Ice sheets and glaciers form the largest component of perennial ice on this planet. Over 75% of the world's fresh water is presently locked up in these frozen reservoirs. Their extents, however, undergo considerable fluctuations. The expansion and reduction of ice sheets and glaciers from glacial periods to interglacial periods has been one of the most dramatic climatic signals in Earth history. Even on short time scales of a few years, changes in ice volume have modulated and continue to modulate, sea level. This has never been more important than it is today, given the increase in economic development of coastal areas. Recent evidence of the rapidity with which sea level has changed due to rapid changes in ice-sheet discharge has heightened our awareness of the dynamic nature of this icy component of our climate. While it is certain that small glaciers are experiencing broad retreat and contributing to sea level rise by increased melting, the contributions of the two largest ice sheets in Greenland and Antarctica, which contain over 98% of the world's ice, remain unknown.
Ice sheets, by their very presence, affect global climate. Greenland and Antarctica both are dominated by large cold ice sheets that rise to high elevations and are among the highest albedo objects on Earth. Despite these similar characteristics, their effects on the climate have important differences. Antarctica has a nearly circular symmetry which encourages the development of strong zonal circulations of the atmosphere and ocean around it. On the other hand, Greenland sits as an obstacle to zonal circulations, deflecting the flow of the jet stream and preventing the establishment of strong zonal circulation.
Understanding these roles of ice sheets and glaciers in climate has been the focus of a small group of scientists for a relatively short period of time. The size of the community and the relative "newness" of glaciology are due, in part, to the remoteness and harsh climate of glacierized regions. These facts are juxtaposed with the climatic importance of ice sheets. Satellite remote sensing has revolutionized this branch of climate science by empowering glaciologists with instruments capable of collecting spatially extensive, yet detailed, data sets of critical parameters. SAR, with its ability to collect ice-sheet data through cloud and throughout the extended polar night has already demonstrated unique capabilities that make it an indispensable new tool for the glaciologist. This is only the beginning: more recent techniques, such as interferometry, have expanded the uses of SAR in glacier and ice-sheet research even farther, and promise glaciologists increased capabilities to collect critical data.
There are a number of key questions being addressed by glaciologists to which SAR data can make substantial contributions.
Are glaciers and ice sheets useful indicators of current climate change?
Glaciers and ice sheets most commonly occur at high elevations and in remote areas difficult to access. Such areas are typically not included in climate monitoring networks. If ways can be found to extract useful climatic parameters from observations of the ice, then ice sheets and glaciers provide valuable extensions of more traditional climate monitoring networks.
The interplay of annual cycles of snow accumulation and melting generates a succession of snow facies that serve as valuable indicators of the climatic regime characteristic of any point on an ice sheet or glacier (Williams et al., 1991). Figure 5-1 illustrates the essential subsurface stratigraphic characteristics. Beginning with the highest elevations, areas that experience no melt comprise the dry-snow facies. At lower elevations, melt-water percolates into the underlying, sub-freezing snow pack as a network of vertical ice pipes and horizontal ice lenses, forming the percolation facies. The amount of melt tends to increase at lower elevations. Eventually, melt is extensive enough that the latent heat released by the internal refreezing process warms the snow pack to the melting temperature throughout, and a fraction of additional melt water is retained within the snow pack in liquid form (the remainder leaves the ice mass as runoff). This situation characterizes the soaked-snow facies. At the lowest elevations is the bare-ice facies, formed when all surface snow is removed by ablation. The upper edge of the bare ice facies marks the snow-line at the end of the mass balance year.
As climatic conditions change, so will the positions of these boundaries. Conveniently, the low surface slopes of glaciers (typically 10-2) and even lower surface slopes of ice sheets (typically 10-3) transform subtle shifts of only a few meters in the elevations of these boundaries to large horizontal shifts on the order of a few kilometers. In addition to indicating the altitudinal extent and intensity of melt, these facies differ in spectral albedo. Thus, as facies extents change, so does the net radiation balance of the ice sheets.
Glacier length is the glaciological parameter with the longest history of observation. While glacier-length changes clearly illustrate that climate has changed, glaciers (and especially ice sheets) are one of the most sluggish climate components. Thus, the inverse problem of detecting or inferring climate change from measurements of ice extent is very difficult. This general view does not hold, however, on smaller, steeper mountain glaciers whose response times are of the order of years to decades.
