814 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 49, NO. 2, FEBRUARY 2011 Experimental Study on Radiometric Performance of Synthetic Aperture Radiometer HUT-2D—Measurements of Natural Targets Juha Kainulainen, Kimmo Rautiainen, Juha Lemmetyinen, Jaakko Seppänen, Pauli Sievinen, Student Member, IEEE, Matias Takala, and Martti T. Hallikainen, Fellow, IEEE Abstract—This paper describes the analysis of L-band radiometric measurement data gathered with the synthetic aperture radiometer HUT-2D during several ground-based and airborne measurement campaigns. The radiometric data are analyzed from the instrument’s performance point of view, aiming to verify the theoretical performance of an instrument of this kind and to assess the performance of the HUT-2D radiometer system in particular. The data sets considered for the study consist of measurements of well-known natural targets, such as cosmic background radiation, and measurements of pure water scenes, the brightness temperature of which is possible to model based on in situ measurements. We define four figures of merit, which are applicable for synthetic aperture radiometers. These are radiometric resolution, image bias, pixel-to-pixel random error, and temporal stability. Then, we use the selected data sets to assess these in the case of HUT-2D. The experimental results are discussed and compared to the theoretical values, where applicable. Also, we discuss possibilities to improve the presented performance. The main results of this paper are the consolidated performance parameters of the HUT-2D instrument. We study and discuss the properties of the error components related to the technology in a general level, and study the scalability of the errors as a function of the measured targets. In particular, the stability of the direction-dependent error component is pointed out, and a mitigation guideline is proposed. Index Terms—Interferometry, radiometry, remote sensing, synthetic aperture imaging. I. I NTRODUCTION HE SYNTHETIC aperture radiometer HUT-2D, designed and manufactured by the Helsinki University of Technology (TKK), has made demonstrations of its remote sensing capabilities during the past years [1]–[3]. An important step in the evaluation of the instrument’s usability and applicability, as well as the whole concept of synthetic aperture radiometry, is to gather practical experience of the radiometric performance provided by the technology. T Manuscript received December 15, 2009; revised April 13, 2010 and June 3, 2010; accepted June 26, 2010. Date of publication September 27, 2010; date of current version January 21, 2011. This work was supported in part by the Graduate School in Electronics, Telecommunications and Automation and in part by the Department of Radio Science and Engineering of the Aalto University School of Science and Technology. J. Kainulainen, J. Seppänen, P. Sievinen, and M. T. Hallikainen are with the Department of Radio Science and Engineering, Aalto University School of Science and Technology, 02150 Espoo, Finland (e-mail: juha.kainulainen@tkk.fi; jaakko.seppanen@tkk.fi; pauli.sievinen@tkk.fi; martti.hallikainen@tkk.fi). K. Rautiainen, J. Lemmetyinen, and M. Takala are with the Arctic Research, Finnish Meteorological Institute, 00101 Helsinki, Finland (e-mail: kimmo.rautiainen@tkk.fi; juha.lemmetyinen@fmi.fi; matias.takala@fmi.fi). Digital Object Identifier 10.1109/TGRS.2010.2061857 The main characteristics of the synthetic aperture radiometer HUT-2D are designed to be similar to those of the European Space Agency’s (ESA) Soil Moisture and Ocean Salinity (SMOS) mission scientific payload, i.e., Microwave Imaging Radiometer using Aperture Synthesis (MIRAS) [4]. Therefore, the HUT-2D instrument was developed in close co-operation with ESA, giving practical experience on the development of synthetic aperture techniques and calibration algorithms together with other early SMOS demonstrators, such as airborne AMIRAS and 2D-STAR [5], [6]. Also today, after completion in 2006 and being one of the few complete airborne instruments of its kind, HUT-2D data analyses and performance studies will help to develop the essential framework required in the further development of this novel technology. Despite the early demonstrators, there is a lack of experience of the radiometric performance of a complete operating synthetic aperture radiometer. In this paper, we address a number of fundamental performance parameters used to characterize radiometric measurements, and analyze them from the end products of the HUT-2D instrument. Namely, we define and study the following figures of merit for the use of synthetic aperture radiometer performance analysis [7]: 1) radiometric resolution; 2) image bias; 3) pixel-to-pixel random error; 4) instrument stability. We present briefly the principle of synthetic aperture imaging and the HUT-2D radiometer system in Section II. In Section III, we discuss the requirements of experimental determination of the figures of merit and discuss the suitable natural targets available. The HUT-2D data sets selected for the study are described in Section IV, and the actual performance analysis is detailed in Section V, where we also compare the results with theoretical expectations and summarize the different performance parameters. Finally, we discuss the main error sources and possibilities to decrease their impact, and estimate the usability of the results for the performance studies of other synthetic aperture radiometers, such as MIRAS. II. S YNTHETIC A PERTURE R ADIOMETER I MAGING W ITH HUT-2D S YSTEM A. Imaging Using Interferometric Aperture Synthesis An imaging radiometer using aperture synthesis measures the spatial frequency components of the brightness temperature 0196-2892/$26.00 © 2010 IEEE KAINULAINEN et al.: EXPERIMENTAL STUDY ON PERFORMANCE OF SYNTHETIC APERTURE RADIOMETER HUT-2D distribution in its field of view (FOV) using several antenna pairs, i.e., the so-called baselines. A measurement of each baseline is described by the visibility equation presented in detail in [8]. A two-dimensionally populated receiver constellation establishes several baselines and measures a 2-D spatial spectrum of the brightness temperature distribution. From the 2-D spectrum, a 2-D brightness temperature image, i.e., a “snapshot,” can be retrieved with proper inversion algorithms. Such processes are discussed, e.g., in [9] and [10]. B. HUT-2D Instrument The HUT-2D instrument is described in detail in [1]. In short, it samples the visibility function using 36 receivers in a 2-D U-shaped configuration over a 7-MHz band centered at 1.4135 GHz. The instrument is designed for airborne use, and for several test campaigns, it has been mounted on the Short SC-7 Skyvan research aircraft of the Helsinki University of Technology. Calibration of HUT-2D is based on correlated noise injection to all receivers from a single thermally stabilized noise diode unit. For more details on HUT-2D calibration, the reader should review [1] and [11], with the latter describing the calibration of MIRAS in detail and being applicable to HUT-2D calibration in many contexts. Each of the HUT-2D receivers has two orthogonally polarized antenna feeds, namely, X- and Y-probes, which are measured in turns switching typically at 1-Hz frequency between them. Due to 2-D sampling of the visibility domain, the reconstructed output brightness temperature is also a 2-D image of the target. In nominal operation mode, HUT-2D measures one 2-D image, or snapshot, per 250 ms, which is the integration time of the radiometer. The image is retrieved in a rectangular grid in the antenna reference frame, which is typically defined with direction cosines (ξ, η). The brightness temperature image is disturbed by image replicas, i.e., aliases, caused by the undersampling of the visibility domain. The area, which remains between the main replicas, is called alias-free FOV (AF-FOV), and it is used for scientific data retrieval [4]. Fig. 1 (left) shows the imaging geometry in the antenna frame of HUT-2D. The rectangular image of the antenna frame is rectified into the ground reference frame using the attitude and position information of the aircraft and a digital elevation model of the target area. In this paper, we use measurement results only in the antenna reference frame, since the performance of the instrument is related closely to the