar sampling volume increases with the square of the distance. Since the variability of weather is high in the sampling volume at all ranges, radar echoes are blurred. The systematic component affected by the amount of blurring as well as overshooting with range can be investigated and compensated with a

⁎ Corresponding author at: Dipartimento di Elettronica, Politecnico di Torino, Torino, Italy. E-mail address: [email protected] (M. Gabella). 0169-8095/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.atmosres.2011.06.001

range-adjustment technique. It is known that, on average, the weather signal significantly decreases with height. At longer ranges, the lower part of the sampling volume can be in rain, whereas, the upper part of the same pulse can contain mixed-phase particles or even be without an echo. This overshooting phenomenon at longer ranges is amplified by the decrease of the vertical resolution and the Earth's curvature. These argumentative effects cause an apparent decrease in the sensitivity of the Ground-based Radar (GR) with range: images of cumulative radar-derived rainfall amounts, using large data sets spanning several months or years clearly show unnatural circular features. The reader can refer, for instance, to Kracmar et al. (1999) for one year cumulative amounts in

M. Gabella et al. / Atmospheric Research 119 (2013) 120–130

the Czech Republic, Gabella et al. (2005) for two years cumulative amounts in Switzerland, as well as the works by Vignal and Krajewski (2001) and Nelson et al. (2003). But how does one quantitatively assess the range-dependence? Gabella et al. (2006) proposed to use the first ever spaceborne weather radar onboard the Tropical Rainfall Measuring Mission (TRMM) satellite. Indeed, ground-based and spaceborne sensors provide a complementary view: the GR measures rain from a lateral direction, while the spaceborne radar sees it from the top. On the one hand, the GR measures precipitation using a lateral view from close to long ranges. Because of the large variation, the scattering volume changes dramatically, increasing with the square of the distance. On the other hand, the TPR has the advantage of similar sized scattering volumes in all locations. This objectiveness stimulated the idea of using the TPR to estimate the influence of sampling volume of ground radars. There are two other important facts that would suggest using the spaceborne radar as a reference for the GR: 1) a great deal of effort has been made to provide the TPR with long-term, continuously monitored electronic stability and 2) the calibration factor is assumed to have an accuracy of within 1 dB (Kumagai et al., 1995). TRMM was launched in November 1997, while the onboard precipitation radar was turned on in observation mode in December 1997. Since the activation of the first spaceborne weather radar onboard the TRMM satellite, it has been possible to monitor meteorological Ground-based Radar (abbreviated hereafter in this paper as GR) throughout the world (at latitudes covered by the satellite, namely, within ±35°) using the TRMM Precipitation Radar (abbreviated hereafter in this paper as TPR), despite the fact that there are enormous differences between the TPR and GR. In this respect, the different sampling volumes, geometrical viewing angles, operation frequencies, attenuation, sensitivity and times of acquisition should be noted. Hence, a quantitative comparison between spaceborne and ground-based weather radar is a challenge, as can be seen in several references (e. g., Bolen and Chandrasekar, 2000; Anagnostou et al., 2001; Liao et al., 2001; Keenan et al., 2003; Bolen and Chandrasekar, 2003; Houze et al., 2004; Amitai et al., 2004). Results from Gabella et al. (2006) referred to a single site and were limited to the lowest elevation. In this paper, there is a further exploitation of the concept: 1) two sites are analyzed (note that the meteorological focus is still in the same geographical region, namely the southeastern Mediterranean area, where the absence (or presence) of few precipitation events may switch the climate characteristic from semi-arid to arid or vice versa); 2) the overshooting effect is assessed also for the 2nd elevation (as expected it is much larger than for the lowest elevation). Section 2 provides an overview of the geographical, instrumentation and data characteristics of the two GR used in this study, as well as a brief description of TPR. Section 3 shows how TPR data could be used to adjust ground-based radar echoes as a function of the range from the radar site itself. The procedure has been applied to two Mediterranean sites, one in Cyprus and the other in Israel. Section 4 presents the results for both of these sites, using the lowest and the second elevation of both ground-based radars. The discussion and conclusions in Section 5 close the paper.