How do the ice sheets affect atmospheric and oceanic circulations?
The general effects of ice sheets on the atmosphere and oceans were discussed above. Clearly, changes in either the albedo of the surface (through a change in snow facies extents), or changes in the geometric shapes of the ice sheets will alter their climatic effect and perturb the global climate system. Usually these changes are discussed in the extreme cases of glacial periods, when surface temperatures in central North America or Europe fell an average of 10 to 12 degrees Celsius, but smaller changes in ice sheets will also perturb the climate (Dawson, 1992). The relationship is probably highly non-linear; the climate record from Greenland ice cores shows that dramatic climate changes can occur in much less than a decade (Alley et al., 1993).
Remnants of vast armadas of icebergs have been detected in the western North Atlantic, well beyond the reach of ice sheets (Heinrich, 1988). It has been suggested that the melting of these icebergs would introduce a sufficient quantity of fresh water into the North Atlantic to completely transform the global pattern of oceanic circulation, altering the climate in every corner of the world (Broecker and Denton, 1989).
What is the current mass balance of ice sheets and glaciers?
Ice-sheet or glacier mass balance is defined as the annual difference between mass gain and mass loss. It is important globally because it has a dominant effect on sea-level change. During the last glacial maximum, global sea level was approximately 125 meters lower, the water being locked up primarily in the now-extinct Laurentide and Fennoscandian ice sheets of the Northern Hemisphere and an expanded Antarctic ice sheet (Shackleton, 1987; Denton and Hughes, 1981). Present ice volume, which is contained mostly in the Antarctic and Greenland ice sheets, is sufficient to raise sea level approximately 80 meters (Bader, 1961, Drewry et al., 1982).
By comparison, the annual turnover in ice-sheet mass is modest. Annual snow accumulation over the Antarctic and Greenland ice sheets is equivalent to an ocean layer only 6 mm thick (Bentley and Giovenetto, 1991; Ohmura and Reeh, 1991). A nearly equivalent amount of ice is returned to the oceans through melting and iceberg discharge. Estimates of present mass balance are poorly constrained: the Greenland ice sheet appears relatively stable, the Antarctic ice sheet seems to be growing slowly and the remaining small glaciers and ice caps are wasting away rapidly (Meier, 1993). Oddly enough, this range of behaviors may be due to the same phenomenon: global warming. Warmer summer temperatures enhance ice melting but also increase the frequency of precipitation events, which could result in growth of ice sheets and glaciers. In Antarctica almost no summer melt occurs except on the Antarctic Peninsula so the increased snowfall would enlarge the ice sheet. Mountain glaciers experience a net reduction in mass at low elevations due to the warmer summer temperatures, but, in some cases, may be growing at higher elevations. In Greenland, the two opposing effects appear to be roughly balanced.
These conclusions are based on sparse data. A notable weakness in one or more of these present estimates of mass balance is that they fail to add up to the present rate of sea-level change, even when liberal error estimates and other contributing effects are included (Meier, 1993). Confidence in predictions of future sea-level change must be tempered until we better understand the current mass balance of the existing ice.
What are the controls on ice flow and are there inherent instabilities in ice flow that could lead to dramatic changes in the dynamics of ice sheets or glaciers?
The subject of ice dynamics ranges from deformation and recrystalization of individual crystals, to surge-type glaciers, to the flow of the Antarctic ice sheet. Most of the controlling processes lie hidden within or underneath the ice, but the effects are clearly evidenced by deformation and flow of the ice. This topic is relevant to global climate because of the effect altered ice flow has on ice volume and, therefore, on sea level.
The ice-sheet environment determines the magnitudes of both snow accumulation and melting, but ice dynamics determines the rate at which ice is delivered to the oceans or to ablation areas. Thus, ice dynamics is a major component of the mass balance. This becomes obvious when examining the record of sea level since the last glacial maximum 20,000 years ago (Figure 5-2). During this period, sea level rose episodically in a series of brief jumps rather than smoothly. Geologic evidence confirms that these jumps correspond to the partial or complete collapses of marine-based ice sheets. Such behavior apparently caused sea level to rise at rates of at least 35 mm/yr, more than twenty times the present rate of rise. Although no such collapses have occurred for the last 4000 years, the West Antarctic ice sheet is the last marine-based ice sheet on the planet, and it contains enough ice to raise sea level more than five meters (Drewry et al., 1982). There is a pressing need to determine if, when, and how rapidly this ice sheet may collapse.
