Introduction
Spaceborne microwave sensors provide valuable data on physical characteristics of surface and upper snow layers of the ice sheets. Significant regional variations of the emission and backscattering properties have been observed over Antarctica and Greenland (Reference Swift, Hayes, Herd, Jones and DelnoreSwift and others, 1985; Reference Rott and PampaloniRott, 1989; Reference Zhang, Pedersen and GudmandsenZhang and others, 1989; Reference Fily and BenoistFily and Benoist, 1991). Explanations of microwave signatures over the permanently dry snow areas include variations of snow morphology related to accumulation rate and temperature (Reference ZwallyZwally, 1977; Reference Rotman, Fisher and StaelinRotman and others, 1982), surface roughness properties (Reference Remy and MinsterRemy and Minster, 1991), sub-surface layering (Reference Gurvich, Kalinin and MatveyevGurvich and others, 1973), as well as combinations of surface roughness and volume scattering (Reference Ridley and PartingtonRidley and Partington, 1988). Currently available theoretical models on the interaction of microwaves with dry snow assume spherical or spheroidal ice grains as scatterers (Reference Fung and ChenFung and Chen, 1989; Reference Tsang and DingTsang and Ding, 1991), but do not take into account the large number of internal boundaries observed in polar firn.
To provide a data base for the analysis of spaceborne microwave measurements, field studies of microwave signatures of snow and ice were initiated by the German Alfred-Wegener-Institut in cooperation with other institutions. During the first of these field campaigns, carried out in summer 1989–90 in Dronning Maud Land, Antarctica, we performed signature measurements in the X- and C-bands. Further information on the angular dependence of backscattering in the C-band was derived from scatterometer measurements of the European Remote Sensing Satellite ERS-1.
Field Measurements
During an oversnow traverse from the German coastal station Georg-von-Neumayer (GvN) to a traverse across Heimefrontfjella mountains between 28 December 1989 and 28 February 1990, we measured backscattering and emission properties of snow in the C- and X-bands ( Fig. 1). Measurement sites were at GvN and along a profile of 200 km length across the Heimfrontfjella from Veststrau-men (V) to Base Camp (B) and Amundsen Ice (AI) at elevations between 500 m and 2300 m a.s.l.
Firn properties
In the area of the field campaign temporary summer melt is observed at elevations below 500 m a.s.l. up to 150 km inland from the coast. Near GvN the stratigraphie profile of the top 4 m of the snowpack typically shows two to three ice layers per m with thicknesses between 5 mm and 20 mm, as well as two depth hoar layers per m. Microwave measurements at GvN were made at the end of February, when the snowpack was completely frozen.
Along the Heimefrontfjella profile no indications of summer melt were observed. Nevertheless, the morphology of the snowpack across the profile was quite variable, caused primarily by differences in accumulation rate and mean temperature. Table 1 summarizes snow characteristics at three sites located in the northern, central, and southern part of the profile, respectively. The highest annual accumulation rate and highest mean snow density was observed at site B, which was located at an elevation of 1200 m, 20 km northwest of the main ridge of the Heimefrontfjella, the highest peaks of which rise to 2700 m. Increased snowfall due to orographic effects and deposition of drift snow by katabatic winds downstream from the mountain slopes are the reasons for the comparatively high accumulation rate of 350 kg m−2 a−1.
Fig. 1. Location map of the oversnow traverse 1989/90 of the Alfred-Wegener-Institut. Ice shelves are shaded. Measurement sites: G: Georg-von-Neumayer station, V: Veslstraumen, B: Base Camp, AI: Amundsen Ice.
Density profiles ( Fig. 2) reflect the differences in the accumulation regime. The density measurements were made for layers of 5 cm thickness, which is not adequate to resolve densities of thin layers. At site Β horizontal layering was less pronounced than at the two other sites, and grain sizes were more homogeneous. At site AI strong density variations were related to depth-hoar layers, which are formed in late summer and autumn below thin wind-crusts. Here in each of two snow pits eight distinct depth hoar layers were found within the top 2 m, with very loose snow and hexagonal plates of up to 4 mm diameter. During the field campaign, wind velocities were in general low, and surfaces at all measurement sites were remarkably smooth.
