Since 1993 ESA has been conducting several feasibility studies and breadboarding activities for the development of a 'Microwave Imaging Radiometer with Aperture Synthesis', known as MIRAS. This Earth-observation instrument is intended particularly for the measurement of soil moisture and ocean salinity on a global scale. The relevance of these two geophysical parameters has been repeatedly emphasised by the scientific community and MIRAS could eventually form the core of an ESA Earth Explorer mission devoted to the global measurement of these two parameters.
Soil moisture and ocean salinity are key parameters for the understanding of the Earth's climatology and the global water cycle. The Microwave Imaging Radiometer with Aperture Synthesis (MIRAS) works in a protected frequency band between 1400 and 1427 MHz and is designed to measure both of these parameters. The particular feature of MIRAS is that it achieves the required ground spatial resolution by employing a 'sparse antenna' (explained below) and using interferometry to synthesise a large aperture.
Although two-dimensional aperture synthesis has been used in radio-astronomy for several decades, its application to a downward-looking sensor for Earth observation is new. Because MIRAS does not work like a conventional total-power or Dicke radiometer, it differs from them in several important respects. During the MIRAS study, much has been learnt about two-dimensional aperture synthesis for Earth observation, and the results obtained with the airborne version and its calibration system thus far have demonstrated the technical feasibility of exploiting a spaceborne sensor of this type.
During the 1980s, in the framework of its Earth observation programme, NASA organised several workshops at which scientists demonstrated the roles of soil moisture and ocean salinity in the global environmental system. Passive microwave radiometry could be used to measure these two geophysical parameters, but the most suitable frequency bands were those below 5 GHz and it was difficult to achieve the required spatial resolution with an antenna of reasonable size. NASA's Goddard Spaceflight Center, in collaboration with the University of Massachusetts at Amherst and the US Department of Agriculture, proposed the use of aperture synthesis as a solution to this problem for the first time and started to build an aircraft-borne prototype to test the concept. This NASA ESTAR (Electronically Scanned Thinned Array Radiometer) sensor was designed to be an L-band hybrid real- and synthetic-aperture radiometer and the instrument's validity was demonstrated in several USDA campaigns.
In May 1991, ESA organised a Workshop on 'Advanced Microwave Radiometer Techniques' in Copenhagen within the scope of a contract with the Technical University of Denmark (TUD). Leading experts from the microwave communities in Europe and the USA discussed conical scan, push-broom and interferometric techniques in an attempt to identify new avenues of research. One of the recommendations from this Workshop was the study of a two-dimensional aperture-synthesis radiometer and its calibration system, which were little understood at that time. The first step was the building of a laboratory demonstration model and this was successfully undertaken by TUD (at X-band).
In view of the recommendations of the Copenhagen Workshop and the encouraging results of ESTAR and the TUD demonstrator, in 1993 ESA initiated the MIRAS feasibility study as a 'microwave radiometry critical-technology development' within its Basic Technology Research Programme (TRP). The contract was awarded to Matra Marconi Space (Toulouse, F).
In April 1995, the Radiofrequency Systems and the Earth Science Divisions of ESTEC co-organised SMOS, a Consultative Meeting on Soil Moisture and Ocean Salinity Measurement Requirements and Radiometer Techniques, which brought together both the international scientific community and representatives of industry. This dialogue reconfirmed the need for and importance of mapping soil moisture and ocean salinity from space.
More recently, in June 1996 on the occasion of the presentation for selection of the first Earth Explorer candidate missions, the Joint Scientific Committee for the World Climate Research Programme and the Global Energy and Water Experiment Scientific Steering Group reiterated their interest in having global soil moisture (upper 5 - 10 cm) data and expressed their wish that urgent consideration be given to an experimental soil-moisture mapping mission.
By the end of 1996, the MIRAS feasibility study had been completed and both the theoretical analysis of the spaceborne instrument and the development of an aircraft instrument had been performed. The success of the theoretical work, together with all of the theoretical and experimental advances related to the demonstrator and its calibration system, show that L-band two-dimensional aperture synthesis is now mature for large field-of-view Earth observation from space.