separate antenna feeds rather than ground frame polarizations. For the reader’s convenience, the projection of the antenna frame geometry on the ground reference frame from a flight altitude of 1000 m is shown on the right in Fig. 1. The image reconstruction process of HUT-2D utilizes the flat target transformation (FTT), which incorporates a measurement of a known source, i.e., the instrument’s flat target response (FTR), to the image reconstruction process. Typically, a measurement of galactic microwave background radiation is 815 used for the purpose. This process is formulated in [12]. It is also discussed in [12] that the processing should be efficient in correcting the direction-dependent error component of the 2-D output images. In [12], the performance is demonstrated with galactic measurements. In this paper, we expand this evaluation to measurements of water scenes by comparing water measurements of HUT-2D obtained with and without FTT processing. In addition to this, we introduce an experimental correction method for further evaluation. Alternative and improved image reconstruction methods have been proposed for synthetic aperture radiometry, e.g., the ones presented in [10] and [13]. However, they are not implemented for use in HUT-2D data processing for reasons that will be discussed later in Section III-C. III. F IGURES OF M ERIT IN P ERFORMANCE A NALYSIS We study the performance of the synthetic aperture radiometer HUT-2D using the four figures of merit. In the following, we define these parameters and discuss their experimental determination. Also, we assess the applicability and availability of different natural sources for this purpose. A. Radiometric Resolution Radiometric resolution describes the temporal variance of the instrument’s output, which, in the case of synthetic aperture radiometers, is the brightness temperature snapshot (see Fig. 1). This figure is finite essentially due to the noisy nature of the measured signal: Finite integration results in variance between measurements of the same target. According to Camps et al. [14], the radiometric resolution of synthetic aperture radiometers can be approximated as follows: σTB (ξ, η) =  ΩA αol  TA + TR ·A √ αw NV αf BηC τ 1 − ξ2 − η2 (1) where ΩA is the antenna solid angle (2.1 sr); A is the area of one pixel in the visibility domain (0.72 ); TA and TR are the antenna and receiver noise temperatures (320 K–450 K, depending on the physical temperature of receivers), respectively; B is the equivalent noise bandwidth (7 MHz); τ is the integration time (250–4000 ms); ηC is an efficiency factor of a digital correlator (1/2.46 for 1-b/two-level correlators [15]); NV is the number of measured visibility samples (575); and αw , αol , and αf are the windowing, local oscillator, and filter factors (0.45, 1, and 1.19), respectively. The values in parentheses indicate the parameter values established for HUT-2D. The noise power of (1) is not equally distributed in the 2-D brightness temperature image due to correlation of visibility noise. The noise level increases with the magnitude of the measured signal. However, when a constant brightness temperature distribution is measured, noise is approximately equal in every direction [16]. Following this, the radiometric resolution of a synthetic aperture radiometer is possible to analyze from measurements of constant, or flat, brightness temperature distributions, which are stable in time. 816 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 49, NO. 2, FEBRUARY 2011 Fig. 1. Illustration of the imaging area of HUT-2D in the (ξ, η)-frame, i.e., (left) the antenna reference frame, and in (right) the ground reference frame from a flight altitude of 1 km. In both figures, the AF area is enclosed at the center between the bold asymptotes. The circles in the right figure indicate the incidence angles of the measurement geometry. The brightness temperature of the sky at L-band has been measured in radio astronomy to good accuracies [17]. Also, the emission and transmission loss of the atmosphere at L-band can be accurately modeled [18]. The brightness temperature of the galactic pole area is also very flat, which should distribute the thermal noise power in the measurement evenly over the image. There are strong radio sources outside the galaxy, namely, quasars, appearing in the galactic pole areas. These are, however, pointlike sources, and their impact is attenuated to a negligible level due to the coarse angular resolution of remote sensing synthetic aperture radiometers [19]. The brightness temperature of water is also possible to model at L-band, e.g., using models described in [20]. Accordingly, the brightness temperature is mainly a function of sea surface temperature (SST) and sea surface salinity (SSS). Sea surface roughness has also an impact on the apparent emissivity, but in very low wind conditions, which we consider in this paper, the effect is small [21]. The brightness temperature is different on H- and V-polarizations, depending on the viewing, or incidence, angle. The sky and water views stand for low-emissivity sources, with the former resulting in approximately 6 K and the latter in 120-K antenna temperatures for both of the HUT-2D antenna probes. In order to examine the performance also at the high end of the operation range, we study a measurement acquired over a homogenous coniferous forest area. As described, e.g., in [22], the typical emissivity of such a forest is on the order of ∼0.9. It is understood that the emissivity of the source is not well known in this case, but since the sensitivity of (1) to an error in antenna temperature is low, we can study the radiometric resolution, even though the antenna temperature is not perfectly known. For the study, we approximate that the antenna temperatures of HUT-2D are at a level of 250 K during forest measurements. This value is established by assuming a constant emissivity of 0.9 for the test area, as suggested by tower-based measurements in [22]. In this paper, we use these three natural sources, identified as galactic pole, open water, and forest views, to study experimentally the radiometric resolution of the HUT-2D instrument. From a measurement including several snapshots, we calculate the radiometric resolution image as the temporal standard deviation of the measured images T̂B (ξ, η, t)      N t T̂ (ξ, η, t ) − T̂ (ξ, η) 2  B i B ′ ∆TSEN (2) (ξ, η) ≡  Nt − 1 i=1 where . . . denotes the temporal average and N t represents the number of snapshots that the data consist of. Note that (2) results in a 2-D image in the antenna reference frame. To obtain a single value describing the radiometric resolution of the measurement, we calculate and present the average value of the radiometric resolution in the AF-FOV area of the instrument (see Fig. 1, left) ∆TSEN ≡ Np  ∆T ′ SEN (ξi , ηi ) i=1 Np (3) where N p is number of pixels in the AF-FOV area, and each point (ξi , ηi ) belongs to the AF-FOV area. It is noted that the calculation of radiometric resolution using (2) may be affected by the temporal drift of the instrument’s output. To avoid this, we calculate the experimental radiometric resolution from reasonably short data acquisitions. The suitable time will be discussed later in Section V-D, where we conclude that the resolution can be reliably calculated from a 30-s acquisition without being affected by instrumental drifts. B. Image Bias Image bias describes the offset component, or the bias, which is a common error component to all the pixels within the measured image. It is the difference of the average brightness temperature level of the measured scene (T̂B ) and the one actually caused by the target. Here, we define the image bias to be the difference of the measured and expected [TBmodel (ξ, η)] spatial brightness temperature averages in the AF-FOV area ∆TBIAS ≡ T̂B (ξ, η) − TBmodel (ξ, η) (4) KAINULAINEN et al.: EXPERIMENTAL STUDY ON PERFORMANCE OF SYNTHETIC APERTURE RADIOMETER HUT-2D where the overbar stands for the spatial average over the AF-FOV area. The expected brightness temperature is obtained by convolving the high-resolution brightness temperature model with the theoretical unit response function of the instrument. In synthetic aperture radiometry, image bias is closely related to the determination of the so-called zero-baseline visibility, which is the value of the visibility