121

2. Study area, instrumentation and data description 2.1. The Doppler radar in Cyprus at 1325 m in altitude (Kykkos site) In 1995, the Meteorological Service of Cyprus purchased a C-band Doppler radar, designed for nowcasting use. Since its installation on the Kykkos site, weather forecasters have been using the radar to issue warnings for hazardous weather. The interpretation of the radar products is purely qualitative for this application. The radar was installed on the northwestern, mountainous region of the island. The radar site (Latitude: 34.98°; Longitude: 32.73°), is at 1310 m above-sea-level; the antenna tower is ~ 15 m. Fig. 1 in Gabella et al. (2006) shows a digital elevation map of the island, the radar site and, above all, the two sectors with considerable beam occultation, which is caused by the Troodos massif in the SE direction and the Tripylos hill in the NW direction. The nearby (10–15 km range) high Olympus peak (1951 m above sea level) in the Troodos massif causes considerable ground clutter, in addition to beam shielding behind it in a “large” sector (approximately between 100° and 140° azimuth). A much narrower sector (between 190° and 200° azimuth) is shielded by the closer Tripylos hill (1450 m above sea level). With the antenna focus at 1325 m above-sea-level and in standard refractivity conditions, the beam axis at the lowest elevation (0° elevation) reaches a maximum altitude of ~2000 m at a 110 km range, which is the maximum distance used in this paper. The beam axis of the 2nd scan (1° elevation) reaches ~4000 m at the 110 km range. In this study, 2 μs pulses were transmitted with a pulse repetition frequency of 250 Hz. The raw reflectivity values were sampled using a 1° interval in azimuth and 500 m radial resolution range-bins. The main features of the GR are listed in Table 1. 2.2. The radar in Israel at 65 m in altitude (Shacham site) The Israeli C-band radar is located close to Tel Aviv airport, which is about 15 km off the coast. The radar site (Latitude: 31.99°; Longitude: 34.90°), named Shacham, is at 42 m above-sea-level; the antenna tower is ~23 m. In their paper, Morin and Gabella (2007) display a digital elevation map of the country and the radar site: Israel's physiography consists of 3 main longitudinal strips: the coastal plain, the hilly regions (Galilee, Samaria and Judean Mountains) and the Jordan Rift valley. The hilly ridge east of the radar causes both ground clutter and beam blockage, which represent two major difficulties of radar rainfall estimation in complex terrain. The lowest elevation (1°, see below) is obviously more affected than the second one (1.6°). Additional ground clutter areas surround the radar at closer ranges, up to ~25 km. With the antenna focus at 65 m and in standard refractivity conditions, the beam axis at 1° elevation (the lowest scan) reaches an altitude of ~2700 m at a 110 km range, which is the maximum analyzed distance in this work. The beam axis of the second scan (1.6° elevation) reaches ~3700 m at 110 km range. In this study, 2 μs pulses were transmitted with a pulse repetition frequency of 250 Hz. The raw reflectivity values were sampled using 1.4° intervals in azimuth and 1000 m radial resolution range-bins. The main features of the GR are listed in Table 1 (see also Gabella et al., 2011).

122

M. Gabella et al. / Atmospheric Research 119 (2013) 120–130

Table 1 Attributes of the Cypriot (Kykkos) and Israeli (Shacham) Ground-based Radars (alternative configurations are listed in parentheses). Cyprus

Isreal

Transmitter–receiver: Peak power Carrier frequency Pulse repetition frequency Pulse duration Minimum detectable signal

158 kW 5.7 GHz 250 Hz (1180 Hz) 2 μs, (0.5 μs) − 110 dBm, (− 104 dBm)

250 kW 5.5 GHz 250 Hz 2 μs − 108 dBm

Antenna: Half Power Beam Width (HPBW) Polarization

1.1° Horizontal

1.1° Horizontal

Data: Maximum range used Resampled radial resolution Resampled azimuthal resolution Number of power levels Radiometric resolution

120 km (480 km) 500 m (125 m) 1° 80 1 dB steps per level

120 km (max. usable range is 186 km) 1000 m 1.4° 80 0.5 dB steps per level

As suggested by Morin and Gabella (2007), prior to any computation and averaging (see Section 3.1), the radar reflectivity values were increased by 6 dB to compensate for system losses not thoroughly taken into account in the implemented conversion from received power (in dBm) to radar reflectivity (in dBZ). 2.3. The TRMM Precipitation Radar (TPR) Kummerow et al. (1998) offered a comprehensive description of the TRMM sensor packages. A complete description of the Ku-band TRMM Precipitation Radar can be found in Kozu et al. (2001). The TPR data used in this study are attenuation-corrected radar reflectivities obtained at 13.8 GHz with the TRMM 2A25 algorithm described in Iguchi et al. (2000); this algorithm produces the best estimate of radar reflectivity close to the ground level. The TPR vertical resolution, V, at the nadir is dominated by the “equivalent” (chirp radar signal) pulse length: V is ~ 250 m. With increasing distance from the nadir, the resolution of the TPR samples becomes poorer in the vertical. This is caused by the inclination of the TPR pulse volume, as can be seen in Fig. 1 in Joss et al. (2006). The TPR beam is scanned electronically from the nadir using 24 adjacent beam positions on both sides. The antenna phase shifters are programmed to increment in constant 0.75° steps from angle bin to angle bin. This scanning program leads to a swath scene made up of 49 footprints (24 for each side plus one at the nadir). Let α be the off-nadir angle, Ψ the beamwidth of the TPR pencil-beam antenna, and H the altitude of the satellite (above the Earth). At increasing off-nadir angles, the vertical layer Depth, Δ, becomes larger, according to the following law: Δ = H⋅Ψ⋅tanðαÞ + V⋅cosðαÞ