Several contracts have been started within the Technology Research Programme (TRP) and the General Support Technology Programme (GSTP) since the completion of the initial MIRAS study, aimed at the breadboarding of the different key subsystems, namely LICEF (Light Weight Cost Effective Antenna Front-end Assembly) by MIER (E) and DICOS (Advanced Digital Correlator Unit for Aperture Synthesis Application) by DSS (D). The breadboarding of the other three main MIRAS elements - the signal harness, the structure and mechanisms and the calibration system - is expected to start in the near future, mainly through already confirmed GSTP funding.
The coordination, integration and testing of all of these breadboarding activities, which will together produce a representative prototype of MIRAS, has been proposed as a Pilot Project within the new TRP programme, with the endorsement of the Earth Observation Preparatory Programme (EOPP). The MIRAS Demonstrator Pilot Project could facilitate the proposal by the scientific community of a candidate mission for the Earth Explorer Programme. Initial studies of such a mission concept have been proposed for EOPP Extension 2 in support of the next round of Earth Explorer candidates.
The scientific goals for MIRAS were established by CESBIO, taking into account the previous European and American work on soil-moisture and ocean-salinity measurements and the conclusions of the Copenhagen Workshop on Advanced Microwave Radiometer Techniques. The soil moisture, expressed in percentage terms, is defined as the ratio of water volume to soil volume in the first 5 cm of depth. The ocean salinity is defined in practical salinity units (1 psu = 0.1%), and ranges from 32 to 37 psu.
The choice of operating wavelength for MIRAS is determined by the increase in sensitivity of the brightness temperature to soil moisture (ground) and to ocean salinity (ocean) as the observation frequency decreases. L-band (1400 - 1427 MHz) is optimum because the frequency is sufficiently low and the Faraday rotation is still negligible (<0.2 deg under average conditions and <3.3 deg during magnetic storms). The sensitivity to other parameters diminishes accordingly, i.e. the vegetation and soil-roughness influence almost vanishes in the case of soil-moisture observation at L-band (Fig. 1). A key factor in the MIRAS specifications is the polarisation ratio's independence from the soil physical temperature, which leads to the choice of dual-polarisation measurements at ground incidence angles between 40 and 55 deg.
Figure 1. The choice of wavelengths for MIRAS is driven by the high sensitivity of L-band emission to soil moisture content. Dual-polarisation measurements remove physical temperature effects
Similarly, for the ocean-salinity measurements the impact of wind speed and ocean roughness on brightness-temperature measurements decreases strongly at a few GHz (Fig. 2). The water temperature's influence also diminishes at low frequencies. Nevertheless, it is useful to obtain the water temperature by using other sensors.
Figure 2. Over the ocean, wind and surface-temperature variations are small at L-band. Still an independent temperature measurement is required for accurate salinity retrieval
The scientific requirements for MIRAS are summarised in Table 1. The performance figures achieved after the various trade-offs are indicated in grey. The 3K brightness temperature accuracy, with 20/50 km ground resolution and three-day revisit cycle, are consistent with soil-moisture applications. The ocean data will be spatially and temporally averaged (in 1 deg x 1 deg zones, i.e. 111 km x 111 km, over 1 month) in order to provide the climate community with the data that it needs.
Table 1. MIRAS scientific requirements
H or V
0.9K, above ground
0.5K, above ocean
The spaceborne instrument collects the flux radiated by the Earth's surface via an antenna array tilted by 31.2 deg with respect to nadir (Fig. 3). This array consists of 133 elements with 70 deg half-power beamwidth distributed along three equi-spaced, 8.3 m-long coplanar arms. They operate in dual linear polarisation (horizontal and vertical). Each antenna is connected to an MMIC L-band receiver which amplifies and down-converts the H and V signals sequentially (Fig. 4). After one-bit digitisation (sign operator), the baseband digital signals are routed to the correlators via optical fibres, to minimise the effects of gain and phase drifts. The 8778 correlations, which are simply the complex multiplication and integration of any pair of receiver outputs, are performed by 1 bit digital correlators implemented in a single piece of equipment. MIRAS weighs 230 kg and consumes 300 W of power.