function in the origin. It determines the overall level of brightness temperature snapshots, and according to , eq. (29)[8], it becomes V (0, 0) = TA − Tphys 817 TABLE I HUT-2D DATA S ETS U SED IN E XPERIMENTAL P ERFORMANCE A NALYSIS (5) where TA is the antenna temperature and Tphys is the physical temperature of receivers. According to this, uncertainty in antenna temperature determination is directly propagated into the uncertainty of the zero-baseline visibility. Furthermore, this translates into uncertainty of the brightness temperature level, i.e., it creates a bias into the snapshots. Experimental determination of the image bias is more challenging than determination of the radiometric resolution, since it is closely affected by the antenna temperature through (5), which, in turn, is affected by the complete brightness temperature distribution in the front (and even back) hemisphere of the instrument. In addition to the accurate knowledge of the target emission, a good estimate of emission from every direction is required in order to make the uncertainty of the modeled average in (4) reasonably low. Determination of the image bias experimentally requires well-known sources, which fill properly not only the AF-FOV area of the instrument (approximately ±30◦ for HUT-2D) but also a maximal portion of the front hemisphere. Also, the target’s emissivity at different incidence angles must be well known. Here, we analyze the image bias from several measurements made in nominal airborne operation mode of HUT-2D, when pure water scenes are observed. For these measurements, we require the condition of not having land in the vicinity of the instrument in order to prevent their influence through antenna sidelobes. In practice, this requirement is fulfilled by flying at low altitudes over the water area concerned. Also, we require that the wind conditions at the moment of measurement be light so that the impact of roughness on brightness temperature can be omitted. Also, we analyze the bias from measurements of the galactic pole area. characterization and errors in receiver system temperature calibration can be considered to be major sources of errors. The origin of visibility errors and their impact on output brightness temperature are studied in the frame of the SMOS mission, e.g., in [23]–[25]. A direction-dependent error component can be introduced also through the Gibbs phenomenon related to synthetic aperture radiometers. The phenomenon intensifies with the increasing amount of high-frequency components, i.e., sharp changes, in the brightness temperature of the whole FOV. For example, in the case of SMOS, a sharp brightness temperature change occurs at the edge of the visible crest of the Earth against outer space, which is in the instrument’s FOV due to positive pitch angle during measurements. Also, transitions between land and water create strong Gibbs effects. Various methods to mitigate this are proposed, e.g., in [13] and [26]. In the case of HUT-2D, which is a nadir-pointed instrument flying at significantly lower altitudes than SMOS (typically 300–3000 m), sharp brightness temperature changes do not appear in FOV as often. The Earth–sky border is very far near the edge of the front hemisphere, and the test areas are very seldom in the vicinity of coastlines. Because of these reasons, the influence of the Gibbs phenomenon is of second order in terms of HUT-2D errors, and mitigation algorithms are not implemented for nominal processing. Here, we define the pixel-to-pixel random error to be the root-mean-square (rms) error of a measurement, from which the offset, or image bias, component defined earlier in Section III-B is compensated. The random error is calculated using only the pixels (ξi , ηi ) in the AF-FOV area   ⌢  Np  T B (ξi , ηi ) −∆TBIAS −TBmodel (ξi , ηi ) ∆TRE ≡  Np i=1 . (6) C. Pixel-to-Pixel Random Error Pixel-to-pixel random error describes the smallest change in brightness temperature that can be detected from the 2-D output image spatially, i.e., it is the random spatial error component. The pixel-to-pixel random error component of a synthetic aperture radiometer is caused by the amplitude and phase errors of the measured visibility samples. These errors cause direction-dependent errors in the brightness temperature snapshots, and they originate from the instrument’s nonidealities and inaccurate calibration. For example, errors in antenna pattern 2 Here, we want to make a clear difference between temporal and spatial performances. Thus, we use the temporal average ⌢ of the measurement, i.e., T B , in the definition to stress that the measurement must not be influenced by thermal noise due to the finite integration time. To achieve this, the natural source needs to be measured for a sufficient period. This yields to a requirement of a temporally constant source. In a static ground-based measurement, the galactic pole area can be measured for a long time during the night, depending on 818 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 49, NO. 2, FEBRUARY 2011 Fig. 2. (Left) Model of the radiation of the sky during FTR measurements and measurements of the galactic pole. (Middle) FTR of the HUT-2D system’s X-probe. (Right) Measurement of the galactic pole area with the HUT-2D system’s X-probe. the season and geographic location (latitude). In airborne operation, however, the requirement is more difficult to fulfill. In order to achieve a reasonable integration time, the homogenous source must have a large dimension so that the target remains in the instrument’s FOV for the duration of the required measurement period. Such large homogenous areas can be found at sea, with the emissivity of the sea surface being determined primarily by SST, SSS, and surface roughness. Other possible natural targets for airborne operations are, e.g., homogenous forested areas, where emissivity is dominated by the biomass of vegetation. Here, we calculate the experimental pixel-to-pixel random error of HUT-2D using measurements of two sources, namely, the galactic pole area and the open water area. In addition, we present a demonstration of the instrument’s ability to measure spatially separated small differences in the brightness temperature of the target. This is done using a measurement of the galactic plane area. The galactic plane radiates more strongly at L-band than the surrounding regions, as the radiative medium in the galaxy is strongly concentrated in the plane. In the angular resolution of HUT-2D (∼10◦ ), the galactic plane should appear as a band that is brighter than its surroundings. D. Temporal Stability An instrument’s ability to detect two different brightness temperature levels in different moments of time depends on radiometric resolution σTB and temporal stability, which, in the case of synthetic aperture radiometers, can be split into stability of image bias and random error. The influence of radiometric resolution becomes negligible after averaging the measurement in the time domain, i.e., increasing the integration time of the measurement. After a certain point in time, this increment will no longer enhance the accuracy of the measurement, since the drifts in bias and random error become the dominant sources of error. Here, we study the instrument’s stability using the twosample variance, or the so-called Allan variance, which is the IEEE standardized definition of time stability [27] σAllan    (τ ) ≡  N τ −1 1 (TB τ,i+1 − TB τ,i )2 2(N τ − 1) i=1 (7) where TB τ,i is the ith sample of the brightness temperature obtained using integration time τ and N τ is the number of τ –length intervals within the complete test time. The analysis requires a long measurement—as long as it takes for instrumental drifts to become dominant over temporal sensitivity. To obtain a 1-sigma confidence for the variance, a measurement of approximately eight times longer than the stability time is required, as the uncertainty follows the square root of the sample numbers. To cope with the drifts that start to dominate, a new calibration of the instrument is performed. The time of recalibration is established by the accuracy of calibration. This accuracy can be studied by analyzing the instrument’s measurement over the same target, applying a new calibration for every measurement. In this paper, we assess the accuracy of HUT-2D calibration using multiple measurements of a water