ð1Þ

In other words, the vertical resolution is no longer dominated by the equivalent pulse length of the TPR: the “pulse limited case” is replaced by the “beam width limited case”. For definitions and explicative figures, the reader is referred to Nathanson et al. (1991; p. 73).

Of the many output variables that are available from the TRMM 2A25 product, this study uses the attenuationcorrected radar reflectivity calculated for the lowest TPR pulse volume, the so-called NearSurfZ. The echo heights range between 1.5 and 3 km above-sea-level (most echoes are at about 2 km above-sea-level). Below this altitude, TPR echoes are influenced by ground clutter: this limitation depends on the backscattering coefficient of the surface, the height of the topography over the land and the rain intensity. 3. Range-adjustment of ground-based radars, using the TPR as a reference 3.1. The concept The concept proposed by Gabella et al. (2006) uses the TPR as a reference to range-adjust the GR radar reflectivity estimates. The TPR, on the one hand, allows this assessment to be made, because its measurements originate from similar 400–420 km distances. The GR, on the other hand, has to measure rain from “close” to the radar (10 km in the present analysis) to long distances from it (110 km, in the present analysis). Consequently, the GR scattering volume, which increases with the square of the distance, changes significantly. This beam broadening-with-distance effect, together with the average decrease of the vertical reflectivity profile with height and the non-homogeneous beam filling cause an underestimation of the GR with range, on average. As an example, at longer ranges, the lower part of the sampling volume could be in rain, whereas, the upper part of the same pulse could be filled with snow, or even be without an echo. On the contrary, the change in distance from the TPR is small (between 400 and 420 km range) and can only be slightly or not even correlated to the range, D, from the GR site (see Section 3.2 for more details). Therefore, an eventually larger underestimation with increasing range from the GR site can be estimated using the TPR as a reference. To check the existence of the above-mentioned range dependence, the linear radar reflectivity averaged in circular rings around the GR site, is computed. This average reflectivity, bZ(D)N2π, which is a function of the distance, D, from the GR site, is computed

M. Gabella et al. / Atmospheric Research 119 (2013) 120–130

in the same circular ring for both radars. We use 7 rings “centered” at 25, 50, 65, 75, 85, 95 and 105 km: the last five rings are 10 km wide, respectively; the first and second ones are 30 and 20 km wide. Hence, the regions used to determine the average radar reflectivity values are large, considerably larger than the rather coarse TPR horizontal resolution. The selected large sampling rings reduce mismatches caused by different beam widths (mismatches in space) and GR–TPR time lag (mismatches in time, up to 25 min for the Cyprus radar). Let bGR(D)N2π and bTPR(D)N2π be average reflectivity values (azimuth-averaged in mm6/m 3) at a distance D from the GR site for both the GR and the TPR. These two variables show similar behavior, although they are hidden by the underestimating trend (with increasing distance) of the GR. Deviations caused by mismatches in space and time are reduced by averaging over the large area of the rings. While bTPR(D)N2π does not correlate with the distance from the GR site, bGR(D)N2π tends to decrease with the distance. The Factor F(D) = (bGR(D)N2π)/(bTPR(D)N2π), which is the ratio between the radar under investigation and the reference, is introduced as the dependent variable. The distance from the GR site, D, is used as the explanatory variable. The regression between the dependent and the independent variable is performed in the Logarithmic domain, where power laws lead to linear relationships. Obviously, the relationship between Log(F) and Log(D) is much more complex than a linear dependence. In purely stratiform rain for instance, as shown by Fabry et al. (1992) and Rosenfeld et al. (1993), we can first expect an almost constant value of F(D), then an increase (bright band contamination), followed by a rapid decrease (more parabolic than linear). However, we prefer to deal with an easy to understand, essential, “first-order” correction model, which is simply based on two coefficients: the first coefficient represents the bias at the “reference” distance; the second one represents the tendency of GR to underestimate with range. In mathematical notation:

 10⋅Log10

hGRðDÞi2π hTPRðDÞi2π



 = FdB ðDÞ = a0 + aD ⋅Log10

 D : D0

ð2Þ

To reduce the range-dependence of a0, we divide the predictor D by an intermediate, “reference” value, which is within the used 10–110 km range (D0 = 40 km, in this paper). In this way, a0 can help us to modify the calibration of the GR