How much of the heat absorbed by the surface snow pack is retained by the ice mass and how much escapes as meltwater?
Warm temperatures melt surface snow, and the resulting meltwater most often drains into the underlying snow pack. Residual colder winter temperatures in this snowpack conduct heat away from the meltwater. If the residence time of the meltwater is sufficiently long, or if the snow pack is sufficiently cold, the meltwater refreezes. In this case, melting recorded at the surface does not represent mass that leaves the glacier; rather, this mass is captured at depth within the snow pack. On the other hand, if a system of snow-pack drainage is well developed with surface streams and moulins (vertical cavities) that quickly transport water through the glacier or ice sheet, not only can the initial quantity of meltwater leave the glacier, but additional melting can occur at the interface between rushing water and the surrounding ice.
The importance of these two extreme cases is apparent when viewed from the perspective of mass balance (Pfeffer et al., 1991). In one case none of the surface meltwater actually leaves the glacier, while in the other case the mass lost from the glacier is actually more than the amount of meltwater produced at the surface. The relevance of this effect to climate change lies in the fact that different snow facies respond differently to changing temperatures. For example, a formerly dry-snow area that begins to experience melt will retain most of the meltwater for many years as the system of internal drainage (comprised of vertical ice pipes and horizontal ice lenses) develops. However, a percolation facies that warms will be less effective at retaining increased amounts of meltwater, possibly delaying its eventual transformation to a soaked-snow facies.
SAR provides the obvious benefits of a weather-independent, day-night imaging system. These advantages are particularly crucial in the ice-sheet and glacier environments where persistent clouds continue to hamper data acquisitions by visible imagers and where the polar night imposes a prolonged period of darkness. In addition, unlike visible imagers, radar penetrates the snow surface, which provides glaciologists with a tool capable of sensing internal properties of the ice sheet or glacier.
Before ERS-1 was launched, limited SAR data of ice sheets only hinted at the potential glaciological uses of SAR. ERS-1 data have allowed full demonstration of many of these uses and have expanded the glaciological applications of SAR to even broader horizons.
The list that follows identifies the key parameters of ice sheets and glaciers that can be measured with SAR, and describes how glaciologists will be able to use this information to answer the key questions identified above.
Radar penetrates the snow surface, so the measured backscatter is a combination of surface and volume scattering. This characteristic enables SAR to discriminate clearly between all the snow facies described above using backscatter amplitude data alone. This discrimination is most effective during middle to late winter when surface water is absent. Dry snow appears dark in SAR because both surface scattering and volume scattering are low. At lower elevations, in the percolation facies, backscatter rises dramatically due to volume scattering from the network of subsurface ice bodies. The soaked- snow facies is composed of larger snow grains than the dry-snow facies because both melt-water and warmer temperatures serve to accelerate grain growth. Thus, the soaked-snow facies is radar dark when water is present, but in winter the backscatter increases to a level intermediate to the radar-dark dry-snow facies and radar-bright percolation facies. Finally, the bare-ice facies is radar dark due to strong specular surface scattering. Figure 5-3 clearly shows these different snow facies in a SAR-image mosaic of the Greenland ice sheet. The northeast portion of the dry-snow facies is slightly brighter than the rest of this facies. This is believed to be due to a decrease in the accumulation rate in this region (Jezek, 1993; Ohmura and Reeh, 1991). Figure 5-4 shows that the facies correlate closely with surface elevation. The ability to discriminate all the snow facies, which is impossible with visible imagery, establishes the unique use of SAR in a monitoring program of ice sheets for indications of climate change over their broad expanses. SIR-C/X-SAR data of the Patagonian ice fields indicate that facies discrimination is also possible using multifrequency data (Forster and Isacks, 1994).
Water is strongly radar-absorptive. This permits the use of SAR for monitoring of the summer melt season on ice sheets and glaciers. Time-series SAR data of Greenland have demonstrated that the gradual refreezing of free water at depth can be detected and have shown that this free-water component exists for a surprisingly long time after the
snow surface has cooled below freezing (Figure 5-5).
