Development and Application of Small Spaceborne Synthetic Aperture Radars (1998)

Following a decline in imaging radar research in the 1970s and 1980s, the 1990s have witnessed a renewal of activity as researchers apply active and passive microwave capabilities for Earth observations. The past few years, in particular, have witnessed an increase in research based on European, Canadian, and Japanese free-flying SARs, as well as the series of Shuttle-based SAR flights (SIR-A, SIR-B, and the U.S.-Germany-Italy SIR-C/X-SAR). SAR interferometry is enabling many applications in solid-earth studies. In addition, biomass estimation, ecosystem delineation, ice dynamics modeling, and biological water monitoring have progressed. Among many others, reports by Dixon (1995), Evans et al. (1995), Goldstein (1995) (as well as the earlier paper by Goldstein et al., 1988), and Dobson et al. (1997) are rekindling interest in the variety of unique parameters that can be measured by multifrequency and multipolarization SAR systems. The following discussion of selected capabilities indicates the breadth and strength of research validating the applications described in these documents but is not intended to be exhaustive. In the committee's opinion, the important question is, What should the application focus and resulting design of this system be?

The high-priority science goals for the JPL LightSAR baseline design support the ESE solid-earth objectives to (1) monitor crustal deformation through millimeter-level interseismic vector deformation mapping along faults and plate boundaries; (2) monitor crustal displacements associated with volcanic activity and determine the spatial extent of newly erupted material; and (3) study other natural hazards such as floods and natural subsidence related to human withdrawal of water from subsurface reservoirs. These goals require repeat-pass interferometry, preferably with a long wavelength (such as L-band) and a short baseline. A set of subsidiary science goals supports ESE hydrology and carbon-cycle objectives. These are (1) to monitor the cryosphere (extent, topography, and velocity of glaciers and ice sheets); (2) to monitor annual forest disturbance and regeneration of tropical, temperate, and boreal forest zones; (3) to monitor near-surface soil moisture in nonforested areas; (4) to map the extent and water equivalence of snow cover; and (5) to monitor oceanic storms in selected regions of research interest. These goals would require the addition of a polarimetric capability and wide-swath modes.

Given the three potential user sectors (science, public, and commercial) and their potentially conflicting needs for SAR data, it is important that NASA explore alternative mission foci and system designs.

Both ground-based and airborne measurements derived from multiparameter systems have been used to map higher forms of vegetation. Lower frequencies (P-band to S-band) are better for mapping multilayered forests, where deeper signal penetration is required; higher frequencies (X-band to K-band) appear better suited for crops, grasses, and tundra.¹ The accuracy of

interpretation seems to improve if both like- and cross-polarized signals are collected. SAR is particularly important for filling gaps in crop and other vegetation data sets collected by optical sensors in cloudy regions.

Research dating from the 1960s shows that K_a- and X-band images can be used to produce vegetation cover maps of both natural and cultivated surfaces (Ulaby et al., 1986a, and references therein), Most of the research between 1964 and 1984 concentrated on temperate agricultural applications, but since then, the focus in the United States has turned toward ecosystems and natural vegetation. The European and Canadian research communities have continued their work on crops. A review of some of the work on crops appears in Chapter 21 of Ulaby et al. (1986a,b); Anys and He (1995) and Foody et al. (1994) report more recent research. Much early work in the 1960s and early 1970s was based on extensive imagery acquired by the K _a-band (35 GHz) like- and cross-polarized AN/APQ-97 side-looking airborne radar systems. Earlier validation can be found in Dellwig et al. (1975) and Banhart (1981, 1984). The extensive scatterometer work by Ulaby (1974, 1984) indicates that the K_u-band is the best frequency for crop discrimination, with the X-band a close second (Ulaby et al., 1986a,b, and references therein).

