Short overview on SAR remote sensing

SAR remote sensing

Imaging radars are airborne or spaceborne radars which generate a reflectivity map of an illuminated area through transmission and reception of electromagnetic energy. Among other types of microwave sensors, special attention has been paid in the past to synthetic aperture radar (SAR) because of its high spatial resolution and multifarious information content.

The development of the synthetic array radar originated in 1951 with Carl Wiley, who postulated the use of DOPPLER information for increasing the azimuth resolution of the conventional side-looking aperture radar (SLAR) [1]. Based on this idea and following developments, the first SAR image was produced by researchers at the University of Michigan in 1958, using an optical processing method [2]. Precision optical processors and hologram radars were developed and fine resolution strip maps were obtained by the mid 1960's. Later on in the early 1970's digital signal processing methods were introduced to obtain off-line or non-realtime SAR images of high quality [3],[4]. Since these early days SAR systems evolved to an essential and powerful tool in geosciences and remote sensing. SAR data are applicable in many scientific fields. Besides traditional applications in geography and for topographic and thematic mapping, nowadays SAR sensors are also utilised in areas such as oceanography, forestry, agriculture, urban planning, environmental sciences and prediction and evaluation of natural disasters.

SAR sensors operate in the microwave region of the electromagnetic spectrum with typical wavelengths between 1cm and several metres. As an active system, a SAR emits by itself microwave radiation to the ground and measures the electric field backscattered by the illuminated ground patch. These measurements are transformed into a high resolution image. Because SAR systems are operated with an illumination of its own, they can perform equally well during day and night.

Looking at its transmission spectrum, it can be observed that the Earth's atmosphere is nearly transparent in the microwave region (see Fig. 1.1). Electromagnetic waves with wavelengths longer than 1cm even permeate almost undisturbedly small water drops. Thus the operation of a SAR is possible even in the presence of clouds, fog and rain, which are a limiting factor in optical remote sensing. This is of great importance for areas which are regularly covered by haze or clouds, like for example rain forests in the tropics. Here topographic and thematic mapping was nearly impossible before the introduction of imaging radars. Weather independence combined with day and night operation capabilities makes SAR an operational monitoring device for the entire earth's surface, a task which cannot be achieved with optical sensors.

Figure 1.1: The electromagnetic spectrum of the sun (top), and the transmission spectrum of the Earth's atmosphere (middle). At the bottom the most important spectral bands in remote sensing are marked.

Radar images contain quite different information than images obtained from optical or infrared sensors. While in the optical range mainly molecular resonances on the object surfaces are responsible for the characteristic object reflectivity, in the microwave region dielectric and geometrical properties become relevant for the backscattering. Radar images therefore emphasise the relief and morphological structure of the observed terrain as well as changes in the ground conductivity, for example caused by differences in soil moisture. Because of the sensitivity to dielectric properties, SAR images, in principle, also can provide information about the condition of vegetation, an important fact for agricultural and forestry applications.

Another important feature of SAR data results from the propagation characteristics of microwaves. Due to their long wavelength, microwaves are capable to penetrate vegetation and even the ground up to a certain depth [5],[6]. The penetration capabilities depend on the wavelength as well as on the complex dielectric constants, conductivities and densities of the observed targets. Shorter wavelengths, like the X-band, show typically a high attenuation and are mainly backscattered on the surface or on the top of the vegetation. Consequently, at these wavelengths mainly information about this layer is collected. Longer wavelengths, like L- and P-band, normally penetrate deep into vegetation and often also into the ground. The backscattering then contains contributions from the entire volume.

