The SIR-C / X-SAR radar observations of rain storms were the first multi-polarization and multi-frequency observations of precipitation from space. In addition to numerous often dramatic images of severe weather systems obtained by forming a synthetic aperture in the usual side-looking attitude, several data takes were performed while the radar antennas were parallel to the ground and the radar beams were pointing at nadir. These opportunities coincided with the passage of the Shuttle over Tropical Cyclone Odille in the southern Indian Ocean during the first flight, and over Typhoon Seth in the Western Pacific during the second flight. The resulting observations, or, more appropriately, the resulting measurements, demonstrate for the first time the capability of a spaceborne multi-frequency multi-polarization microwave radar system to quantify precipitation rates, to detect hydrometeor phase, and to classify rain type.

Introduction

This report summarizes the SIR-C / X-SAR radar observations of rain storms, the first multi-polarization and multi-frequency measurements of precipitation from space. These observations were made on SRL-1 in April, 1994, and SRL-2 in October, 1994. The main goals of the rain experiment were to verify the potential for the estimation of rainfall using multiple microwave radar frequencies, to demonstrate the use of polarization in detecting the freezing layer (the radar "bright band"), and to determine the effect of rainfall on the ocean surface backscatter. Detailed descriptions of the results have appeared in Jameson et al, 1997.

Data were collected from two observation geometries: nadir-pointing and side-looking. The nadir observations were obtained by rolling the Shuttle so that the radar antenna beams were normal to the Earth. The radars were then operating effectively as profiling scatterometers. In this geometry, the contamination of the rain echoes by the surface return within the main antenna beam is minimized. The resulting measurements were therefore quite readily amenable to quantitative interpretation. These nadir observations are described in section 2. The side-looking observations were processed using conventional synthetic aperture radar processing. The resulting images and the effects of the rain on the ocean surface are described in section 3.

Nadir Observations

When operating in a traditional side-looking observation geometry, the vertical extent of the radar antenna beams essentially integrates the rain returns over the entire depth of the troposphere. Since the change in atmospheric parameters is generally greatest in the vertical dimension, this "integration" in effect obscures much of the important atmospheric stratification. Therefore, in order to minimize this problem, the Shuttle was rolled on a few occasions so that the beams were orthogonal to the surface of the Earth for the duration of the data collection period. The challenge of this procedure, however, was the timing of the roll so that there would likely be precipitation in the radar antenna beams some time during the brief period of data collection. By developing decision rationales and procedures over the last few years and, quite frankly, because of good luck, we succeeded in collecting some interesting observations of precipitation in this mode. Before considering these measurements, it is important to mention that precipitating systems fall into two general categories, namely "convective systems" extending vertically over a significant fraction of the troposphere, and "stratiform systems" with negligible vertical air motion, in which the precipitation is largely confined to the lower troposphere (see, e.g., Houze, 1981).The former are associated with more intense precipitation, and, consequently, greater backscattered signals over several kilometers, while the latter are usually characterized by weak signals with the strongest backscatter occurring near the melting level. Consequently, it is a priori feasible to use SIR-C / X-SAR measurements to compute vertical profiles of precipitation in convective systems, but not in stratiform ones. Nevertheless, as we shall see, measurements in stratiform systems can still be quite useful.

Nadir measurements were obtained during one data take (DT 34.55) on SRL-1, and during three data takes (DTs 45.5, 54.0 and 103.0) on SRL-2. During these periods, measurements were collected with VV and VH polarizations at C- and L-bands, and with VV polarization at X-band. DT 34.55 from SRL-1 produced clear but somewhat disappointingly weak reflectivity profiles from several rain showers within the outer rain band of Tropical Cyclone Odille in the Indian Ocean on April 11, 1994, around 12:42 GMT. The L- and C-band radar operation parameters were incorrectly set during DT 45.5 from SRL-2. Fortunately, DTs 54.0 and 103.0 from SRL-2 produced data with excellent SNR. Both were performed over Typhoon Seth in the Western Pacific, DT 54.0 on October 3, 1994, around 17:44 GMT, and DT 103.0 on October 6, 1994, around 18:14 GMT. Both contain localized echoes from several convective cells as well as extended returns from stratiform rain. The remainder of this section will concentrate on describing one stratiform and two convective cases.