123

so as to improve the average agreement between the radar estimates aloft and the measurements on the ground. The slope aD in Eq. (2) reflects the deviation of the radar sensitivity from the common 1/r 2 law (i.e., it reflects the rate of change of the calibration with distance). Since in Eq. (2) the Log10 is used, the value of aD represents how many dB compensation should be used for the Factor F(D) in correspondence to an increase in the range of a factor of 10; in other words, the units of aD as defined in Eq. (2) are dB per decade. Negative values can be expected and are in fact found (see next sections), since the sampling volume of the GR increases with the distance, as already mentioned. This is a consequence of non-homogeneous beam filling, caused mainly by the overshooting of precipitation. Section 4.1 presents the results of the regression for the lowest GR scan of both the Cypriot and the Israeli sites. Section 4.2 deals with the second elevation, which, on the one hand, is less affected by partial beam occultation, but, on the other hand, is much more affected by the overshooting problem. 3.2. Most suitable satellite ground-track distances from the ground-based radar site The horizontal resolution of the TPR at the nadir is ~ 5 km. This value is obtained if we assume that the effective instantaneous field of view is represented by the “3 dB one-way” angular resolution. This is the usual assumption, which, as already stated, is too optimistic. A more conservative value, e.g. 10 km, is probably more realistic and representative. The horizontal resolution of the GR is 1 km in Israel and 500 m in Cyprus (lowest horizontal scans). The vertical resolution of the GR decreases with range from the radar site. At the intermediate range of 60 km, it is ~ 1 km, using the (too optimistic) “3 dB one-way” angular resolution; again, 2 km is more realistic and conservative. The vertical resolution of the TPR is 250 m at the nadir and deteriorates off-nadir according to Eq. (1). However, in addition to the selected values for the beam width, pulse length or equivalent pulse length (the TPR being chirp radar), other attributes complicate the comparison: the minimum and maximum range of the GR (10 and 110 km in this paper) and most of all, the viewing angle of both radars. While the elevation angle shows little variability for the GR (between 0° and 1.6°, see Sections 3.1 and 3.2), the angle of incidence varies between 0° (at the nadir) and ± 18° (edges of the 240 km swath) for the TPR. As seen in Section 2.3, the TPR vertical resolution decreases with increasing angle of

Table 2 Coefficients in dB used to explain the GR/TPR ratio (on a Logarithmic scale), as a function of the Logarithm of the distance from the GR (Eq. (2)). The “offset” coefficient a0 reflects the radar “calibration” at the “intermediate” distance of 40 km. Even more interesting are: A) The “slope” coefficient aD, which shows that, on average, the apparent decrease in sensitivity of the GR with range, is almost 10 dB per decade (lowest ground-based radar scan); B) The coefficient of determination, R2, which shows that in three overpasses (out of four) only about 50% of the variability can be statistically explained by a power-law between the GR/TPR ratio and the distance from the GR site. GR site

TRMM overpass date and UTC time

Cyprus Cyprus Israel Israel

Feb. Feb. Feb. Feb.

11, 12, 24, 27,

2002 2002 2003 2003

22:50 00:28 02:03 00:54

Nearest-in-time GR scan (UTC)

a0 (dB)

aD (dB/decade)

R2

23:15 00:15 02:08 00:57

− 3.4 − 4.8 + 2.5 + 1.3

− 6.7 − 11.1 − 8.7 − 9.9

0.47 0.53 0.88 0.45

124

M. Gabella et al. / Atmospheric Research 119 (2013) 120–130

Table 3 Same as Table 2, but for the second elevation of the ground-based radar scan program. Note that the values of the coefficient of determination (last column) are considerably larger than for the lowest elevation (Table 2). Also the GR underestimation with range (second column from the right) is considerably worse than in Table 2. GR site

TRMM overpass: date and UTC time

Cyprus Cyprus Israel Israela

Feb. Feb. Feb. Feb.

a

11, 12, 24, 27,

2002 2002 2003 2003

22:50 00:28 02:03 00:54

Nearest-in-time GR scan (UTC)

a0 (dB)

aD (dB/decade)

R2

23:16 00:16 02:09 00:58

− 9.1 − 7.2 + 0.1 − 4.5

− 29.3 − 26.2 − 29.0 − 31.2

0.93 0.88 0.93 0.88

For this overpass the regression is based on 6 values rather than 7. The last ring, in fact, contains just a few, too weak (especially for the GR) weather echoes.