area. It is assumed that the stability of the instrument is strongly driven by the ambient thermal environment. Hence, we study the stability only in the instrument’s nominal operation environment, i.e., the aircraft installation, which establishes unique and dynamic thermal conditions for the instrument. IV. M EASUREMENT DATA S ETS We analyzed HUT-2D data from several airborne and ground-based test campaigns according to the reasoning presented in the previous section. The test setups selected for this analysis are summarized in Table I. The table shows also the experimental performance parameters analyzed from each measurement. We also present the measured FTR of the instrument, from which we draw some conclusions about the performance of the instrument and sources of uncertainty. Next, we briefly describe each of the data sets presented in Table I. A. FTRs and Galactic Pole Measurements For the measurements of the instrument’s FTR and the measurement of the galactic pole area, an HUT-2D was installed outside, pointing toward the east, 25◦ off the zenith. After sunset, the instrument measured continuously overnight, alternating between the X- and Y-probes. A forward model for the brightness temperature of the sky was established using [17] and [18]. The outcome of these models applied in the actual KAINULAINEN et al.: EXPERIMENTAL STUDY ON PERFORMANCE OF SYNTHETIC APERTURE RADIOMETER HUT-2D 819 Fig. 3. (Left) Reference brightness temperature of the sky during the galactic plane measurement. The angular resolution of the model is 0.25◦ × 0.25◦ , and the color scale is truncated to make the galactic plane more visible. (Middle) Reference brightness temperature map convolved with the HUT-2D synthetic antenna pattern with an angular resolution of approximately 10◦ × 10◦ . (Right) HUT-2D measurement of the galactic plane area using FTT image processing. The local radiation maximum of Cygnus is clearly visible at approximately (ξ = −0.1; η = −0.05). The color bar shown next to the leftmost image is common to all the images. measurement time is shown in Fig. 2 (left). Accordingly, the brightness temperature should be rather flat in the AF-FOV area, slowly increasing from 6.0 K at the zenith to 6.5 K at the edge of the area. The first acquisitions after sunset are considered as the FTRs of the instrument. For the X-probe, this measurement is shown in Fig. 2 (middle). The FTR acquired in this point of time is used in the image processing of all the other measurements presented in this paper. From this FTR of the instrument, we note strong error components: Fluctuations of ±30 K can be recognized once comparison of the measurement with the model in Fig. 2 (left) is made. For the galactic pole measurement, we consider measurements of HUT-2D approximately 1 h after the FTR measurements. The measurement result is shown in Fig. 2 (right). In comparison with the model, as shown in Fig. 2 (left; this model applies to both FTR and galactic pole measurements), we note small error levels and approximately correct overall level of brightness temperature. B. Galactic Plane Measurement The galactic plane was measured using the same configuration as that of the galactic pole measurement. After several hours from the pole measurement, the galactic plane area crossed the instrument’s FOV with the rotation of the Earth. Fig. 3 (left) shows the modeled brightness temperature in the HUT-2D FOV at the time of the measurement based on [17] and [18]. We convolved the reference model with the HUT-2D synthetic antenna pattern to obtain a reference model of the same angular resolution as that measured by HUT-2D. The convolved reference model is shown in Fig. 3 (middle). The model reveals the local radiation maximum at the galactic plane, which is almost at the center of the HUT-2D FOV. This radiation maximum is a well-known extended radiation source located in the direction of the constellation of Cygnus. A 4-min acquisition was considered for the data set. The brightness temperature measured by the HUT-2D X-probe is shown in Fig. 3 (right). The result clearly reveals the radiation maximum of the Cygnus area at the expected location. Fig. 4. (Top left) Brightness temperatures based on forward modeling of the open water test site. The presented image is for X-probe measurements of HUT-2D. Due to the rotational symmetry of the X-and Y-probes, the response for the Y-probe is similar, but it is rotated by 90◦ . (Top right) HUT-2D measurements of the sea during the SSS signature test. (Bottom) One-minute average of the measured brightness temperature of the open water test site using the (left figure) X-probe and (right figure) Y-probe antennas of HUT-2D. A detail worth mentioning is that the weather during the experiment was cloudy, but as its impact on atmospheric emissivity and attenuation is rather low at L-band, the galactic plane imaging was successful regardless. The galactic plane data set is used to study the image bias of HUT-2D and to demonstrate the imaging capabilities of the instrument. C. Open Water Measurements In this paper, we consider four different open water measurement setups for different purposes. For all the cases, sea surface emissivity was modeled for each measurement, as proposed in [20], using in situ SST and SSS information. Also, we require that the wind conditions at the area be light so that the impact of water surface roughness can be discarded. One exemplary outcome of these models is shown in Fig. 4 (top left). It corresponds to the emissivity of water at SST = 820 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 49, NO. 2, FEBRUARY 2011 Fig. 5. Radiometric resolution maps of HUT-2D calculated from the (left) galactic pole, (middle) open water, and (right) forest data sets. Only X-probe images are shown, since Y-probe measurements do not have a meaningful difference. The color bar shown next to the leftmost image is common to all the images. TABLE II N ORMALIZED 1- S R ADIOMETRIC R ESOLUTION OF HUT-2D 20 ◦ C and SSS = 4 psu (practical salinity units). The figure shows the brightness temperature at the instrument’s X-probe antenna. Due to the rotational symmetry of the X-and Y-probes, the response for the Y-probe is similar, but it is rotated by 90◦ [28]. The first setup is selected in order to study the experimental radiometric resolution and pixel-to-pixel random error of HUT-2D. For this purpose, we consider two 30-s acquisitions made over the pure water area, one using each of the HUT-2D antenna probes (X and Y). Fig. 4 (bottom row) shows the results of these measurements. The second setup is selected in order to examine the image bias and its behavior, particularly its stability. For this purpose, we consider all the applicable water scene measurements from HUT-2D data acquisition history. The minimum requirement is that reference SST and SSS information must also be available from the measured area. During open water acquisitions, the aircraft’s attitude was kept as stable as possible, and the flights were always conducted after sunset to avoid any influence from the Sun. An infrared thermometer onboard was used to monitor the physical temperature of the test areas during acquisitions. Furthermore, it is required that acquisitions be made on separate flights so that recalibration and restabilization of the instrument could be performed between measurements. We established altogether nine such cases between 2007 and present. The third open water measurement setup is established in order to study the stability of the instrument during a single acquisition. The measurement data consist of a 30-min continuous acquisition of an open water area. In this case, the instrument was calibrated before the test area, and no recalibration was applied. The fourth open water measurement setup is established to study the accuracy of the instrument’s calibration. The data set consists of a water area, which was measured during ten overflights spanned over a 130-min time frame. The data set is a part of the data gathered for SSS gradient detection over a coastal area of Finland. The measurement campaign is described in detail in [3], which should be referred to for a detailed description of in situ measurements at sea. D. Sea Signature Measurement In order to demonstrate the instrument’s performance and data acquisition capabilities, we present an open water measurement with extended