Figure 1 shows the measured VV effective radar reflectivity factors at the three radar frequencies. No surface clutter subtraction was performed on the SIR-C reflectivities because the surface return was saturating the receiver too strongly. However, we were able to subtract the clear-air surface return from the X-SAR data, although some residual contamination was unavoidable. The discrepancy between the C- and X-band reflectivities, on one hand, and the L-band reflectivity, on the other, is due largely to the difference in the along-track antenna beam widths: while the 3-dB beam width is 0.25 at C-band and a comparable 0.15 at X-band (which, with a nominal Shuttle altitude of 220 km, represent along-track beam widths of 1 and 0.6 km on the ground), it is 1.1at L-band (4.2 km on the ground). Thus, the obviously significant along-track variation of the intensity explains the large-scale difference between X/C- and L-bands. The along-track variation of the intensity also strongly suggests that the rain was very non-uniform across track too. While a gain corresponding to the nominal 5-6 across-track 3-dB beam widths (or about 20 km on the ground) was assumed to produce these images, it is much more likely that the rain echo was produced by precipitation occupying only a fraction of the actual antenna beam. In fact, since the along-track dimension of the "pockets" producing the strongest rain reflectivities manifestly did not exceed 2 km, one can reasonably assume that the same must have been true of the across-track dimension. Since this would correspond to a 0.5 beam width, one can therefore conclude that the measured reflectivities are very likely about 10log(5/0.5) = 10 dB too low because of this partial beam-filling effect.

While no significant cross-polarized return was measurable in this data segment, the return from a second convective cell did produce a distinct VH return at C-band. Figure 2 shows the C-VV, L-VV and C-VH reflectivities. The C-VH reflectivities were plotted using a specific color scale to enhance what appears to be the return from melting ice particles. Leakage from the V channel can be ruled out since the stronger VV returns from the lower altitudes do not appear in the VH data. Indeed, similar signatures were observed routinely during the Convection and Precipitation/Electrification (CaPE) project in Florida (see, e.g., Jameson et al, 1996), and were found to be due to the melting of denser ice particles.

The four panels in figure 3 show the VV relfectivities from stratiform precipitation within the outer edges of Typhoon Seth, as well as the C-VH reflectivity. As before, no surface-subtraction was performed except for the X-band data (where subtracting the clear-air surface return is necessary because of the long illuminating pulse).In all panels of figure 3, the appearance of the "melting layer" at an altitude of about 5 km is striking. While this observation is not new (the so-called "bright band" has been studied over the last 50 years), these observations are not only the first reports from a spaceborne radar, but they are the first collected simultaneously at three frequencies and at two polarizations. The rapid increase in signal intensity responsible for the banded appearance is due largely to the change in the index of refraction as ice begins to melt (see, e.g., Austin and Bemis, 1950; Browne and Robinson, 1952). The magnitude of this effect, however, is governed not only by the transition to a liquid having a much greater index of refraction, but also by other factors such as the kind of icy hydrometeor, any growth (accumulation of mass) of ice just before melting, and changes in fall speed as melting proceeds.

As to the measured reflectivities from the convective cells, they can be used to estimate quantitatively the amount of precipitation as a function of altitude. Consider the case of figure 1, where the absence of any detectable echo in the C-VH channel allows one to reasonably assume that the hydrometeors consisted mostly of liquid water. Using mutually independent hydrometeor size distribution parameters, given the measured C-VV and X-VV reflectivity profiles, one can retrieve the rainrate profile and the values of the parameters that are most likely to have given rise to the observed reflectivities. Since the retrieval problem in this case is stochastic rather than deterministic, due to the fading, system and surface-clutter noise contributions, and to the unknown partial beam-filling effect, one can apply stochastic filtering techniques (Haddad et al, 1996) to estimate the values of the variables involved, given a forward equation specifying the relation between the measured reflectivities and the DSD variables. An algorithm similar to the one described in Haddad et al, 1996, was used to estimate the rainrate profile through the center of the cell shown in figure 1. The details can be found in Jameson et al, 1997.