Figure 1.2: Wavelength dependency of the penetration capabilities of microwaves in vegetation and ground

One main problem in the analysis of SAR data is to determine exactly this superposition of different scattering contributions. Even though certain desired information is contained in the data, it is not accessible because only the total backscattering can be measured. An inversion of the measured data to parameters of interest is often ambiguous and cannot be solved without `a priori' information. Another problem is that the exact origin of the backscattering is, in principle, unknown as the entire SAR geometry has an elevational symmetry. Thus the elevation angle and, respectively, the topographic height of the observed scatterer remains unknown. Two main extensions of conventional SAR have been pursued in the past to resolve these limitations: SAR interferometry and polarimetry.

SAR interferometry

SAR interferometry (INSAR) is a technique, which analyses the phase difference between two SAR images acquired from slightly different positions. This phase difference is related to the terrain topography of the scene and can be used to generate high resolution digital elevation models (DEMs). The first basic experiments were made in 1974 by L.C. Graham demonstrating the capability of interferometric SAR for topographic mapping [7]. In the following years only little attention was paid to SAR interferometry. Experimental efforts were made by R.M. Goldstein and H.A. Zebker in the 80's, using a simple two-antenna modification of the NASA/JPL airborne AIRSAR system to demonstrate single-pass SAR interferometry [8]. In 1988 A. Gabriel and R.M. Goldstein showed the possibility of repeat-pass interferometry using spaceborne L-band data acquired by the SEASAT sensor [9]. Finally, first results from repeat-pass interferometry using the Canadian airborne CCRS sensor, in X- and C-band, were published in 1992 by A.L. Gray and P.J. Farris-Manning [10].

In 1991, the European Space Agency ESA launched the ERS-1 remote sensing satellite, carrying beside a number of other sensors a C-band SAR system. This was the starting point for numerous advances in SAR interferometry [11]-[14]. ERS-1, originally designed mainly for oceanographic applications, quickly turned out to possess great potential for the generation of large scale DEMs of high precision from areas all over the world. Main reasons for this have been the continuous and nearly global mapping performed by ERS-1, combined with a very good orbit restitution necessary for operational DEM generation. ERS-2 followed ERS-1 in 1995 to continue its operation; and, for a while during the `TANDEM' mission period, it was possible to operate both in parallel. In 1994, two missions with the Shuttle Imaging Radar SIR-C/X-SAR were flown, for the first time providing spaceborne multi-frequency (X-, C- and L-band) interferometric data in a repeat-pass mode, and over some areas even fully polarimetric [15],[16]. Parts of the same instrument, augmented by a second antenna mounted on a 60m boom, was used again in February 2000 for the Shuttle Radar Topography Mission (SRTM). Its purpose was to generate a high-resolution global DEM with single-pass interferometry within 60$^\circ$N and 60$^\circ$S [17].

Additionally to these and some other spaceborne sensors, also several airborne single-pass interferometric systems are in use. Some significant examples currently are the NASA/JPL TOPSAR system (C- and L-band) in the US; the Japanese NASDA/CRL airborne SAR (X- and L-band); the EMISAR system (C-band) of the Danish Center for Remote Sensing; the commercial German AeS-1 (X-band) of the company AeroSensing; and the E-SAR system of the German Aerospace Center (DLR) which operates in X-band in a single-pass mode and in L- and P-band fully polarimetrically in a repeat-pass mode [18]-[21].

Besides topographic mapping, an extended version of SAR interferometry, called differential interferometry, can be used for precise mapping of elevation changes. This technique allows the detection of surface deformations on a scale smaller than the radar wavelength, usually in the millimetre range. This extremely high precision enables a large scale detection and monitoring of ecological stress-change processes like sudden co-seismic displacements and long-term tectonic movements with spaceborne sensors. Also volcanic bulging before eruptions, land subsidence in mining areas, land sliding in mountainous areas as well as ice deformations and glacier dynamics can be detected with this method [22]-[27].