incidence angle coverage. Throughout the data acquisition, which consists of a 1-min acquisition alternating between the X- and Y-probes, the aircraft was performing the so-called “wing wag” maneuvers. During this period, the aircraft’s roll angle was rapidly alternated between ±25◦ . By this means, the incidence angle range measured by the instrument was spanned to cover the range from 0◦ to 50◦ for directions across the flight track. By collecting acquisitions in the across-the-track direction, we retrieved H- and V-polarized brightness temperatures at the local pixel reference frame. The measured brightness temperatures are shown in Fig. 4 (top right), plotted against forward model estimates established using [20]. The measurements clearly follow the expected signature at each polarization. E. Forest Measurement The forest measurement considered in this paper consists of a single 30-s airborne acquisition made over a homogenous boreal coniferous area. During the acquisition, the instrument alternated between the X- and Y-probes. The HUT-2D antenna temperatures during the acquisition were on the order of 250 K, according to calibration. This data set is used only in the calculation of experimental radiometric resolution. KAINULAINEN et al.: EXPERIMENTAL STUDY ON PERFORMANCE OF SYNTHETIC APERTURE RADIOMETER HUT-2D 821 TABLE III I MAGE B IAS C ALCULATED F ROM S EVERAL WATER S CENE M EASUREMENTS V. E XPERIMENTAL P ERFORMANCE A NALYSIS In this section, we apply the experimental performance parameters discussed in Section III to the HUT-2D data sets presented in Section IV. The data used to assess each parameter are summarized in Table I. A. Measurements of Radiometric Resolution The radiometric resolution maps of the galactic pole, open water, and forest measurements are calculated according to (2) and shown in Fig. 5. The figures show only X-probe maps, since Y-probe maps have no significant visible differences. The average radiometric resolution is calculated from the maps according to (3). The corresponding theoretical values are calculated according to (1) and compared with the experimental ones in Table II for the three data sets. Note that Fig. 5 and Table II present resolution values normalized to an integration time of 1 s. In general, the measured radiometric resolutions are well in accordance with the theoretical values. Note that the values for the galactic pole and open water measurements are of the same order despite differing antenna temperatures. This is due to the fact that, in airborne operation, the instrument hardware is at a lower ambient temperature due to the instrument’s location outside the aircraft. This yields to significantly lower receiver noise temperatures compared with the static groundbased measurements. In the two cases, the system temperatures of receivers are approximately the same, which yield to similar radiometric resolutions. Y-probe radiometric resolutions are generally slightly better than the ones of the X-probe. This is due to the lower receiver noise temperatures of the Y-probe receivers, which, in turn, is due to the higher losses in X-probe microstrip feeds. In the case of the forest measurement, the measured resolution is slightly worse than the expected one. This is most likely due to the fact that the forested area is not completely homogenous but includes denser and sparser parts. The presented resolution maps show that noise is distributed very evenly in the image, as was assumed in Section III-A based on the flatness of the targets. B. Measurements of Image Bias Image bias is calculated according to (4) using all the applicable measurements of open water areas available in HUT-2D data acquisition history. The bias values calculated from the nine applicable water areas are shown in Table III. The table shows also the reference SST that was used in forward emission modeling, as well as the operation temperature of the HUT-2D hardware. This temperature is equal to the average physical temperature measured from the RF front end of the receivers. In the last row of the table, we present the average bias and standard deviation values of the samples. As a conclusion of the values, we state that biases in the measurements of the HUT-2D X- and Y-probes are on the order of 3 and −5 K, respectively. From one measurement flight to another, this holds with 1 − σ uncertainties of 2 and 3 K approximately. It must be noted that these uncertainties include also the uncertainty of the forward models used in the calculation. Throughout the open water measurements, HUT-2D has operated at very different temperatures: The highest temperature (31 ◦ C) occurred during a sea salinity measurement campaign in southern Finland in August 2007 [3]. The measurement flights were flown at the altitude of approximately 300 m above sea level, so the air temperature was reasonably high, also due to the late summer season. The coldest measurement environment (12 ◦ C) was during early springtime, when Lake Starnberg was measured in Germany as a part of SMOS calibration and validation activities. In these data sets, HUT-2D biases remain in the same order regardless of the operation temperature. Close observation shows no correlation between bias and temperature. This suggests that HUT-2D radiometric performance is not dependent on the physical temperature of the receiver RF front ends, provided that calibration of the instrument is carried out at the same temperature. We have estimated image bias using measurements with similar antenna temperatures (100–120 K). We can also calculate the bias from the galactic pole and plane measurements shown in Figs. 2 and 3, respectively, and conclude that the 822 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 49, NO. 2, FEBRUARY 2011 Fig. 6. Pixel-to-pixel random error, or residual, maps of HUT-2D once the open water data set is considered. Columnwise, different processing is applied. (Left) FTT processing is not applied. (Middle) FTT processing is applied. This is the nominal image processing technique for the instrument. (Right) Residual map, when the measurements are corrected with a residual map stored from another water measurement acquired four months earlier. The top and bottom rows show the residuals for the X- and Y-probes, respectively. bias is very small, which is < 1 K in both cases and for both antenna probes. This bias is meaningless from the scientific point of view, since it is related to low antenna temperatures, which never occurs in practical cases of observations of scientifically important targets (such as water and land targets). The fact that the bias increases with the antenna temperature leads to an assumption that the bias related to observations of high antenna temperatures is higher than the bias with water measurements. This, however, cannot be consolidated, as none of the “hot” sources measured with the instrument so far is known well enough. C. Measurements of Pixel-to-Pixel Random Error Pixel-to-pixel random error is assessed using the galactic pole and open water measurements. The influence of pixel-topixel random error, i.e., the direction-dependent error component, on the galactic pole measurement (Fig. 2, right) can be seen by comparing it with model predictions (Fig. 2, left). The direction-dependent error of the image is random ripple, the rms error of which with respect to the model is calculated using (6). This yields to 0.4 K for the X-probe measurement and 0.2 K for the Y-probe one. This error level is extremely small, and as was demonstrated in Section IV-B with the galactic plane measurement with successful imaging of the Cygnus area, the error level is stable. This extreme performance in the case of galactic measurements is due to the FTT algorithm, in which it can be said that the more it becomes efficient, the more the target resembles the target of the FTR measurement. Note that if the FTT was not applied, the result would resemble very closely the FTR measurement (Fig. 2, middle). The pixel-topixel random error calculated from the FTR is 10.1 K for the X-probe measurement and 15.7 K for the Y-probe one. TABLE IV P IXEL - TO -P IXEL R ANDOM E RROR ( IN K ELVINS ) In the case of water measurements, the brightness temperature of the target is no longer almost equal to the FTR target. To evaluate the efficiency of FTT processing, we show in Fig. 6 (in the left and middle