Radar at lower frequencies penetrates more deeply into snow. Thus, multi-frequency data permit a depth-variable view of the snowpack. This has been most clearly demonstrated by SIR-C/X-SAR data of the Patagonian ice fields (Figure 5-6). While a capability does not yet exist to quantify either the amount of melting or refreezing, the multi-frequency SARs sensitivity to conditions at different depths is already useful in monitoring the thermal and hydrologic evolution of ice sheets as climate changes. Future development of this capability will increase SARs utility in this area even further.
Mass loss by iceberg calving exceeds mass loss by melting for the Antarctic ice sheet and is approximately equal to the amount of surface melting for the Greenland ice sheet (Bader, 1961). Thus, this is a critical term in determining ice-sheet mass balance (although it does not directly affect sea level). Icebergs are visible in cloud-free visible imagery, but the requirements of a monitoring program include routine and dependable acquisition of high-resolution imagery, even during the extended polar night. These requirements match the capabilities of SAR and make it the preferred instrument for this activity. Figure 5-7 shows that icebergs as easily identified in SAR imagery.
Most surface morphological features are seen by either SAR or a visible imaging system. SAR holds the dual advantages of its all-weather, day-night capability while visible imagers avoid image degradation caused by speckle. Specific features that can be identified include streams and lakes, flowbands (linear forms stretched longitudinally in the direction of motion), ice edge, moraines and crevasses (Figure 5-8). The ability to identify such features opens the door to monitoring their evolution.
Lakes can be an especially good indicator of surface hydrologic activity. They tend to form in the same surface depressions each summer (locations fixed in space by the flow of ice over bed undulations), and their size and numbers are indications of the intensity of melt. Thus, they serve as ancillary data to the climate monitoring activity already described.
Ice margins are important because they change in response to changes in the flow of the ice. The radar contrast at many ice margins is less distinct than the contrast in visible imagery. However, there are many situations where this generality does not hold. These include areas where persistent cloud cover impedes collection of visible data, where fresh snow has covered the visible contrast between ice and adjacent terrain, and where ice near the edge is covered by rock or other debris. In the last case, differences in the polarization signatures of the moraine-covered ice and ice-free debris may permit identification of the ice margin.
Crevasses present serious hazards to field personnel but are one of the most useful natural features for glaciological study. Their orientations provide information on the state of stress within the ice. A more quantitative use is the measurement of ice motion accomplished by following unique crevasses or crevasse patterns over time (see next section). Additionally, SAR has demonstrated an ability to detect crevasse fields where visible imagery cannot (Hartl et al., 1994; Vaughan et al., 1994). This is due to either the detection of micro-cracks, below the resolution of the visible imagery, or from detection of buried crevasses by penetrating through the surface layers of the snow.
Ice velocity is one of the most fundamental parameters in the study of ice dynamics. The proven ability to obtain this information from time-sequential imagery using a cross-correlation technique significantly expanded the amount and density of such data available to glaciologists (Bindschadler and Scambos, 1991). This technique tracks patches of the surface containing crevasses and other surface features from one image to another by searching for the matching pattern of surface features in a second image. Displacements can be measured to sub-pixel accuracy, but typical displacements should be at least a few pixels to minimize the impact of systematic errors in coregistration of images (typically 1 to 2 pixels) (Scambos et al., 1992). Although developed initially for visible imagery, this technique also has been shown to work with SAR imagery (see Figure 5-9) (Fahnestock et al., 1993). A requirement of this technique is that the same sets of surface features be resolvable in both images. This requirement is not met over most of the ice sheets, but is usually met in the most active flow regions.
The application of interferometric techniques using SAR data holds the potential of obtaining ice velocity data from any ice sheet region. The technique utilizes the phase measurement of the backscattered radar pulse from every ground pixel to make a sub-wavelength scale measurement of displacement (usually a few millimeters) at every pixel (Goldstein et al., 1993). In the ideal case, the two images would be collected from the same point in space (zero-baseline). In practice, however, the baseline between observation points is finite so the interferogram contains a combination of motion and topographic information. Images must be coregistered to sub-pixel accuracy and the backscatter signatures from the same pixel in each image must be correlated for the phase difference to have a physically meaningful interpretation. Either a different viewing geometry or metamorphic changes in the surface or subsurface of a target pixel can destroy the phase correlation for a particular pixel.