Microwave penetration
For the interpretation of microwave signatures, knowledge on the depth of the snow volume contributing to emission and backscattering is essential. Penetration of microwaves in snow is limited by losses due to scattering at inhomogeneities and absorption. Typically, at frequencies below 10 GHz, scattering losses are neglected, and the penetration depth is calculated by means of the effective dielectric constant ϵ, which is derived from volumetrically weighted averages of the dielectric constants of the constituents (air and ice in case of dry snow) by means of mixing models (Reference Stiles and UlabyStiles and Ulaby, 1982).
According to equations given by Mätzler (1987), we calculate the real part ϵ′ and the imaginary part ϵ″ of the dielectric constant of dry snow at −15°C with a density of 440 kg m−3: ϵ′ = 1.87; ϵ″ (5.2 GHz) = 1.0 × 10−4 ϵ″ (10.3GHz) = 2.1 × 10−4. The absorption coefficient k a derived by
amounts to k a = 0.008 for 5.2 GHz and k a = 0.0332 for 10.3 GHz for the snow conditions specified above.
In stratified polar firn, losses due to specular and diffuse reflection at internal boundaries between layers with different densities, as well as volume scattering at ice grains, have to be considered. Taking into account the dimension of the inhomogeneities in relation to wave-length, at frequencies ≤ 10 GHz scattering losses at the grains should be significantly smaller than absorption losses. However, reflections at internal boundaries of polar firn show up very clearly in X-band radar measurements (Reference Forster, Davis, Rand and MooreForster and others, 1991). Because the volume inhomogeneities of polar firn, from grains to layers, cover two orders of magnitude and show considerable regional variations, calculations on propagation of microwaves in the medium is a difficult task. Our estimates of penetration depth of dry polar firn are based on field measurements of transmissivity and on satellite radiometer measurements.
Table I. Snow characteristics at measurement sites
Fig. 2. Snow density as function of depth at three sites V: Veststraumen, B: Base Camp, AI: Amundsen Ice.
Figure 3 shows vertical profiles of the extinction coefficient ke at 5.2 GHz and 10.3 GHz, derived by means of radiometric transmission measurements on snow blocks of 0.3 m thickness, which were cut out of a snow pit at Base Camp. Transmittance ts of the radiation flux through a snow block was determined according to
where Ts is the thermodynamic temperature of the snow, Tm is the brightness temperature measured by the radiometer looking toward the sky with the snow block in front of the antenna, and Tsky is the brightness temperature of the sky.
Fig. 3. Extinction coefficient ke (m−1) of the top 3 m of firn at Base Camp, determined by means of transmission measurements at 5.2 GHz (broken line) and 10.3 GHz (full line).
The extinction coefficient ke, which includes scattering and absorption losses, is related to transmittance by
with δz = 0.3 m. Mean ke values for the top 3 m of the snow pit were:
These values are close to those reported by Mätzler (1987) for dry alpine snow, and are also in agreement with ke values which we derived by inversion calculations from Nimbus-7 Scanning Multichannel Microwave Radiometer (SMMR) measurements (Reference Rott and PampaloniRott, 1989). With (dp = 1/K e) we obtain the penetration depths shown in Table 2. The SMMR-based d p values were derived by numerical calculations of radiative transfer, based on the comparison of annual amplitudes of snow temperatures down to 10 m depth with annual amplitudes of SMMR brightness temperatures (Reference Rott and PampaloniRott, 1989).
Table 2. Penetration depth dp of polar firn at various frequencies f derived (I) by means of field measurements at the Base Camp; (II) by inversion calculations of annual brightness temperature amplitudes measured by SMMR in the areas of the East Antarctic stations, Plateau and Mizuho (Rott, 1989)
Fig. 4. C- and X-band brightness temperatures Tb at vertical polarization (□) and horizontal polarization ( + ) versus incidence angle, measured at Veststraumen, Base Camp and Amundsen Ice.
C- and X-band emission and backscattering measurements
Measurements were carried out with a dual frequency scatterometer/radiometer, which was mounted 3.5 m above the surface on the sidewall of a container for measurements at incidence angles between 10° and 80°, and on the crane of a snow-mobile for nadir measurements. The two radiometers are of the unbalanced Dicke type with center frequencies respectively of 5.25 GHz, bandwidth 0.56 GHz, and 10.3 GHz, bandwidth 1.06 GHz. The scatterometer operates with a separate noise transmitter in each channel and uses the radiometers as receivers. Two turnable horn antennae per channel, with half-power beamwidths of 18° enable measurements at different linear polarizations. Advantages of the noise system are the large number of independent samples and the possibility of using the same receiver for active and passive measurements. A disadvantage is the lack of range information.