column images) pixel-to-pixel random error maps for two ways of processing. First, in the left column, we show the error in the measurement without using FTT processing. Second, in the middle column, we show the error in the case when FTT processing is used. From both cases, we calculate the average pixel-to-pixel error component and present them in Table IV. The error levels using nominal processing (FTT) are 6.0 K and 6.2 K for the X- and Y-probe measurements, respectively. Without using FTT, the levels are 7.4 and 8.4 K. Therefore, a modest improvement of 20%–30% is introduced by the FTT. This is smaller than expected, since the influence should be proportional to the change in contrast between the measured brightness temperature and receiver physical temperature [12]. Thus, an improvement of ∼60% was expected. The modest efficiency of the FTT, in this case, can be due to the change in antenna patterns from the FTR measurement, for which the instrument is mounted on tables, looking upward. In this setup, e.g., the minor flexibility of the arms may cause different alignments than that during the more robust aircraft KAINULAINEN et al.: EXPERIMENTAL STUDY ON PERFORMANCE OF SYNTHETIC APERTURE RADIOMETER HUT-2D 823 Fig. 7. (Left) Allan deviation of the average brightness temperature in the HUT-2D AF-FOV once the open water target was measured. (Right) Average brightness temperatures in the HUT-2D AF-FOV for 11 individual overpasses of the same test area. In this case, recalibration of the instrument was conducted before every measurement. The standard deviation of the average values was 0.5 K for both X- and Y-probe measurements. installation. Also, when installed below the aircraft, antenna backlobe patterns are definitely different. There are also minor physical objects in the front hemisphere of the instrument, such as the landing gears. These all decrease the applicability of the FTR measured in a different physical setup. However, we still assume that the direction-dependent error component is dominated by the antenna pattern uncertainty. Again, the error is static by nature from one aircraft installation to another. To benefit from this knowledge, we propose a method to compensate the error component. The procedure is similar to the FTT processing presented in [12], but uses a water measurement as the FTR of the instrument. Thus, the processing requires forward modeling of the water scene. This type of correction is preliminarily tested also with SMOS measurements, and those results were presented in [29]. One outcome of this correction is shown in Fig. 6 (right column). There, we show pixel-to-pixel error components that remain in an open water measurement after we have subtracted the error component stored from a measurement four months before. Now, the error components are 2.3 and 3.2 K for the Xand Y-probe measurements, respectively. Definitely, the error level is decreased from nominal processing (FTT) significantly with the correction. several measurements of the same water target (open water setup 4 in Section IV-C). This setup provides 11 measurements, each consisting of approximately 100 samples measured with both probes of the instrument. Overpasses were separated by approximately 10–12 min in time, and recalibration of the instrument was conducted for each overpass separately. We calculate the average brightness temperature within the instrument’s AF-FOV area over the whole 100-sample acquisition to minimize the impact of thermal noise in the measurement. The average brightness temperatures of the 11 measurements are shown in Fig. 7 (right). The standard deviation between the measured brightness temperatures averages at 0.7 K for both probes. This value can be considered to be the uncertainty of the calibration in terms of image bias within one measurement flight, when the measurement conditions are stable. From the calculated Allan deviation (Fig. 7, left), we see that 0.7-K uncertainty is achieved after a measurement of approximately 300 s. Thus, this is the time after which the recalibration of the instrument should optimally be performed. VI. D ISCUSSION Prompt qualitative conclusions from the previous sections are that HUT-2D can be considered to have the following: D. Instrument Stability In order to assess instrument stability and to determine the frequency needed for calibration, we calculate the Allan deviation defined by (7) from the 30-min-long open water measurement (open water setup 3 in Section IV-C). The Allan deviation is shown in Fig. 7. The result shows that the optimum integration time for instrument operation, i.e., the integration time with which minimum sensitivity is achieved, is on the order of ∼100 s. After this time, averaging no longer improves the sensitivity as instrumental drifts start to dominate. From the result, we see that averaging follows the square law very firmly up to ∼30 s, after which the first signs of instrumental drifts can be seen. Therefore, for the calculation of instrument radiometric resolution, we use 30-s acquisitions, as already mentioned earlier in this paper. To estimate the accuracy of the instrument’s calibration within one measurement flight, we calculate the standard deviation of the average brightness temperature levels measured from 1) low radiometric resolution; 2) moderate pixel-to-pixel random error, i.e., directiondependent error component; 3) slight image bias; 4) reasonable stability. The first two items establish a general requirement for averaging: in the time domain to overcome the low radiometric resolution and in the incidence angle domain to decrease the pixelto-pixel random error component. The good stability and the well-characterized bias establish a firm base for averaging: For measurements carried out in different conditions, even when separated in time, instrument properties are similar enough for averaging to enhance sensitivity and radiometric resolution. The low radiometric resolution is expected, and accepted, in the context of aperture synthesis radiometers—the low resolution of the technique is to be compensated with the benefits emerging from 2-D multiangular imaging. 824 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 49, NO. 2, FEBRUARY 2011 Pixel-to-pixel random error, i.e., direction-dependent error component, restricts one of the most fundamental benefits of aperture synthesis radiometers: 2-D imaging. Following from the visibility equation, antenna pattern characterization errors are among the most dominating sources of error. This property is pronounced in the presented measurements of HUT-2D: The FTR measurement is characterized by strong error components. Antenna pattern errors tend to be static by nature. This is seen from the fact that the galactic plane was imaged successfully (Fig. 3) despite the strong fluctuations of the FTR using the elegant FTT algorithm. However, low effectiveness of the FTT in the processing of open water measurements implies that the antenna patterns are different in the aircraft installation and in the FTR measurement installation. To study and, finally, to mitigate this, we proposed a correction method, which showed promising results. In the presented test case, the pixel-to-pixel error component decreased to a level better than the radiometric resolution of the instrument. The final efficiency of the correction method depends on the stability of the pixel-to-pixel random error component from one flight to another, as well as the scalability of the component with the changing brightness temperature level. These topics will be addressed in near-future studies concerning the performance of HUT-2D. Similar topics are also timely in the analysis of the performance of SMOS measurements, for which a similar correction is recently proposed [29]. In the case of MIRAS, deformation of the antenna patterns should also be considered. The antenna patterns of MIRAS are characterized to a high accuracy. However, the impact of the launch and the effects of the space environment can be surprising. The efficiency of FTT processing in the case of MIRAS will be an interesting research topic in the near future. Based on the results presented in this paper, we propose the use of HUT-2D measurements, specifically for applications where a high flight altitude, and thus a mediocre ground resolution, is accepted. This enables long measurement times of the