Figure 3. Artist's impression of MIRAS aboard an Envisat-type platform
Figure 4. Functional block diagram of the MIRAS instrument
The aperture-synthesis technique
The requirements for passive imaging from space of soil moisture and ocean salinity at L-band (lambda=21 cm) lead to large antenna apertures (up to 20 m) for which 'thinned' arrays using aperture-synthesis principles offer clear advantages compared with mechanically or electronically steered antennas and push-broom radiometers. A two-dimensional interferometric radiometer like MIRAS achieves the greatest degree of antenna thinning.
Several T- and U-shaped configurations with regularly spaced elements were studied for the MIRAS array. Early in the MIRAS study, the T-shape approach suited for space-platform accommodation was merged with the design used for the Very Large Array (VLA) radio telescope in New Mexico, namely a Y-shape with an exponential element distribution. The result was a Y-shaped configuration with equally spaced antenna elements, which has important advantages in terms of ground resolution and grating lobes and is suitable for spaceborne applications (Fig. 5).
Figure 5. Left, the Y-shaped MIRAS array; right, the array's spatial frequency coverage; centre, circle of spatial frequencies in standard resolution mode
The complex correlation (at zero delay) between each possible pair of antenna elements of the interferometric array gives a point of the so-called 'visibility function' at a spatial frequency defined by that particular antenna element baseline. The visibility function is ideally the Fourier transform of the brightness temperature of the scene, weighted by the element gain pattern, which can then be recovered via an inverse Fourier transform.
MIRAS's novel Y-shaped antenna achieves a dense sampling of the visibility function for a given spatial resolution, the larger spacing between antenna elements compared to rectangular arrays leading to savings of 13% in antenna/receiver elements and 28% in the number correlators needed.
MIRAS's useful swath is a defined region more than 900 km wide, comprising all incidence angles between 40 and 55 deg (green domain in Fig. 6). Its element spacing has been increased beyond the strictly alias-free limit to the point where only the useful swath remains alias-free, resulting in further hardware savings (Fig. 6).
Figure 6. The useful field of view of MIRAS represented by the alias-free green box; the red contour is the Earth's disk, and the blue lines represent the Earth disk alias
In a standard-resolution mode (only the visibility-function samples inside the small circle in Fig. 5 are taken), the antenna half-power beam width is 1.2 deg, corresponding to a worst-case ground resolution of 35 km. Even with this narrow reconstructed beam, the beam efficiency of MIRAS is better than 93%.
The radiometric sensitivity of the MIRAS snapshot is greater than that of the corresponding conventional real-aperture radiometer by the array filling factor. This loss of sensitivity is, however, recovered via the larger integration time available in the case of the interferometric radiometer. MIRAS's sensitivity is 0.9 K over land and 0.5 K over the ocean.
Antennas and receiver chains are likely to be affected by phase errors. Their impact on the visibility function is partially cancelled thanks to the phase closure properties which allow the phase of the visibility samples to be recovered via a phase-restoration algorithm. The effect of in-plane array deformation is also cancelled by this means.
The second half of the MIRAS contract was devoted to a demonstration of the spaceborne sensor's capabilities by flying a scaled-down instrument on an Hercules C-130 aircraft. The central core of this MIRAS breadboard is identical to that of the spaceborne instrument (Fig. 7), with a Y-shaped antenna with 0.65 m-long arms and 11 antenna elements in total. One additional element outside the Y-array provides the phase-restoration capability. The aliasing conditions onboard an aircraft are different from those on a spacecraft, but an aliasing-free region exists in MIRAS demonstrator field of view for the 50 deg ground incidence cone when the antenna array is tilted by 45 deg from nadir. Both V and H polarisations are sent alternately to the receivers. Every 3 or 5 s, according to operator selection, calibration signals are also presented at the receiver's input ports: either independent (uncorrelated) loads supposed to produce a zero output, or correlated signals supposed to produce a known output.
Figure 7. An 11-element L-band MIRAS aircraft breadboard suitable for a Hercules C-130
The MIRAS airborne demonstrator has addressed several system issues as well as critical technologies which are briefly discussed below. The calibration system has also been investigated using the demonstrator, and is addressed separately below.