The polarization behaviours of brightness temperatures T b in Figure 4 clearly reflect the differences in snow stratification at sites V, Β and AI. At 60° incidence angle the T b differences (δT b) between vertical and horizontal polarization increase from Β (δT b = 28 Κ in the C-band, δT b, = 34 Κ in the X-band) to site V (41 Κ in both bands), and to AI (48 Κ in the C-band, 50Κ in the X-band). Mean emissivity decreases from Β to V and to AI. Emissivities are in general slightly lower in the X-band than in the C-band, indicating increased effects of reflections and scattering at the higher frequency.
The behaviour of the active signatures is complementary to the emissivities. Because we could not separate the backscattering contributions of individual layers, we calculated the backscattering coefficient γ for given frequency and polarization from the measurements of the reflected power according to
where P r is the received power, P t the transmitted power, and G s is the effective scatterometer gain:
In our system the gains g r and g t of the receiving and transmitting antennae are identical in each channel. The 3 dB beamwidth for the power measurements, given by the convolution of g r and g t, is 7–8°, and shows a slight dependence on the range due to the parallax resulting from the spacing between the two antennae. For calculating γ of the volume scattering medium, we introduced an effective range:
where R 0 is the range from the scatterometer to the snow surface and δR is the weighted distance for the backscatter contribution of the snow volume measured from the surface (Reference Rott and SturmRott and Sturm, 1991). By means of radiative transfer calculations we determined values of δR = 3 m for 5.2 GHz and δR = 1.5 m for 10.3 GHz. The backscattering coefficient γ is normalized to the solid angle and is related to σ°, the radar cross section per unit area, by
where Θ is the incidence angle of the radar beam measured from the vertical.
The like-polarized backscattering coefficients at all three sites show a clear decrease in the incidence angle range from 10° to 40–50°, which is of the order of 15 dB at C-band ( Fig. 5). In the X-band the incidence angle dependence of the like-polarized γ is more pronounced at sites Β and V than at AI. At all three sites the like-polarized backscattering intensities at C-band are lower by a few dB compared to the X-band.
At site B, where the snow density is more homogeneous, the like-polarized γ values are slightly lower than at the two other sites. The γ values at HH and VV polarizations are in close agreement, with the exception of high incidence angles. The differences between co-polarized and cross-polarized γ are larger in the C-band than the X-band. This can be explained by increasing scattering efficiency of the rough internal interfaces and snow grains with increasing frequency.
Fig. 5. C- and X-band backscatterinng coefficients at the polarizations VV (□), UH ( +) J, VH (x), HV (◊) versus incidence angle, measured at sites Veststraumen, Base Camp and Amundsen Ice.
Active and passive signatures of refrozen firn
In Figure 6 results of T b and γ measurements near GvN are plotted. The backscattering coefficients show little dependence on the incidence angle and are – with exception of near nadir incidence – clearly higher than at the other measurement sites, which were located over permanently dry snow areas. Polarization differences of T b at GvN are comparable to those at B, but the emissivities (e = Tb/T) are clearly lower at GvN, where the mean thermodynamic snow temperature Τ is about 10°C higher. Backscattering and emission signatures at GvN are typical for a diffusely scattering medium. Ice lenses and ice trunks are strong scatterers at X- and C-band wavelengths.
Fig. 6. Brightness temperatures and backscattering coefficients at C- and X-band versus incidence angle at Georg-von-Neumayer station, 28 February 1990. Symbols as in Figures 4 and 5.
Backscattering Signatures Measured by the ERS-1 Scatterometer
Our first analysis of σ°, measured by the scatterometer aboard the European Remote Sensing Satellite ERS-1, was based on the period 25–30 November 1991. The Active Microwave Instrument (AMI) of ERS-1 operates at 5.3 GHz, VV polarizations. In scatterometer mode, AMI measures the backscattered intensity across a swath of 500 km width in three beams, perpendicular to the sub-satellite track (mid-beam), and at azimuth angles ±45° from the mid-beam (fore-beam and aft-beam respectively). The spatial resolution of a scatterometer cell is 50 km, the grid spacing of the processed data is 25 km (Reference FrancisFrancis and others, 1991).