target from different viewing angles. To name one field of interest, HUT-2D data will be used in the future to study the emission modeling of forested areas, where, for example, the dependence of emissivity as a function of incidence angle is not well characterized. For example, setups similar to that in the case of the presented SSS signature measurement (resulting in the incidence angle coverage visible in Fig. 4, top right) could be applied. VII. C ONCLUSION In this paper, we have analyzed measurements of a complete synthetic aperture radiometer in order to understand, demonstrate, and validate the performance of the novel technology in general and the performance and usability of the HUT-2D instrument in particular. We have defined four figures of merit, namely, radiometric resolution, image bias, random error, and stability, which we assessed using the data of several groundbased and airborne measurement campaigns carried out with HUT-2D during the past two years. The radiometric resolution calculated from the HUT-2D instrument measurements was closely in line with theoretical expectations. During the three test setups, namely, galactic pole, open water, and forest, the measured radiometric resolutions (presented in Table II) were always close to theoretical expectations regardless of the target. The slightly higher receiver noise temperatures with the X-probe measurements yielded lower radiometric resolutions on that probe, as predicted by theoretical calculations. The image bias was assessed using galactic and open water measurements. The bias of the former was small (under 1 K). The bias of the open water measurement was found to be larger, as presented in Table III. However, this remains reasonably constant from one flight to another. The level of bias within the measurements of the same flight is low, which is on the order of < 0.7 K (1σ) for the water measurements. The circumstances, which change from one installation to another, affect the bias more significantly, resulting in biases with a standard deviation on the order of 2–3 K for the water measurements. The pixel-to-pixel random error component of the instrument was calculated from a measurement of the galactic pole and open water areas. The measurements conclude that the direction-dependent radiometric resolution of HUT-2D was less than 1 K for the galactic pole and plane measurements and 6 K for the open water measurements. As discussed, this directiondependent error is dominated by a static error from antenna pattern characterization. That component can be corrected by using stored error component maps. With this processing, the pixel-to-pixel error component was decreased by a factor of two in the presented test case. The scalability of the image bias and random error emerges from the image processing approach, i.e., the FTT, which suppresses multiplicative uncertainties more effectively the more the target resembles the FTR of the instrument. This was well seen in the error components calculated from the galactic and water measurements. This phenomenon will be seen also in the forthcoming measurements of the MIRAS instrument. Future actions in ongoing performance studies aim at the consolidated usage of the stored pixel-to-pixel random error correction. Also, we aim at better characterization of antenna losses of individual antennas, which are among the main sources of error in the HUT-2D calibration process. The better characterization will yield to an enhanced pixel-to-pixel random error, since it mainly affects the amplitude of the measured visibility sample. Also, the performance of the instrument as a function of rapidly changing thermal conditions is an open issue. In this paper, utilizing data sets of a complete synthetic aperture radiometer HUT-2D, we have established a solid understanding of the radiometric performance of the instrument. HUT-2D measurements are used by a large scientific community, e.g., in the frame of SMOS calibration and validation activities, and the results of this paper will enable more thorough analysis and interpretation of measurement data. By understanding the behavior, nonidealities, and limitations of HUT-2D, we also develop our readiness to confront issues related to the performance of MIRAS and other future instruments utilizing interferometric imaging principles. KAINULAINEN et al.: EXPERIMENTAL STUDY ON PERFORMANCE OF SYNTHETIC APERTURE RADIOMETER HUT-2D R EFERENCES [1] K. Rautiainen, J. Kainulainen, T. Auer, J. Pihlflyckt, J. Kettunen, and M. Hallikainen, “Helsinki University of Technology L-band airborne synthetic aperture radiometer,” IEEE Trans. Geosci. Remote Sens., vol. 46, no. 3, pp. 717–726, Mar. 2008. [2] J. Kainulainen, K. Rautiainen, S. Tauriainen, T. Auer, J. Kettunen, and M. Hallikainen, “First 2-D interferometric radiometer imaging of the Earth from an aircraft,” IEEE Geosci. Remote Sens. Lett., vol. 4, no. 2, pp. 241–245, Apr. 2007. [3] J. Kainulainen, K. Rautiainen, J. Lemmetyinen, M. Hallikainen, F. Martin-Porqueras, and M. Martin-Neira, “Sea surface salinity retrieval demonstration using datasets of airborne synthetic aperture radiometer HUT-2D,” in Proc. IEEE IGARSS, Cape Town, South Africa, 2009, pp. III-737–III-740. [4] K. D. McMullan, M. A. Brown, M. Martin-Neira, W. Rits, S. Ekholm, J. Marti, and J. Lemanczyk, “SMOS: The payload,” IEEE Trans. Geosci. Remote Sens., vol. 46, no. 3, pp. 594–605, Mar. 2008. [5] M. Martin-Neira, I. Cabeza, C. Perez, M. A. Palacios, M. A. Guijarro, S. Ribo, I. Corbella, S. Blanch, F. Torres, N. Duffo, V. Gonzalez, S. Bereza, A. Camps, M. Vall-Ilossera, S. Tauriainen, J. Pihlflyckt, J. P. Gonzalez, and F. Martin-Porqueras, “AMIRAS—An airborne MIRAS demonstrator,” IEEE Trans. Geosci. Remote Sens., vol. 46, no. 3, pp. 705– 716, Mar. 2008. [6] D. M. Le Vine, T. J. Jackson, and M. Haken, “Initial images of the synthetic aperture radiometer 2D-STAR,” IEEE Trans. Geosci. Remote Sens., vol. 45, no. 11, pp. 3623–3632, Nov. 2007. [7] J. Randa, J. Lahtinen, A. Camps, A. J. Gasiewski, M. Hallikainen, D. M. Le Vine, M. Martin-Neira, J. Piepmeier, P. W. Rosenkranz, C. S. Ruf, J. Shiue, and N. Skou, “Recommended terminology for microwave radiometry,” Nat. Inst. Standards Technol., Gaithersburg, MD, NIST Technical Note 1551, Aug. 2008. [8] I. Corbella, N. Duffo, M. Vall-Ilossera, A. Camps, and F. Torres, “The visibility function in interferometric aperture synthesis radiometry,” IEEE Trans. Geosci. Remote Sens., vol. 42, no. 8, pp. 1677–1682, Aug. 2004. [9] B. Picard and E. Anterrieu, “Comparison of regularized inversion methods in synthetic aperture imaging radiometry,” IEEE Trans. Geosci. Remote Sens., vol. 43, no. 2, pp. 218–224, Feb. 2005. [10] I. Corbella, F. Torres, A. Camps, N. Duffo, and M. Vall-Ilossera, “Brightness-temperature retrieval methods in synthetic aperture radiometers,” IEEE Trans. Geosci. Remote Sens., vol. 47, no. 1, pp. 285–294, Jan. 2009. [11] I. Corbella, F. Torres, A. Camps, A. Colliander, M. Martin-Neira, S. Ribo, K. Rautiainen, N. Duffo, and M. Vall-Ilossera, “MIRAS end-toend calibration: Application to SMOS L1 processor,” IEEE Trans. Geosci. Remote Sens., vol. 43, no. 5, pp. 1126–1134, May 2005. [12] M. Martin-Neira, M. Suess, J. Kainulainen, and F. Martıacute;nPorqueras, “The flat target transformation,” IEEE Trans. Geosci. Remote Sens., vol. 46, no. 3, pp. 613–620, Mar. 2008. [13] A. Camps, M. Vall-Ilossera, I. Corbella, N. Duffo, and F. Torres, “Improved image reconstruction algorithms for aperture synthesis radiometers,” IEEE Trans. Geosci. Remote Sens., vol. 46, no. 1, pp. 146–158, Jan. 2008. [14] A. Camps, I. Corbella, J. Bara, and F. Torres, “Radiometric sensitivity computation in aperture synthesis interferometric radiometry,” IEEE Trans. Geosci. Remote Sens., vol. 36, no. 2, pp. 680–685, Mar. 1998. [15] J. B. Hagen and D. T. Farley, “Digital correlation techniques in radio science,” Radio Sci., vol. 8, no. 8, pp. 775–784, Aug./Sep. 1973. [16] R. Butora and A. Camps, “Noise maps in aperture synthesis radiometric images due to cross-correlation of visibility noise,” Radio Sci., vol. 38, no. 4, p. 1067, Jul. 2003. [17] D. M. Le Vine and S. Abraham, “Galactic noise and passive microwave remote sensing from space at L-band,” IEEE Trans. Geosci. Remote Sens., vol. 42, no. 1, pp. 119–129, Jan. 2004. [18] F. T. Ulaby, R. K. Moore, and A. K. Fung, Microwave Remote Sensing, Active and Passive. Norwood, MA: Artech House, 1981, pp. 256–343. [19] J. D. Krauss, Radio Astronomy, 2nd ed. Powell, OH: Cygnus-Quasar Books, 1986. [20] L. A. Klein and C. T. Swift, “An improved model for the dielectric constant of sea water at microwave frequencies,” IEEE Trans. Antennas Propag., vol. AP-25, no. 1, pp. 104–111, Jan. 1977. [21] S. Zine, J. Boutin, J. Font, N. Reul, P. Waldteufel, C. Cabarró, J. Tenerelli, F. Petitcolin, J.