incidence. If the satellite passes exactly above the GR site, there is a remarkable positive correlation between the distances from the nadir line (parallel tracks) and the distances from the GR site (circular rings). This unwanted transfer should be avoided; otherwise, the influence of the decreasing TPR vertical resolution would be partly transferred to the radar under test, namely the GR. A similar negative transfer happens if the TPR nadir line is tangent to the GR maximum range (110 km, in this paper): in this case, the unwanted coupling between the two systems is anticorrelated. What are the “best” TRMM overpasses from this viewpoint? Little influence is found, when the satellite flies at approximately half way of the GR maximum range. The investigation on the interaction between the GR-range and the TPR-distance of the pulse volume, combined with the characteristics of the reference radar, were an important part of our analyses. However, its description has been left to another paper with a more mathematical content (Gabella et al., in press). The main consequence, for this paper, is that we selected the overpasses in which the TRMM satellite happened to fly approximately “half way” (say, between 40 and 70 km) from the GR site. We have also restricted the analysis to “winter” rainy events with 0 °C isotherm at ~ 2000 altitude (February 2002 and 2003). Adding up, the results in Section 4 refer to the following orbits and (GR site–TRMM nadir line) distances. Cyprus: orbit #24205, 49 km (Feb. 11, 2002); orbit #24206, 62 km (Feb. 12, 2002). Israel: orbit #30083, 71 km (Feb. 24, 2003); orbit #30129, 57 km (Feb. 27, 2003). 4. Range dependence of the ground-based radars, as seen by the TRMM radar 4.1. Results using the lowest ground-based radar scan Table 2 shows the values of the coefficients in Eq. (2), derived using two successive TRMM overpasses in February. GR data were acquired using the lowest scan both in Cyprus and Israel. The coefficient a0 reflects the calibration of the GR with respect to the TPR at the intermediate distance D0 = 40 km: a0 is found to be negative in Cyprus (around −5 dB) and positive in Israel (around + 2 dB). It is worth noting that the Israeli raw reflectivity values were increased by 6 dB (see Section 3.2).

The retrieved values of the coefficient aD are more interesting, because they show whether radar echoes acquired at various elevations should be adjusted with range. As expected, the values of aD are negative, indicating that both ground-based radars underestimate precipitation at longer distances. On average, the slope of the adjustment factor, FdB, as a function of the Logarithm of the distance from the radar site, is almost 10 dB per decade for both the Israeli and the Cypriot radar. Since 10 dB per decade corresponds to a factor of two per octave, this means that an increase of the ground-based radar range by a factor of two (e.g. from 50 km to 100 km) seems to require a compensation of ~3 dB to the radar reflectivity Z (in mm6/m3) expressed in dBZ; as stated, such compensation factor refers to the lowest elevation of both radars. Furthermore, by looking at the values of the determination coefficient (last column in Table 2), it can be noted that only a fraction of the variability of FdB (from a minimum of 45% to a maximum of 88%) is statistically explained by the Log of the range from the GR site. By converting reflectivity to rain rate (in mm/h) using a single power-law with exponent set to 1.5 (1.6) for, it would imply a range-adjustment factor of approximately 6.66 (6.25) dB/decade. 4.2. Results using the second elevation Increasing the elevation of the GR antenna, we expect the overshooting problem to become even worse: this implies a smaller value of a0 and most of all a larger underestimation with range. This implies more negative values of aD in Eq. (2). The results, which are shown in Table 3, are in line with the expectations. At the reference distance of 40 km, on the one hand, the Israeli radar no longer overestimates with respect to the TPR (24 February 2003) or it even underestimates by 4.5 dB (27 February 2003); on the other hand, the bias of the Cypriot radar becomes −9.1 and −7.2 dB, respectively. As in Section 4.1, we are even more interested in the retrieved values of the slope and the explained variance of the regression. With data acquired using the second elevation, the underestimation with distance becomes significantly larger for both ground-based radars with respect to the lowest elevation presented in Section 4.1. In terms of radar reflectivity, aD is as bad as −30 dB per decade for both ground-based radars. By converting reflectivity to rain rate (in mm/h) using a single power-law with exponent set to 1.5

Fig. 1. 24 February 2003 latitude–longitude representation for the 1.6° (top) and 1° (center) elevations of the Israeli, C-band, ground-based radar (GR) and the Ku-band, spaceborne TRMM Precipitation Radar (bottom) closest to the Earth surface (NearSurfZ). GR data were acquired at 02:09 and 02:08 UTC, while TPR data at 02:03 UTC. Radar reflectivity values are in dBZ. For this orbit, the distance between the TRMM ground track at the nadir and the GR site is 71 km.