Figure 8. Measurement of spatial resolution and beam efficiency
Figure 9. Testing of the phase-restoration algorithm using a noise source in the near field of the MIRAS breadboard
A near field point source generates aberrations in the reconstructed image (left image). The near field influence, in the case of a point source, is mainly a phase error applied to each antenna receiver chain. This phase error is fully corrected by the phase restoration algorithm, as demonstrated by the right-hand image. A phase offset remains (displacement of the reconstructed spot) which can be calibrated out
Figure 10. Measurement of the antenna pattern of the MIRAS breadboard simulating the side walls of its box
Figure 11. The microwave receiver unit, 11 of which were manufactured
The correlation measurement is based on a 1-bit by 1-bit multiplication, sampled at the Nyquist rate of the incoming video signal (> 50 MHz) and the result accumulated during a (snapshot) integration time of 0.3 s. The 300 ms clock is generated by the correlator unit. The correlation results, which constitute the visibility function, are provided in a digitised form. The correlator unit also measures, averages and digitises the Power Measurement Signal (PMS), which provides the total brightness temperature of the scene. The integration time is synchronised with the 300 ms clock. The visibility and PMS data, together with the housekeeping data, are multiplexed and transmitted to the data-management unit.
The two critical performances of the correlators are the noise and the accuracy. As demonstrated in the 1-bit correlator theory, the signal-to-noise ratio (S/N) is equal, for small correlations, to 0.64 times the analogue correlator S/N. MIRAS correlators fit perfectly with the theory, with no additional noise being observed. As far as accuracy is concerned, it was found during the measurements that the major contributor is the test setup digital voltmeter linearity. Measured with independent noise input signals, some correlation offset errors are as low as 10-4, whilst others reach 3 x 10-3. This has a negligible impact on MIRAS performances (Fig. 12).
Figure 12. Measured performance of the correlator unit of the MIRAS breadboard (correlation error after offset calibration is less than 10-4
In the MIRAS demonstrator, the image-reconstruction process assumes identical antenna patterns and identical transfer functions for the 11 receiver chains. The mutual coupling between antennas is assumed to be negligible, as well as the decorrelation effect at large viewing angles from the boresight (the so-called 'fringe wash factor'). Those assumptions enable a straightforward image reconstruction and give a very good result in the case of the MIRAS demonstrator. In a real instrument, the receiver bandwidth centre frequencies and group delays are slightly different, the antenna radiation patterns are not identical, and the antenna mutual couplings are not null. All of those discrepancies have to be taken into account to achieve the ultimate image accuracy. Such a study has been accomplished in parallel with the MIRAS demonstrator's development, and the results are briefly presented below. In the meantime, a powerful Fourier transform using hexagonal sampling has been proposed based on methods available in the literature.
Identification of instrument errors
The instrument errors that can play a role have been classified into three categories:
Hardware calibration system
A centralised common noise-source reference as implemented in the MIRAS demonstrator enables the calibration of both the receiver and baseline errors. However, such a coaxial distribution network is critical in the case of the spaceborne instrument because of mechanical constraints at arm hinges. The phase stability of the reference signal is also questionable. An alternative calibration system is therefore proposed for the spaceborne instrument based on a distributed network in which several reference sources feed overlapped groups of antenna/receiver chains (Fig. 13). It achieves the calibration of all receiver errors and the separable part of the baseline errors. The non-separable contribution in the baseline errors cannot be calibrated out with the distributed noise injection and must be minimised and bounded by the error in estimating the visibility amplitude and phase.
Figure 13. Distributed network of correlated overlapping noise sources for MIRAS calibration
In addition to the correlated reference signal, another calibration signal is provided to calibrate out correlation offsets. This signal is generated by independent matched loads connected at the time of calibration to the input of each receiver.
From the results obtained with the airborne demonstrator and the related studies, the best calibration system for the spaceborne MIRAS instrument is one based on the injection of correlated and uncorrelated noise, followed by the phase restoration technique.
The baseline image-reconstruction algorithm assumes identical antenna patterns and receiver frequency responses. An iterative algorithm has been defined to solve the equation set describing the complete system. As an example, a 31-antenna MIRAS-type radiometer has been simulated with antenna and receiver discrepancies. A synthetic scene is perfectly reconstructed after just four or five iterations (Fig. 14). Despite the system complexity, the spatial and radiometric resolution are not degraded compared to the inverse Fourier transform used for the ideal system.