-L. Vergely, M. Talone, and S. Delwart, “Overview of the SMOS sea surface salinity prototype processor,” IEEE Trans. Geosci. Remote Sens., vol. 46, no. 3, pp. 621–645, Mar. 2008. 825 [22] B. I. Vichev, E. N. Krasteva, and K. G. Kostov, “Study of seasonal evolution of tree emission using zenith-looking microwave radiometers,” in Proc. IEEE IGARSS, Florence, Italy, Jul. 10–14, 1995, pp. 981–983. [23] F. Torres, A. Camps, J. Bará, and I. Corbella, “Impact of receiver errors on the radiometric resolution of large two-dimensional aperture synthesis radiometers,” Radio Sci., vol. 32, no. 2, pp. 629–641, Mar./Apr. 1997. [24] A. Camps, J. Bará, F. Torres, I. Corbella, and J. Romeu, “Impact of antenna errors on the radiometric accuracy of large aperture synthesis radiometers,” Radio Sci., vol. 32, no. 2, pp. 657–668, Mar./Apr. 1997. [25] I. Corbella, A. Camps, F. Torres, and J. Bará, “Analysis of noise-injection networks for interferometric radiometer calibration,” IEEE Trans. Microw. Theory Tech., vol. 48, no. 4, pp. 545–552, Apr. 2000. [26] E. Anterrieu, “On the reduction of the reconstruction bias in synthetic aperture imaging radiometry,” IEEE Trans. Geosci. Remote Sens., vol. 45, no. 4, pp. 1084–1093, Apr. 2007. [27] F. L. Walls and D. W. Allan, “Measurements of frequency stability,” Proc. IEEE, vol. 74, no. 1, pp. 162–168, Jan. 1986. [28] P. Waldteufel and G. Caudal, “About off-axis radiometric polarimetric measurements,” IEEE Trans. Geosci. Remote Sens., vol. 40, no. 6, pp. 1435–1439, Jun. 2002. [29] J. Font, J. Boutin, N. Reul, P. Spurgeon, A. Chuprin, C. Cabarro, J. Gourrion, S. Lavender, N. Martin, M. McCullogh, F. Petitcolin, M. Portabella, R. Sabia, J. Tenerelli, M. Talone, J.-L. Vergely, X. Yin, and S. Zine, “Retrieving sea surface salinity from SMOS measurements: First assessment,” presented at the 11th Specialist Meeting Microwave Radiometry Remote Sensing Environment, Washington, DC, Mar. 1–4, 2010. Juha Kainulainen was born in Lappajärvi, Finland, in 1979. He received the M.Sc. degree in technology from the Helsinki University of Technology (TKK), Helsinki, Finland, in 2004. Since 2005, he has been a Doctoral Student and a Project Manager with the Department of Radio Science and Engineering, Aalto University School of Science and Technology, Espoo, Finland. His duties include development and testing of the department’s synthetic aperture radiometer system HUT-2D and managing and working in several European Space Agency (ESA) projects in the frame of the Soil Moisture and Ocean Salinity (SMOS) mission. From 2004 to 2005, he was a Young Graduate Trainee with ESA. His work was related to signal processing of the payload instrument of the SMOS mission. From 2001 to 2004, he was a Research Assistant with the Laboratory of Space Technology, Helsinki University of Technology (currently part of Aalto University). From 1998 to 1999, he was a Trainee with Nokia Networks and Nokia Telecommunications. His research interests include radiometry, interferometry, and signal processing. Kimmo Rautiainen received the M.Sc. degree from the Helsinki University of Technology (TKK), Espoo, Finland, in 1996. He was a Research Scientist with the TKK Laboratory of Space Technology focusing on microwave radiometer systems, with emphasis on interferometric radiometers. His main work in TKK was within the TKK airborne 2-D interferometric radiometer HUT-2D and on Soil Moisture and Ocean Salinityrelated projects. He is currently a Research Scientist with the Arctic Research, Finnish Meteorological Institute, Helsinki, Finland, continuing his research on radiometers, and a Project Manager in mobile high-frequency radar development project. Juha Lemmetyinen received the M.Sc.(tech.) degree from the Helsinki University of Technology (TKK), Espoo, Finland, in 2004. From 2004 to 2008, he was a Researcher with the TKK Laboratory of Space Technology and the TKK Department of Radio Science and Engineering, specializing in radiometer calibration techniques and remote sensing. He is currently a Scientist with the Arctic Research, Finnish Meteorological Institute, Helsinki, Finland. His current research interests include applications of microwave radiometers, radiative transfer modeling, and remote sensing of snow. 826 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 49, NO. 2, FEBRUARY 2011 Jaakko Seppänen received the M.Sc.(tech.) degree from the Helsinki University of Technology (TKK), Espoo, Finland, in 2008. He has been working toward the Ph.D. degree in the Department of Radio Science and Engineering, School of Science and Technology, Aalto University, Aalto, Finland, since 2008, specializing in microwave remote sensing. His current research interests include applications of microwave radiometers, radiative transfer modeling, and remote sensing of vegetation. Pauli Sievinen (S’09) is currently working toward the M.S. degree in electrical engineering at the Aalto University School of Science and Technology, Espoo, Finland. Since 2008, he has been Research Associate with the Department of Radio Science and Engineering, Aalto University School of Science and Technology. Matias Takala received the M.Sc. degree in technology from the Faculty of Electrical Engineering, Helsinki University of Technology (TKK), Espoo, Finland, in 2001. From 2001 to 2006, he was a Research Scientist with TKK. Since 2007, he has been a Research Scientist with the Arctic Research, Finnish Meteorological Institute, Helsinki, Finland. His research interest includes microwave remote sensing of snow cover and developing remote sensing applications. Martti T. Hallikainen (M’83–SM’83–F’93) received the Doctor of Technology degree from the Faculty of Electrical Engineering, Helsinki University of Technology (TKK), Espoo, Finland, in 1980. Since 1987, he has been a Professor of space technology with TKK (Aalto University starting 2010). He was a Visiting Scientist with the Jet Propulsion Laboratory and the NASA Goddard Space Flight Center/University of Maryland Goddard Earth Sciences and Technology Center, Baltimore, in 2007– 2008 and with the European Union’s Joint Research Centre, Institute for Remote Sensing Applications, Italy, in 1993–1994, and he was a Postdoctoral Fellow at the Remote Sensing Laboratory, University of Kansas, Lawrence, in 1981–1983. His research interests include development of microwave sensors for air- and spaceborne remote sensing, development of methods to retrieve the characteristics of geophysical targets from satellite and airborne measurements, and cryospheric applications of remote sensing. Dr. Hallikainen was was a member of the Geoscience and Remote Sensing Society (GRSS) Administrative Committee in 1988–2006 and the General Chair of the IGARSS’91 Symposium held in Espoo. He was the President of IEEE GRSS in 1996–1997, and he is currently the Cochair of the IEEE GRSS Awards Committee. He was the Chair of International Union of Radio Science (URSI) Commission F in 2002–2005 and has been the Vice President of URSI since 2005. He was the Chair of the URSI Finnish National Committee in 1997–2005 and has been the national official member of URSI Commission F since 1988. He was also a member of the European Space Agency’s Earth Science Advisory Committee in 1998–2001. He has been the Vice Chair of the Finnish National Committee for Space Research since 2000. He was the Secretary General of the European Association of Remote Sensing Laboratories (EARSeL) in 1989–1993 and the Chairman of the Organizing Committee for the EARSeL 1989 General Assembly and Symposium held in Espoo. He has been a member of the EARSeL Council since 1985. He has also been an honorary life member of IEEE GRSS since 2007. He was a recipient of three IEEE GRSS Awards: the 1999 Distinguished Achievement Award, the IGARSS’96 Interactive Paper Award, and the 1994 Outstanding Service Award. He was awarded the IEEE Third Millennium Medal in 2000 and the Microwave Prize for the best paper in the 1992 European Microwave Conference.