M. Gabella et al. / Atmospheric Research 119 (2013) 120–130

125

Israeli GR ~1.6° elevation

Israeli GR ~1.0° elevation

TRMM Precipitation Radar

126 M. Gabella et al. / Atmospheric Research 119 (2013) 120–130

Israeli GR ~1.0° elevation

TRMM Precipitation Radar

Israeli GR ~1.6° elevation

Fig. 2. 27 February 2003 latitude–longitude representation for the 1° (left) and 1.6° (right) elevations of the Israeli C-band radar and (center) the TRMM Precipitation Radar (TPR). GR data were acquired at 00:57 and 00:58 UTC, while TPR data at 00:54 UTC. Reflectivity values are in dBZ. For this orbit, the distance between the TRMM ground track at the nadir and the GR site is 57 km.

M. Gabella et al. / Atmospheric Research 119 (2013) 120–130

127

Cypriot GR ~1.0° elevation

TRMM Precipipitation Radar

Range-adjusted Cypriot GR

Fig. 3. Average radar reflectivites for two consecutive TRMM overpasses (center) and the corresponding ground-based radar observations acquired using a 1° elevation angle: the top picture refers to original data and the bottom picture to range-adjusted data. TRMM overpasses were at 22:50 UTC on 11 February and 00:28 UTC on 12 February 2002. The nearest-in-time GR echoes were acquired at 23:16 UTC and 00:16 UTC. Reflectivity values are in dBZ.

128

M. Gabella et al. / Atmospheric Research 119 (2013) 120–130

Cypriot GR ~ 0° elevation

TRMM Precipitation Radar

Range-adjusted Cypriot GR

Fig. 4. As in Fig. 3, but for the lowest scan of the ground-based radar acquired with a 0° elevation angle. Reflectivity values are in dBZ.

M. Gabella et al. / Atmospheric Research 119 (2013) 120–130

(1.6) for, it would imply a range-adjustment factor of 20 (18.75) dB per decade. Finally, the variability of FdB, which tends to decrease more rapidly with increasing range, is (statistically) better explained than in the case of the lowest scan: it is easy to note (see last column of Table 3) that the explained variance varies from a minimum of 88% to a maximum of 93%. 4.3. Discussion Figs. 1 and 2 show the comparison between the first two elevations of the ground-based radar and the satellite radar over Israel for the 24 and 27 February overpasses. In Fig. 1 the top and center pictures show the second and first elevations, respectively. By looking at the edges of the pictures (that is far from the radar site), the overshooting problem is quite evident: for instance, using 1.6° elevation angle, no echo was detected westward of 34° longitude. The spaceborne radar image (bottom picture) confirms the well-known overshooting problem (see e.g. the rainy area in the upper-left corner, which is almost 150 km far from the GR site). On 27 February, the more intense area of the precipitation field was located close to the radar site: consequently, the focus in Fig. 2 is on the region surrounding Tel Aviv (zoom-in with respect to Fig. 1). Strong ground clutter affects both GR elevations in the area eastward of 35° longitude and southward of 32° latitude: the lowest elevation is obviously more affected and presents also (weaker) sea clutter to the North-West. Because of overshooting (combined with the rapidly decreasing vertical reflectivity profile), the northwestern cell (around 32.5° lat.; 34° lon.) is not detected at 1.6° elevation (right picture). Gabella et al. (2006) showed that more robust rangeadjustment coefficients can be derived by integrating more overpasses so as to increase the sample size. The central picture in Figs. 3 and 4 shows the TPR average reflectivity derived from two consecutive overpasses over Cyprus (orbit #24205 on 11 February 2002 at 22:50 UTC and orbit #24206 at 00:28 UTC on 12 February 2002). The upper and lower pictures show the corresponding ground-based radar echoes before and after the range-adjustment correction. Fig. 3 refers to the (second) 1° elevation: as in Israel, the overshooting problem dramatically affects GR observations (top picture). Range-adjustment causes a tremendous change (and improvements, on average). Fig. 4 refers to the (first) 0° elevation; as in Fig. 3, it is the (arithmetic) average of two consecutive overpasses of the so-called Z unit ([Z] = mm 6/m 3) displayed using a Logarithmic, decibel scale (dBZ). The retrieved (and applied) correction coefficients are a0 = − 4.5 dB and aD = − 9.4 dB/decade; for more details the reader can refer to the 7th line of Table 1, page 506 of the monographic review concerning radar calibration and adjustment by Gabella and Michaelides (2008). Also, for the interested reader, Fig. 1 (page 508) in Gabella and Michaelides (2008) and Fig. 3 (page 130) in Gabella et al. (2006) show the corresponding precipitation fields of each single overpass (GR, TPR and range-adjusted GR, respectively). 5. Summary and conclusions The radar sampling volume increases with the square of the range, which is often referred to as beam broadening. If