Interferometric SAR data additionally have content of a different nature than SAR images alone. The correlation or coherence between two SAR images is very sensitive to changes in the arrangement of the scatterers inside the resolution cells. Particularly, the coherence of multi-temporal, multi-frequency or multi-polarised repeat-pass data sets can be used to analyse and characterise changing processes, for example taking place in vegetation layers [28],[29], or on natural surfaces [30]. Finally, multi-baseline approaches take into account influences of the imaging geometry on the interferometric coherence and even can resolve to a limited extent for the spatial distribution of scatterers in a volume [31],[32].

SAR polarimetry

SAR polarimetry (POLSAR) is another major extension of conventional single channel SAR imaging. Like all electromagnetic waves also microwaves have a vectorial nature, and a complete description of the scattering problem in radar science requires a vectorial matrix formulation. This is the task of radar polarimetry, a technique which was initiated by the introduction of the concept of the `scattering matrix' by G.W. Sinclair in 1948 [33],[34]. Since radar polarimetry requires advanced hardware devices, which have not been available in the late 1940's and the 1950's, radar polarimetry remained only a theoretical concept and its practical use for civil applications was not really recognised.

This situation changed at the latest in the early 1980's with the availability of polarimetric SAR data from the NASA/JPL airborne AIRSAR system, which allowed in practice the implementation of more recent works by E.M. Kennaugh, J.R. Huynen [35] and W.-M. Boerner [36]. Since then, SAR polarimetry has become an established technique in remote sensing. This was supported by the growing number of polarimetric airborne sensors like DLR's E-SAR or NASA/JPL's AIRSAR systems, providing high resolution polarimetric data in several frequency bands. Additionally, in 1994, two SIR-C/X-SAR Shuttle missions took place, recording for the first time spaceborne fully polarimetric data in C- and L-band. During the second mission this was already combined with repeat-pass interferometric data acquisition.

One special characteristic of SAR polarimetry is that it allows a discrimination of different types of scattering mechanisms. This becomes possible because the observed polarimetric signatures depend strongly on the actual scattering process. In comparison to conventional single-channel SAR, the inclusion of SAR polarimetry consequently can lead to a significant improvement in the quality of classification and segmentation results [37]-[39]. Certain polarimetric scattering models [40] even provide a direct physical interpretation of the scattering process, allowing an estimation of physical ground parameters like soil moisture and surface roughness [41], as well as unsupervised classification methods with automatic identification of different scatterer characteristics and target types [42],[43].

SAR polarimetry additionally offers some limited capability for separating multiple scattering mechanisms occurring inside the same resolution cell and can be deemed as a first step in resolving the ambiguous scattering problem in SAR, as mentioned above. With polarimetric decomposition techniques a received signal can be split into a sum of three (intrinsically four in case of SAR, but under the assumption of reciprocal symmetric backscattering the two cross-polar components are equal.) scattering contributions with orthogonal polarimetric signatures [40]. This can be used for extracting the corresponding target types in the image, even in the case that they are occurring superimposed. Also, if a signal is disturbed by undesired orthogonal contributions, in this way the relevant components can be extracted, improving results for diverse applications [44].

Finally, polarimetric SAR interferometry (POLINSAR) combines the capability of interferometry to extract height information with polarimetric decomposition techniques [45]. With POLINSAR the topographic height of the phase centre of each extracted scattering mechanism can be estimated independently. This means that a limited volumetric imaging can be achieved with this technique, allowing for example an estimation of tree heights or underlying ground topography [46]. In recent times also increased attention was paid to model-based parameter estimation based on polarimetric interferometric data. This technique tries to invert scattering models in order to determine from the measured signals the physical parameters of interest [47],[48].

The concepts of SAR polarimetry nowadays are applied in many scientific fields. Significant areas of application are particularly found in agriculture and forestry for crop monitoring, species identification and biomass estimation. Also in geology and hydrology the possibility of a characterisation of ground roughness as well as soil and snow moisture content is of great interest. Finally, in oceanography SAR polarimetry is used for monitoring of wave systems, thermal and current fronts, and for estimating ice-age and thickness in polar regions. [49]-[54].