Because of their sensitivity to structural characteristics, multiparameter SARs offer a means to classify vegetation cover, as discussed in Chapter 2. SAR data can be used to detect deforestation and forest regrowth and to discriminate among up to 10 distinct vegetation types in a region, with an accuracy comparable to that attained with current electro-optical systems (approximately 89 percent). “Types” refers to vegetational communities having distinctive morphologies such as evergreen forest, deciduous forest, shrubland, marshland, and grassland. SAR is also sensitive to temporal-dynamic factors such as moisture content and freeze-thaw status. SAR is able to detect regional flood conditions, especially under variable canopies. For flooded forests, a lower-frequency (L- or P-band), HH-polarized design is preferred. For wetlands mapping, a higher-frequency, HH- or VV-polarized system is more suitable. The all-weather capabilities of SAR allow for repetitive coverage of flooded regions and provide a unique tool for use in disaster relief.

More recent research confirms SAR's sensitivity to forest biomass and plant moisture content, making it a useful tool either as a stand-alone sensor for vegetation applications or as a supplemental sensor. Biomass measurements are possible in both agricultural fields and natural vegetation such as forests and rangeland (Beaudoin et al., 1994; Dobson et al., 1995; Kasischke et al., 1995; Le Toan et al., 1992; Ranson and Sun, 1995; Rignot et al., 1995a; and Saatchi et al., 1995). Single-frequency, single-polarization SARs are sensitive to above-ground biomass differences in forests up to approximately 100 to 150 metric tons per hectare. Multichannel SAR systems, which include low frequencies (L-band at 24-cm wavelength and P-band at 65 cm) and a higher frequency (C-band at 6 cm or X-band at 3 cm), can be used to estimate biomass levels up to 250 to 300 metric tons per hectare. This biomass range includes all forests except mature old-growth forests in temperate regions and some tropical rain forests.

It is the committee's opinion that a small SAR mission focused on the above objectives would enhance the general understanding of SAR capabilities for monitoring rangelands, crops, and forests on a global basis if the data were made widely available for analysis. Such a mission

would also enhance National Wetland Inventory delineation and mapping efforts and would continue building data sets for tropical forest assessment and the dynamics of regrowth. Moreover, collaboration with operational agencies could lead to an experimental test of small SAR for flood detection and relief planning.

There are still many questions concerning the value of SAR for biomass estimation and cover-type mapping in vegetation studies. Among them are the following:

• What range of temporally varying factors influences SAR signatures across the breadth of vegetation and climatic regions worldwide?

• How sensitive is SAR for detecting variations in the mount of foliage in forests?

• What is the potential of interferometric SAR in vegetated regions?

Amplitude data alone permit the determination of snow facies, seasonal melt, surface morphology, ice velocity in rapidly moving regions, and iceberg production. Several studies have demonstrated SAR capabilities for monitoring ice-sheet motion, ice topography, glacial surges, and flow dynamics of the Greenland ice sheet (Joughin et al., 1995, 1996a,b; Rignot et al., 1995b). Multi-image amplitude and phase data add full spatial fields of ice velocity and surface topography. Interferometric SAR is the most important development for determining the surface velocity and topography of glaciers and ice sheets (Kwok and Fahnestock, 1996; Rignot et al., 1995b, 1996). Given suitable orbital parameters, interferometric SAR could provide a unique data set for glaciology that is not obtainable by any other means.

Although single-frequency, single-polarization SAR data have contributed to ice and glacier research (Rott et al., 1995; Shi and Dozier, 1995), multifrequency SAR might extend this research to other cryogenic phenomena such as depth probes of snowpack. Ascertaining the quantitative capabilities of such data, however, requires further research. Snow-water equivalent cannot be measured for wet snow, because the SAR signal cannot penetrate sufficiently into the snow. The snow-water equivalent of dry snow may be amenable to measurement using multifrequency polarimetric SAR, but this also requires experimental verification. A focused small SAR mission would expand the data sets needed to answer many questions regarding the utility of L- or X-band in multifrequency and interferometric investigations of ice sheets and glaciers. Such a mission in a near polar orbit would collect data at latitudes much higher than those of any of the previous Shuttle radar missions and thus be able to gather a uniquely valuable data set for the cryosphere.