Figure 14. Iterative image algorithm tested on a synthetic scene (Cuba and Florida)
Within the scope of the MIRAS contract, two further studies have also been performed which are worth mentioning here: one covering a dual-frequency MIRAS, and one covering a MIRAS demonstrator for small-satellite applications.
While a single-frequency, dual-polarisation L-band MIRAS sensor is capable of providing soil-moisture observations on its own, retrieval of ocean salinity is possible only with an independent measurement of the sea-surface temperature. In order to provide that indepen-dent observation, a dual-frequency MIRAS has been studied. The scientific analysis has shown that a C-band channel can do the job and both the antenna array and the receiver architecture have been analysed. Once again, an extremely convenient array solution for space-borne applications has been found (Fig. 15).
Figure 15. Dual frequency MIRAS array. The C-band interferometer is also full-phase with an homothetic spatial frequency coverage as compared to the L-band array and the same spatial resolution
The C-band array can be easily accommodated mechanically while preserving all of the required interferometric and performance features of the L-band array. The overall mass and power consumption of the dual-frequency MIRAS have been estimated at 300 kg and 600 W, respectively.
MIRAS demonstrator for a small satellite
The main MIRAS contract focused on compatibility of the instrument with an ESA platform of the Envisat type. A short study has been carried out of a reduced MIRAS sensor suitable for small-satellite applications. With an arm length of just 5 m, it is still able to provide valid scientific data at 50 km ground resolution from a 672 km-high Sun-synchronous orbit. Its compact configuration when stowed allows small commercial launchers to be used. Compatibility of this sensor with the Minisat (E) platform has been preliminarily assessed (Fig. 16). It has been proposed that the array geometry and pointing be measured in-flight by placing GPS antennas at the tips of the antenna arms and applying double-difference carrier phase processing. The estimated mass and power requirements are 180 kg and 230 W, respectively.
Figure 16. Presentation of the MIRAS demonstrator on the Spanish Minisat at the International Conference on Small Satellites: Missions and Technology, in Madrid in September 1996 (Mockup courtesy of Sener, Spain)
With MIRAS being based on interferometric principles, much research has been performed at all subsystem levels, from the key geometry of the antenna array to the calibration system for correlated/uncorrelated noise injection, from the theoretical fundamentals to the image processing through iterative methods based on Fourier transformation over hexagonal sampling grids. The airborne MIRAS demonstrator that has been built and tested has performed successfully and in accordance with the predictions. A great deal has been learnt and the results obtained so far with the airborne instrument and its calibration system have shown that L-band two-dimensional aperture synthesis is now a mature technology for large field-of-view Earth-observation from space, paving the way for a candidate Earth Explorer mission devoted to soil moisture and ocean salinity.
The MIRAS activities have been performed within ESA's Basic Technology Research Programme under the Prime Contractorship of Matra Marconi Space (F), which was responsible for the feasibility study of the spaceborne instrument and the design, manufacture and testing of the airborne instrument. The subcontractors were CESBIO (F) for the scientific aspects, DSS (D) for the correlator and thermo-mechanics and a contribution to the system study, LAT/CERFACS (F) for the image reconstruction, MMS (UK) for the microwave receivers, ORS (A) for the breadboard structure, TUD (DK) for the software and experiments, and UPC (E) for the instrument fundamentals, system error calibration and image reconstruction.
The authors would like to acknowledge the efforts of D. Maccoll and A. Resti (at ESTEC) who first initiated the study of MIRAS, and of the whole industrial team, in particular Y. Kerr (CESBIO), Dr. O. Bätz and Dr. U. Kraft (DSS), A. Lannes and E. Anterrieu (LAT/CERFACS), J.C. Orlhac (MMS-F), D.J. Adlam and A.J. Knight (MMS-UK), N. Skou and B. Laursen (TUD), J. Bará, I. Corbella, A. Camps and F. Torres (UPC). The contributions from outside the MIRAS contract from J. Ortiz (INTA) and M. Sierra (Sener) are also gratefully acknowledged. Last but not least, we would like to thank E. Attema (at ESTEC) for his support as co-organiser of the SMOS Consultative Meeting.