Successful ice-sheet interferograms have typically used time separations of only a few days (Goldstein et al., 1993, Rignot et al., in press, Joughin et al., in press). Figure 5-10 shows an example of an interferogram from the Bagley Icefield in Alaska. Interferometrically measured displacements are in the direction of view, which for satellite SARs is in the range of 20 to 40 degrees from vertical. If the general direction of flow is known (along the regional surface gradient), one velocity component is sufficient to estimate the total velocity. Greater precision in velocity requires that a second interferogram be obtained from a different viewing angle. This can usually be accomplished by acquiring image pairs from both ascending and descending orbits.
Because interferograms contain no absolute displacement information, only velocity differences are obtained. Nevertheless, velocity gradient data (strain rates) are highly useful. To obtain absolute velocities, a theoretical minimum of two control values are needed to provide a datum and to correct for along-track variations in baseline. In practice, more control is desirable and may be necessary.
As mentioned above, interferometry with a non-zero baseline includes both topographic and motion information because the measured range difference (in units of phase) is the result of both surface movement and topography- induced parallax. This mixture of topographic and motion information requires that the topography be known with sufficient accuracy to remove its effects from the interferometrically-determined phase differences, in order to extract ice displacements. Fortunately, by using a third SAR image, an extremely clean separation of the topographic and velocity signals is possible if the velocity and topography are constant over the interval spanned by the three images (Kwok and Fahnestock, in press). Figure 5-11 shows an example of this separation.
As with the interferometric velocity output, the extracted topographic output is relative elevation, rather than absolute elevation. In principle, a single absolute elevation is sufficient to provide the datum for an entire interferogram but, again, insufficient knowledge of the precise baselines for each interferogram require that more elevation control points be used. The range of elevations spanned by a two-pi cycle of phase difference depends upon the baseline. For topographic relief of a particular scale, there is an optimal range of baselines between the undesirable extremes of too short a baseline, where insufficient parallax is achieved to resolve elevation variations, and too long a baseline where phases decorrelate. The accuracy of the derived elevations also is dependent upon the baselines of the interferograms. In one study area, shown in Figure 5-12, relative elevation accuracies of less than 2 meters were obtained with ERS-1 data having an effective baseline of 520 meters (Joughin et al., 1994). Eventually, the flow of ice is expected to be well enough understood that it will be possible to invert topographic and surface motion data to obtain basal topography and basal shear stress, which are additional parameters needed for ice-dynamics studies.
The ability to make the measurements described above with SAR data was demonstrated only after the collection of a substantial amount of ice-sheet data by ERS-1. The list is probably complete for the C-band, single polarization SAR carried by ERS-1. Before these data were available to demonstrate these techniques, limited Seasat and airborne data could only suggest the potentials that awaited glaciologists. In the case of interferometry, no mention of this now-proven technique was ever made in the pre-ERS-1 ice-sheet community. To extend this analogy, then, it probably is impossible to predict accurately the future uses of a SAR enhanced with additional polarizations and frequencies because multiple-frequency and multiple-polarization data sets of ice sheets from which to extrapolate remain very sparse.
Therefore, additional necessary work is posed in the form of questions that still need to be answered. In the process of answering these questions, new potential uses of SAR are likely to be discovered.
Is there a "best" SAR frequency for ice sheets and glaciers?
It is known that single-band data (C-band) permit almost all the analyses summarized above because a wealth of such data has been collected, from which these techniques have been developed. Seasat provided a limited amount of L-band data that confirmed it also can be used for snow facies, icebergs, and surface morphology research (Bindschadler et al., 1987). Limited airborne P-band and X-band data have hinted that these frequencies may also be used (Jezek et al., 1993). Recently, Space-Shuttle-based multi-frequency (L- and C-band), multi-polarimetric data have been added to the data pool.
What has been missing is a methodical comparison of data of the same ice- sheet areas using different frequencies and including complex data so interferometric products can be examined. L-band and P-band penetrate deeper than C-band, but the quantitative advantages and disadvantages of sensing deeper, older snow have not been established. A more diffuse volume-scattering component might provide a more temporally stable signature of the various snow facies. Airborne data have highlighted major differences in backscatter signatures when the frequency shifts from P-band to C-band (with the intermediate L-band being more like P-band) (Jezek et al., 1993). Similar backscatter variations have been seen in SIR-C/X-SAR data (cf. Figure 5-5) (Forster and Isacks, 1994). These differences could lead to techniques to derive a number of important variables including: grain-size versus depth distributions (critical for the quantification of accumulation rates from passive microwave data); meltwater production; and the amount of free water retained by the snow pack (by following the depth of the winter freezing wave as it penetrates the snow pack).