Data from Seasat, ERS-1, and the SIR missions show that many oceanographic phenomena are observable from SAR data, although the underlying imaging mechanisms are not always well understood. A SAR designed to enhance marine measuring attributes, and accompanied by appropriate surface measurements, could lead to better understanding of ocean

waves, eddies, slicks, sea-ice formation, and other marine phenomena (Apel, 1994; Beal et al., 1983; Carsey and Garwood, 1993; Drinkwater et al., 1992; Fetterer et al., 1994; and Forget and Broche, 1996). Scientists may well improve their knowledge of the circulation and dynamics of global oceans from these techniques and from data being collected by other international SAR systems.

The driving force for Seasat SAR was ocean wave imaging to determine wave spectra. For many years, most of the research on oceanographic SAR was on wave spectra; indeed, ERS-1 has a special wave mode. Researchers over the years have made much progress extracting spectra from radar images. Although the questions are not all resolved, recent work shows that spectra may truly be determined from SAR by iterative techniques (Hasselmann et al., 1996). SAR is the only sensor capable of obtaining global samples of ocean spectra from space. These spectra are important for wave forecasting as well as for describing meteorological conditions. Lower satellite altitudes are preferred for ocean wave imaging (Monaldo and Beal, 1995). For operational purposes, long-duration missions are necessary. If a small SAR mission were to be optimized for observation of ocean wave spectra, a relatively low orbital altitude (˜300 km) would be preferred.

In coastal oceanography, single-frequency, single-polarization SARs have demonstrated their ability to record internal waves, surface waves, bathymetric features, and the location of ocean fronts (Forget and Brosche, 1996). In the open ocean, the frontal location of major currents such as occur in the Gulf Stream and Gulf of Alaska can be observed and measured (Liu et al., 1994; Thompson et al., 1994). Multifrequency, multipolarization SAR is able to distinguish between oil spills and natural surfactants, but additional research is needed to document the extent of this capability. Along-track interferometry would enable direct measurement of (one component of) surface current motion.

SAR applications for the hydrological sciences focus on biological water (both soil and plant) and on free-standing water in wetland ecosystems. Soil moisture is a parameter long sought by the agricultural community—and one that has growing importance for ecosystem modelers. It is a key environmental variable in both research and operational applications but is highly elusive and thus difficult to measure. Truck-mounted, aircraft, and ERS-1 measurements from the 1960s to the present have confirmed the correlation of radar backscatter with surface soil moisture (Dubois et al., 1995; Ulaby et al., 1996; Waite and McDonald, 1971). The nature of the interaction between a radar signal and the terrain is strongly affected by surface roughness, slope, and vegetation cover. The problem is to find the soil-moisture signal in the presence of other effects. K- and X-band frequencies respond to a combination of soil moisture and surface roughness in the top few centimeters of bare soil but cannot penetrate very deeply into the root zone where measurements often are required. L-band appears to be a good frequency for soil-moisture measurement in flat areas; however, the L-band scattering coefficient is so sensitive to the local angle of incidence that C-band may be superior in rough terrain (Ulaby et al., 1986b). Multifrequency systems with full-polarization capabilities may prove to be unique sensors for making relative comparisons of surface and root zone measurements of soil moisture, but research is required to document this potential.

Biological water can be measured either as canopy moisture or as free-standing water at ground level. Saatchi et al. (1995) have shown the utility of SAR for canopy moisture measurements in forests and grasslands. Rignot and Way (1994) and Kwok et al. (1994) have reported on freeze-thaw detection in forest and tundra environments. At ground levels, detection of seasonal flooding cycles and wetland-dryland ecosystem boundaries has been reported by Hess et al. (1995), Pope et al. (1994), and Wang et al. (1995). Most of these reports focus on humid tropical vegetation, suggesting that mid- and high-latitude ecosystem canopies might be penetrated and monitored even more easily for gross moisture patterns.

Geology was the first major application of airborne SAR. Regional geologic structure is readily apparent in wide-swath SAR images; its value was recognized early in the 1960s. The earliest applications of SAR and RAR on aircraft were to find new lineaments and to show that long fault lineaments previously thought to be segmented were actually continuous. Commercial airborne imaging radar programs that began in 1969 provided regional structural maps of large areas in often cloudy parts of the world. Many of these images and the derived geologic maps remain trade secrets for the companies who generated them, leading to speculation that the data resulted in major mineral discoveries. Although validation is difficult, some of the results are documented in the annotated bibliography by Dellwig et al. (1975). These results are derived from airborne radars that have modest resolution and much larger angles of incidence than satellite radars.