Side-Looking Observations

As part of the experiment, side-looking data over rain were acquired on data takes 13.4, 46.3, 66.6, 72.9, 140.4, 147.8, and 149.1 on SRL-1. On SRL-2, side-looking data over rain were acquired on data takes 13.4, 77.5, 114.0, 114.6, 119.0, 141.5, and 157.5. Images from these data show large modulations of the ocean in areas where there appears to be rain storms. Because of the side-looking geometry, quantitative interpretation is challenging. Previous side-looking SAR imagery of rain over the ocean has been interpreted in terms of the effects of wind and rain on the ocean surface (see, e.g., Atlas, 1994). These analyses, however, have dealt only with single polarization and single frequency data. The multiple frequencies and polarizations used in the SIR-C/X-SAR data provide an opportunity to improve the interpretation of the data.

DT 13.4 from SRL-1 is of particular interest because it captured an extensive rain area using mode "16X", giving the full scattering matrix at L- and C-bands. For our analysis the standard SIR-C and X-SAR calibrated image products were used. Figure 4 shows an example of three-frequency VV polarization imagery from this data take. The incidence angle for these data is 50. Figure 5 shows a contrast enhanced version of the L-band image with several dark areas noted A-G. For these areas, is near -40 dB, while the surrounding areas have near -20 dB. This variability in L-band backscatter is likely to be entirely controlled by ocean variations since, L-band attenuation in rain is small. At C- and X-bands, interpretation becomes more complicated, since both attenuation and scattering in rain could be significant. Areas A-E, which are darker than the surrounding areas at L-band, are brighter than the surrounding areas at C- and X-bands. These areas have C- and X-band up to -10 dB in contrast to -40 dB at L-band, while the surrounding areas have near -20 dB, similar to L-band. The same phenomenon can be see in the area just to the left of F and down to E. This type of increase in brightness with frequency is a common phenomenon, appearing not only in figure 4 but in many other data takes as well. The most likely explanation is that these areas were being subjected to moderate to heavy precipitation. At C- and X-bands backscatter from the precipitation is significant, increasing backscatter from these areas relative to the surrounding areas. At L-band backscatter from precipitation is not significant. Furthermore the ocean may be smooth due to damping of waves by heavy rain, making backscatter lower than the surrounding area (Tsimplis 1992). While A-E become bright as frequency increases, F and G remain dark at C- and X-bands, suggesting that the rainrate in them is low and that the ocean is smooth. In this case, the smooth ocean is probably related to low winds, since the rainrate would probably be too low to damp waves. The area between E and G exhibits decreasing backscatter with increasing frequency. The backscatter cross section drops from -27 dB at L-band to -32 dB at X-band, possibly due to increasing attenuation in precipitation.

Polarization observations provide additional information. The HH to VV ratio (called ZDR in the radar meteorology literature) should be strongly negative for ocean scattering at 50 incidence, but is typically near 0 or positive for scattering from hydrometeors (Jameson, 1989).Over much of the image, including the areas near A-G, HH is around 5 dB lower than VV at C-band; i.e. ZDR = -5 dB. However, within A-G, ZDR is generally larger, reaching a maximum of -2 dB, suggesting hydrometeor scattering is important. At L-band ZDR is consistently around -9 dB, both in A-G and throughout the image, confirming that the scattering is primarily from the ocean. A-E in figure 5 generally have high C-band LDR (up to -10 dB), likely due to backscatter from melting ice particles. Propagation through precipitation could also contribute to depolarizing the signal, as could ocean scattering. The C-band HH-VV phase difference is low throughout much of the image. However, it is large (>100) in many spots in the top portion of the image and particularly in the "hole" regions F and G. At L-band the HH-VV phase reaches 40 in G and is near zero elsewhere.