Figure 7 shows examples of the angular dependence of σ° for four areas with different firn properties: an area covering 100 × 100 km2 around the site AI, an area of 200 × 200 km2 around Dome C (74.5° S, 132.2°E, 3240 m a.s.l.), an area of 200 × 200 km2 around Siple Station (75.9°S, 84.2°W, 1050 m), and an area of 50 × 50 km2 size on the Ekströmisen south of GvN.
Because the limited data set did not allow for detailed analysis of the azimuth dependence of σ°, we averaged over all azimuth angles for each of the beams. Preliminary analysis indicates some variations of σ° with the azimuth angle within an overall range of less than 3 dB at each of the four sites shown. At other areas in Antarctica, in particular in regions with strong katabatic winds, we found more pronounced azimuth dependence of σ°.
Scattering of the σ° data, plotted in steps of 2° in Figure 7, at AI and Ekström Ice Shelf, can be explained by the small data set and the spatial variability of snow morphology due to orographic effects in these areas. Considering these limitations and the differences in the footprint sizes, the correspondence of σ° measured by ERS-1 AMI over AI and Ekström with the field measurements at AI and GvN, respectively, is satisfactory. At the other field measurement sites the inhomogeneities at the AMI footprint scale, due to crevasse zones and mountains, were too big to allow the comparison.
In correspondence with the ground-based scatterometer measurements at GvN, σ° measured by AMI at the Ekström Ice Shelf is clearly higher and shows less variation with the incidence angle than at the other sites. The strongest decrease of σ° with the incidence angle and the lowest σ° values are observed at Siple. At incidence angles below 30° the σ° values at Siple are 10 dB below those at Dome C. The backscattering properties for these two sites, which are both in regions of permanently dry snow, reflect the differences in firn stratigraphy. At Siple the accumulation rate is 560 kg m−2a−1, mean annual temperature is −24°C (Reference Mosley–Thompson, Dai, Thompson, Grootes, Arbogast and PaskievitchMosley-Thompson and others, 1991), versus 34kgm−2 a−1 and −53°C at Dome C (Reference Palais, Whillans and BullPalais and others, 1982). Snow pit studies at Dome C showed horizontally discontinuous strata and close sequences of thin layers with significant differences in density. The variations of small-scale surface topography at Dome C are typically of the order of 10–20 cm over distances of several meters, which can be considered to be a rather smooth surface in the C-band. The surface properties, the angular dependence of σ°, as well as the penetration depth of cold firn, clearly point out that the differences between Siple and Dome C are primarily due to the backscattering contribution of the volume. Surface effects can play only a minor role with exception of back-scattering near nadir.
Fig. 7. Radar cross section σ° in dB versus incidence angle from ERS-1 scatterometer measurements on 25–30 November 1991, at Dome C, Siple, Amundsen Ice and Ekströmisen. ⃝ fore-beam, + : mid-beam, □ : aflbeam. Scale for Ekströmisen is shifted by 10 dB.
Conclusion
In permanently dry firn, areas with high accumulation rate and comparatively homogeneous snow morphology show low backscattering coefficients and high emissivities in the C- and X-bands. Pronounced stratification, including depth hoar layers, results in increased back-scattering intensities and in increased brightness temperature differences between horizontal and vertical polarization. Refrozen firn with sub-surface ice layers and ice lenses shows the highest backscattering coefficients with the exception of near-nadir incidence, and only weak angular dependence of backscattering.
The backscattering and emission signatures derived by means of field measurements and satellite data at the various sites with permanently dry firn are clearly different from the behaviour of an ideally diffusely scattering medium. The strong angular dependence of backscattering and the polarization differences of the emitted radiation emphasize the importance of diffuse and specular reflections at internal boundaries. Interfaces between the snow layers retain in many cases the sastrugi structure of the surface, with different roughness scales, from small-scale ripples in mm to cm scales to slightly inclined facets of cm to tens of cm size. Theoretical models of microwave interaction with polar firn have to take into account the complex structure. Field experiments on microwave signatures, in combination with studies of snow properties, provide an important basis for the analysis of spaceborne microwave measurements over the ice sheets.
Acknowledgements
This work was supported by the Austrian Science Fund (FWF) Project P8476. The Antarctic field campaign 1989/90 was organized and supported by the Alfred-Wegener-Institut for Polar and Marine Research. The European Space Agency made available the ERS-1 Scatterometer data for AO project Nr. A2.