Interferometric applications might be aided by lower frequencies that permit longer baselines and have a relatively greater and more temporally stable volume-scattering component, making them less sensitive to meteorological events on the surface that decorrelate successive images. However, the increased contribution of the deeper volume scattering component could lead to an enhanced geometric decorrelation sensitivity or lower signal-to-noise, thus restricting available interferometric image pairs.
P-band radar might even penetrate the full depth of some glaciers. This would make it possible to map subglacial topography using interferometric techniques. Obtaining both surface and bed topography leads directly to ice volume--one of the critical climatic parameters discussed at the outset of this chapter. If successful, this would substantially improve all existing ice-volume estimates because existing data have been collected along linear ground tracks or flight lines, so the resulting ice-thickness maps have been produced by spatial interpolation.
Can useful scientific information be obtained by studying polarization effects?
Limited polarization data have been used to determine the dielectric constant and to extract the small-scale surface roughness of portions of the Greenland ice sheet (Jezek et al., 1993). The dielectric constant can be used to derive, albeit indirectly, surface albedo. Albedo has obvious importance in the energy balance of the ice sheet. Surface roughness is also a necessary consideration in exchange of energy because it affects the near-surface wind profile. Field measurements of surface roughness suffer from sampling sparseness but would be a necessary component of surface-truth experiments designed to develop the ability to remotely determine surface roughness. Given that different radar frequencies are sensitive to surface roughness on different scales, a wide spectrum of surface roughness may be obtainable.
In winter, the percolation zone displays anomalous backscatter polarization behavior. Similar behavior has been observed elsewhere in the solar system, ranging from icy Jovian moons to the Martian polar cap to, perhaps, a polar cap on Mercury. The common denominator seems to be ice, though not necessarily water ice. Anomalously polarized backscatter is, however, otherwise very rare in the solar system. Theoretical explanations of this are speculative, but all bear on the depth distribution of volume scattering. In the case of the percolation zone in Greenland, this depth distribution is linked to the redistributions of melt water and heat. This is of considerable interest because melting is one of the major mass-loss mechanisms of the Greenland ice sheet and could affect the salinity balance of the North Atlantic. Understanding and using anomalously polarized backscatter as a remote- sensing tool could lead to a unique probe of physical properties of both terrestrial and extraterrestrial icy regions.
What are the physical processes that cause target decorrelation and what are their relative effects?
It has been hypothesized that the occurrence of windstorms, snowfall, surface- and depth-hoar formation, melting, and refreezing between the epochs of two SAR images are meteorological events that can alter the backscatter signature from the target sufficiently to decorrelate the phase data and prevent the generation of interference fringes. Few studies have been done to actually quantify the effects of any of these events on the correlation of successive images. Jezek and Rignot (1994) hypothesized that spatial patterns of decorrelation in one ERS-1 image pair of Greenland may actually be due to variations in the distribution of snow deposited between the images. At C-band, 10 cm of fresh snow adds one additional wavelength (or fringe) to the round-trip radar distance.
More studies are necessary and will lead not only to guidelines for improving the likelihood of collecting correlated image pairs from which interferometric products can be produced, but also will produce meteorologically meaningful data over a spatially broad scale as compared with local data provided by sparsely distributed ice-sheet meteorological stations. Independent views afforded by interferometric SAR could prove valuable in interpreting the data sets provided by passive microwave sensors or radar altimeters, both of which also derive some of their signal from the sub-surface snow pack.
Most of the SAR technique-development investigations for ice-sheet and glacier research require data at frequencies other than C-band. Now that the C-band data set is extensive enough over the ice covered areas, more limited coverage at other frequencies can be placed into a meaningful context. This research is a necessary prerequisite to the development of a satellite SAR mission at any frequency other than the proven C-band. The fidelity and sparseness of JERS- 1 L-band data of ice sheets has failed to provide a data set capable of verifying the utility of L-band data for glaciological studies.
A critical component of the collection of data at frequencies other than C-band is the collection of interferometric pairs spaced in time so that motion information as well as topographic information are included in the phase differences. A particularly useful data set would be the collection of interferometric triplets of a moving ice sheet in at least C-band and L-band.