129

the hydrometeors were homogeneously distributed over the volume itself, beam broadening had no effect on the radar measurements. However, this is rarely the case. Hence, non-uniform beam filling could be one of the causes of the observed residual range dependence of the GR. Furthermore, the vertical radar reflectivity profile tends, on average, to decrease with height. Because of the Earth's curvature, the larger the range the higher the radar sampling volume. Consequently, another possible cause of the systematic range dependence is the old, well-known problem of overshooting, which, combined with the vertical decrease in the radar echo, can lead to serious underestimation. The effect becomes severe at distant ranges or in the case of partial beam occultation by relief. Overshooting of precipitation can even affect the lowest ground-based radar scans. Another possible cause of the underestimation with range can be attenuation along the radar “line of sight”. In this case, attenuation depends on the properties of the atmosphere between the radar site and the weather target zone. Hence, attenuation along the radar “line of sight” contributes not only to resulting negative values of aD but also to their variability. It is worth noting that TRMM PR shows that there was precipitation at the GR sites in all the four analyzed events. If the radome is covered by a film of water, the apparent rainfall intensity could be reduced to less than a half even in case of moderate precipitation (Germann, 1999). Hence, attenuation caused by the wet-radome can also be an important cause of the observed variability of the a0 coefficient. This paper illustrates the possible causes of the apparent decrease in sensitivity of the GR with range and presents a procedure that can be used to assess and eventually compensate this range dependence, using the radar in space as a reference. This analysis has already been applied and described in the literature, using the lowest scan of the Cypriot C-band radar. In this paper, the analysis has been extended to another C-band radar in the Mediterranean region, namely the Shacham radar close to Tel Aviv, Israel. Again a negative slope of the range-adjustment factor is found. The negative slope of the range-adjustment factor is interpreted to be mainly caused by overshooting: as a consequence, when comparing echoes from the second GR elevation with the TPR ones, we expected the range-dependency to be even more evident. In conformity with our expectations, the negative slope of the range-adjustment factor is remarkably larger for both the Cypriot and the Israeli ground-based radars. It can be concluded that when and where available, spaceborne weather radars can be used to monitor and even range-adjust meteorological radars at ground level. The underestimation of ground-based radars at distant ranges has often been verified in the literature by using rain gauges. Here, we use a spaceborne radar that permits more robust results to be obtained because of: 1) the larger number of samples that are available and averaged at similar ranges; 2) the volumetric nature of these samples, instead of gauge “point” measurements; and 3) the possibility of covering land and sea, where no gauges are available. Obviously, the suggested technique adjusts reflectivities but still faces the issue of converting reflectivity to rain rate. However, the possibility of extending the useful range by correcting GR observations, from inside Israel and Cyprus to over maritime

130

M. Gabella et al. / Atmospheric Research 119 (2013) 120–130

areas is highly desirable in the south-eastern Mediterranean. When using radar and gauges, we are forced to extrapolate the results derived over the land to the area over the sea. On the contrary, the combination of the TPR with the GR is valid in the whole (both land and sea) surveillance area of the ground-based and spaceborne radar. Acknowledgments The TRMM data were provided by NASA; the groundbased radar data by the Meteorological Service of Cyprus and by E.M.S (Mekorot). The researchers exchanging visits in Cyprus and Italy have been funded in the framework of the “Protocol of Cooperation between the Governments of the Italian Republic and the Republic of Cyprus on Research and Development” and under the RADVAL project (Conventional and non-conventional observations of rainfall in the Southern Mediterranean Sea, Cyprus Research Promotion Foundation Contract No. KY-IT/0906/02). References Amitai, A., Nystuen, J., Liao, L., Meneghini, R., Morin, E., 2004. Uniting space, ground, and underwater measurements for improved estimates of rain rate. IEEΕ Geosci. Remote S. 1, 35–38. Anagnostou, E.N., Morales, C.A., Dinku, T., 2001. The use of Precipitation Radar observations in determining ground radar calibration biases. J. Atmos. Ocean. Tech. 18, 616–628. Bolen, S.M., Chandrasekar, V., 2000. Quantitative cross validation of spacebased and ground-based radar observations. J. Appl. Meteorol. 39, 2071–2079. Bolen, S.M., Chandrasekar, V., 2003. Methodology for aligning and comparing spaceborne radar and ground-based radar observations. J. Atmos. Ocean. Tech. 20, 647–659. Fabry, F., Austin, G.L., Tees, D., 1992. The accuracy of rainfall estimates by radar as a function of range. Q. J. Roy. Meteor. Soc. 118, 435–453. Gabella, M., Michaelides, S., 2008. Adjusting ground radar using space TRMM Precipitation Radar. In: Michaelides, S. (Ed.), Precipitation: Advances in Measurement, Estimation and Prediction. Springer, Berlin, Germany. ISBN: 978-3-540-77654-3, pp. 491–512. Gabella, M., Duque, D., Notarpietro, R., in press. Comparing meteorological spaceborne and ground-based radars: optimal satellite overpass distance from the ground-based radar site. International Journal of Remote Sensing, vol. 32. Gabella, M., Bolliger, M., Germann, U., Perona, G., 2005. Large sample evaluation of cumulative rainfall amounts in the Alps using a network of three radars. Atmos. Res. 77, 256–268. Gabella, M., Joss, J., Michaelides, S., Perona, G., 2006. Range adjustment for Ground-based Radar, derived with the spaceborne TRMM Precipitation Radar. IEEE T. Geosci. Remote 44, 126–133.