Among the high-priority scientific goals of the JPL LightSAR program are (1) to refine understanding of the earthquake cycle through millimeter-level interseismic vector deformation maps along faults and plate boundaries, (2) to monitor volcanoes for new activity and potential eruptions through millimeter-level deformation maps, and (3) to support additional research on natural hazards using SAR as a rapid and weather-independent monitoring tool.

The committee regards interferometry as among the most compelling uses of SAR for solid-earth studies (Gabriel et al., 1989; Massonnet and Fiegl, 1995; Massonnet et al., 1995; Peltzer and Rosen, 1995; Rosen et al., 1996; Treuhaft et al., 1996; Zebker et al., 1994; and Zebker et al., 1995). Because interferometric SAR has demonstrated the ability to image surface deformation on the order of a millimeter at regional scales, a suitably configured small SAR mission might permit measurement of large-scale topographic changes associated with earthquake cycles, small-scale topographic changes caused by volcanic inflation or deflation, lava flows, erosion, human activities, migration of mobile geologic features (e.g., sand dunes, glaciers), and incipient landslides. However, the added benefit of multipolarization, multifrequency SAR in these applications is unclear.

In detecting ground motion, SAR complements Global Positioning System (GPS) observations at point locations in several ways: (1) the continuous spatial imaging provided by SAR adds surface continuity to the point positions obtained with GPS; (2) GPS can provide full three-dimensional vector motion determinations, whereas SAR gives only displacements along the line of sight; and (3) the continuous monitoring capability of GPS permits the resolution of temporal variations of crustal motions in earthquake or volcanic eruption cycles.

Land use and land cover are critical elements of ESE and the U.S. Global Change Research Program because they focus attention on the rates and directions of global and regional landscape changes that occur on an interannual or decadal basis, that is, time periods that exceed intervals of catastrophic change due to flooding, fire, hurricanes, or tornadoes. The general premise is that human socioeconomic or political endeavors are manifested in the landscape as spectral and morphological changes. Some of these changes are so subtle that it takes years to measure them from satellite altitudes, and then only through indicators of change. For example, for changes associated with expanding urban populations and the consequent conversion of land uses from rural to urban types, one of the indicators typically is “miles of road per square mile.” This is one quantitative measure of expanding access to hinterlands undergoing economic change. This same indicator is used for forest management to indicate the intensity of deforestation.

In the lexicon of radar, land use and land cover are typically referred to in contexts such as “resource monitoring and management” (Winokur, 1996) and “ecology” (Evans et al., 1995), perhaps because the terms “land use” and “land cover” imply merely a backdrop of baseline attributes to be altered by human activities and natural processes. With a few major exceptions, applications of airborne radar, SAR, and interferometric SAR for imaging land use and land cover have been reported in terms of vegetation mapping, agricultural monitoring, soil-moisture mapping, and flood mapping, rather than in terms of land use or land cover. At the 1996 Workshop on Applications of Future U.S. Spaceborne Imaging Radar Missions (Bard and Leon, 1996), none of the agreed-on “key applications ” related directly to land use or land cover per se; rather, they focused on such topics as mineral exploration, military surveillance, topographic mapping, and hydrologic phenomena. Nevertheless, with a small SAR in orbit, users should have access to data sets that would further validate SAR's contribution to observation of long-term landscape changes.

Among several major exceptions in land use and land cover monitoring was Project RADAM in Brazil. This and later surveys in developing countries (especially in the humid tropics where aerial photography and optical satellite imagery are difficult to obtain on demand) used radar images to produce land use maps, including those for urban areas. This application has received little attention in the current literature. However, recent publications suggest important applications for settlement monitoring (Henderson, 1995; Henderson and Xia, 1997; Xia and Henderson, 1997). Since settlement expansion typically modifies the vegetation environment, this has important implications for monitoring and predicting changes in global biomass and carbon dioxide (CO₂).