Ground truth is a mandatory part in the development of any new remote sensing application. This is certainly true with SAR data of ice sheets where volume scattering is often the dominant backscatter component. Field measurements are the only certain means of documenting specific physical properties of the snow pack and the temporal changes in these properties between remote data collections. To the extent possible, these measurements should be contemporaneous with airborne or satellite SAR acquisitions. Scattering measurements made with ground-based radar systems provide a data set for comparison with the remote measurements. Standard techniques allow surface parties to collect depth profiles of density, water content, grain size, conductivity, permittivity, temperature, and icy inclusions. These persons also can record surface meteorological conditions and help optimize the collection of SAR data from remote platforms.
ERS-1 has provided an invaluable SAR data set which has been used to establish the scientific utility of SAR for ice-sheet and glacier research. A few interferometric pairs await analysis, but for the purposes of developing or demonstrating new techniques, the use of this data set is virtually complete. No plans exist to place ERS-1 into a short repeat cycle orbit so that interferometric opportunities from this single satellite have probably ceased. The continuation of the ERS series with ERS-2 and ENVISAT will allow monitoring of the Greenland ice sheet snow facies to proceed.
JERS-1 carries an L-band SAR. It promised the same capabilities as ERS-1, but damage caused by a faulty antenna deployment has compromised the quality of the data. Thus, it has not provided an adequate opportunity to assess the merits of L-band data of ice sheets and glaciers relative to C-band.
The AIRSAR facility (C-, L- and P-band antennas on a DC-8 aircraft) provides the best existing means to collect the multi-frequency and multi-polarization data sets needed to assess the relative merits of these different windows of the electromagnetic spectrum. Multiple antennas have the obvious advantage of collecting simultaneous and coincident data. From knowledge of the positions of snow facies and surface features gleaned from the ERS-1 data set, aircraft missions can be planned in a manner that optimizes the utility of the collected data.
A critical augmentation to AIRSAR is the ability to collect interferometric data by navigating the aircraft with GPS real-time corrections. This capability will be crucial in investigations of the frequency-specific characteristics of interferometric data. The missions should be flown at various times of the year during periods when particular meteorological events (i.e., onset of melting and freezing, snowfall, high surface winds, etc.) are most likely.
ERS-2, when launched, will continue the capability of the ERS-1 SAR. An exciting prospect is an ERS-1/ERS-2 tandem mission for the collection of interferometric data.
RADARSAT, carrying another C-band SAR, is planned for launch in 1995. After an extended initial operational period, a scheduled orbit maneuver will afford RADARSAT the first SAR-view of most of the Antarctic ice sheet, including the regions south of latitude 78deg.S. The primary goal of this maneuver is to map the Antarctic ice sheet with SAR. It will take approximately two weeks, less than one orbit cycle, after which time RADARSAT will return to the nominal north-looking configuration. No possibilities exist for interferometric data collection during this short period. A second mapping may occur later in an extended RADARSAT mission. This mapping is exploratory, and it remains tantalizingly impossible to predict all that may be discovered with these data.
SAR remains a technology that is grossly underutilized in proportion to its proven capability to assist glaciologists in answering some of the most pressing questions in their discipline. These questions have direct relevance to global climate and future sea-level change.
SAR interferometry can provide data sets whose regional collection was never before feasible, yet are crucial to glaciological studies. A mission designed to produce complete interferometric coverage of permanently ice-covered areas promises extraordinary glaciological returns.
The omission of large portions of the polar regions by virtually every satellite mission to date continues a long, but undesirable, tradition that restricts the glaciological utility of satellite data. At present more than 2/3 of the Earth's permanent ice cover cannot be viewed by existing spaceborne SARs. Modern awareness of the climatic importance of the polar regions must be expressed in the ability of new sensors to extend their view to the poles. As in the case of RADARSAT, this polar view need not be available continuously, but, unlike RADARSAT, when available, it should be for a number of repeat cycles so that the enormous utility of SAR interferometry can be applied to the glaciological problems of global significance.
This chapter closes with the following specific recommendations:
(1) An interferometric mission at C-band should be conducted that includes multiple-image views of all ice sheets and glaciers sufficient to yield detailed surface topography and surface-velocity data sets.
(2) Future SAR missions should include maneuvering and data collection capability sufficient to monitor all permanently ice-covered areas at least once per year.
(3) Airborne and surface measurements should be carried out to assess the relative merits of different frequencies and combinations of frequencies and polarizations in deriving parameters needed to answer pressing glaciological questions relevant to the global climate.