Gabella, M., Morin, E., Notarpietro, R., 2011. Using TRMM spaceborne radar as a reference for compensating ground-based radar range degradation: methodology verification based on rain gauges in Israel. J. Geophys. Res. 116, D02114. doi:10.1029/2010JD014496. Germann, U., 1999. Radome attenuation: a serious limiting factor for quantitative radar measurements? Meteorol. Z. 8, 85–90. Houze, R.A., Brodzik, S., Schumacher, C., Yuter, S.E., 2004. Uncertainties in oceanic radar maps at Kwajalein and implications for satellite validation. J. Appl. Meteorol. 43, 1114–1132. Iguchi, T., Kozu, T., Meneghini, R., Awaka, J., Okamoto, K., 2000. Rain-profiling algorithm for the TRMM precipitation radar. J. Appl. Meteorol. 39, 2038–2052. Joss, J., Gabella, M., Michaelides, S., Perona, G., 2006. Variation of weather radar sensitivity at ground level and from space: case studies and possible causes. Meteorol. Z. 15, 485–496. Keenan, T., Ebert, E., Chandrasekar, V., Bringi, V., Whimpey, M., 2003. Comparison of TRMM satellite-based rainfall with surface radar and gauge information. Preprints of the 31st International Conference on Radar Meteorology, Seattle, WA, pp. 383–386. Kozu, T., Kawanishi, T., Kuroiwa, H., Kojima, M., Oikawa, K., Kumagai, H., Okamoto, K., Okumura, M., Nakatsuka, H., Nishikawa, K., 2001. Development of precipitation radar onboard the Tropical Rainfall Measuring Mission (TRMM) satellite. IEEE T. Geosci. Remote 39, 102–115. Kracmar, J., Joss, J., Novak, P., Havranek, P., Salek, M., 1999. First steps towards quantitative usage of data from Czech weather radar network. In: Collier, G.C. (Ed.), COST-75 Final International Seminar on Advanced Weather Radar Systems, Locarno, Switzerland. Commission of the European Communities, Brussels, Belgium, pp. 91–101. Kumagai, H., Kozu, T., Satake, M., Hanado, H., Okamoto, K., 1995. Development of an active radar calibrator for the TRMM precipitation radar. IEEE T. Geosci. Remote 33, 1316–1318. Kummerow, C., Barnes, W., Kozu, T., Shiue, J., Simpson, J., 1998. The Tropical Rainfall Measuring Mission (TRMM) sensor package. J. Atmos. Ocean. Tech. 15, 809–817. Liao, L., Meneghini, R., Iguchi, T., 2001. Comparisons of rain rate and reflectivity factor derived from the TRMM Precipitation Radar and the WSR-88D over the Melbourne, Florida, site. J. Atmos. Ocean. Tech. 18, 1959–1974. Morin, E., Gabella, M., 2007. Radar-based quantitative precipitation estimation over Mediterranean and dry climate regimes. J. Geophys. Res. 112, D20108. doi:10.1029/2006JD008206. Nathanson, F.E., Reilly, J.P., Cohen, M.N., 1991. Radar Design Principles2nd Edition. McGraw-Hill, New York. Nelson, B.R., Krajewski, W.F., Kruger, A., Smith, J.A., Baeck, M.L., 2003. Archival precipitation data set for the Mississippi River Basin: algorithm development. J. Geophys. Res. 108 (D22), 8857. doi:10.1029/2002JD003158. Rosenfeld, D., Wolff, D.B., Atlas, D., 1993. General probability-matched relations between radar reflectivity and rain rate. J. Appl. Meteorol. 32, 50–72. Vignal, B., Krajewski, W.F., 2001. Large sample evaluation of two methods to correct range-dependent error for WSR-88D rainfall estimates. J. Hydrometeorol. 2, 490–504.