In addition to data over tropical oceans, SRL-1 serendipitously captured several rain cells over the Amazon rain-forest. These data are particularly interesting because while interpretation of SAR imagery of rain over the ocean is complicated by the possibility of backscatter and attenuation in rain, as well as the variability of the ocean itself, images of rain over the Amazon do not suffer from any significant surface variability (assuming that rain does not strongly affect backscatter from trees). Figure 6 shows images of a set of rain cells over the Amazon during DT 103.6 from SRL-1 at the three frequencies. At X-band the rain cells show up as dark areas, with a slight brightening at the leading edge of each area. At C-band, no brightening is visible, but some dark areas can be seen in the same location as the dark areas at X-band. At L-band neither bright nor dark areas are visible. In theory, the transmitted wave is both scattered and attenuated by the rain at all three frequencies. However, at L-band, both effects are small. At C-band the attenuation is large enough to reduce the observed by the two-way loss through the rain. The rain backscatter is small relative to the tree-top backscatter and is not seen. At X-band the backscattering from rain (-7 dB) is large enough that it can be seen above the cross section of the tree tops (-9 dB). However, as range increases, the signal backscattered from both surface and rain is attenuated strongly. These observations were successfully interpreted quantitatively using a simple scattering model: Figure 7 plots the observed and modeled versus range. Rain cells such as these can be found in much of the SIR-C/X-SAR data over the Amazon rainforest. See Jameson et al, 1997, for further details.

Conclusions

In summary, the SIR-C / X-SAR mission's data over rain storms successfully produced the first multi-polarization and multi-frequency measurements of precipitation from space. Rainfall was clearly detectable even at L-band. We were able to estimate rainrates from the C- and X-band nadir returns. Melting hydrometeors were clearly detectable in the cross-polarized C-band echoes from stratiform as well as convective systems, much as one would expect given past ground and air-borne measurements.

The data also confirm previous observations that rain and wind effects can produce large variations in 30- 50 incidence-angle sea surface backscatter. At nadir, however, the ocean surface echo could not be estimated due to severe receiver saturation.

Finally, several areas with significantly non-zero HH-VV phase differences and apparently lighter rainfall were observed in the ocean images.

References

Atlas, D.A. (1994), Footprints of storm on the sea: A view from synthetic aperture radar, J. Geophys. Res. 99: 7961-7969.

Austin, P.M., and Bemis, A.C. (1950), A quantitative study of the ``bright band'' in radar precipitation echoes, J. Meteor. 7: 145-151.

Browne, I.C., and Robinson, N.P. (1952), Cross polarization of the radar melting band Nature 170: 1078-1079.

Haddad, Z.S., Im, E., Durden, S.L., and Hensley, S. (1996(a)),Stochastic filtering of rain profiles using radar, surface-referenced radar, or combined radar-radiometer measurements, J. Appl. Meteor., 35: 229-242.

Houze, R.A., Jr. (1981),Structures of atmospheric precipitation systems: a global survey, RadioSci. 16: 671-689.

Jameson, A.R. (1989), The interpretation and meteorological application of radar backscatter amplitude ratios at linear polarizations, J. Atmos. Oceanic Technol. 6: 908-919.

Jameson, A.R., Murphy, M.J., and Krider, E.P. (1996), Multiple-parameter radar observations of isolated Florida thunderstorms during the onset of electrification, J. Appl. Meteor. 35: 343-354.

Jameson, A.R., Li, F., Durden, S., Haddad, Z.S., Holt, B., Fogarty, T.,Im, E., and Moore, R.K. (1997), SIR-C/X-SAR Observations of Rain Storms, Remote Sensing of the Environment 59: 267-279.

Olsen, R.L., Rodgers, D.V., and Hodge, D.B. (1978), The aR^b relation in the calculation of rain attenuation, IEEE Trans. Ant. Prop. 26:318-329.

Tsimplis, M.N. (1992),The effect of rain in calming the sea, J. Phys. Ocean. 22:404-412.