SMOS instrument performance and calibration after six years in orbit

1 1 SMOS INSTRUMENT PERFORMANCE AND 2 CALIBRATION AFTER 5 YEARS IN ORBIT 3 4 5 M. Martín-Neira (1), R. Oliva (2), I. Corbella (3), F. Torres (3), N. Duffo (3), I. Durán (3), J. Kainulainen (4), J. 6 Closa (5), A. Zurita (5), F. Cabot (6), A. Khazaal (6), E. Anterrieu (7), J. Barbosa (8), G. Lopes (8), J. Tenerelli 7 (9), R. Díez-García (10), J. Fauste (2), F. Martín-Porqueras (11), V. González-Gambau (12), A. Turiel (12), S. 8 Delwart (13), R. Crapolicchio (13), M. Suess (1) 9 10 11 Corresponding author 12 13 M. Martín-Neira and M. Suess are with ESA-ESTEC, Keplerlaan 1, 2200 AG Noordwijk, The 14 Netherlands. (e-mail: manuel.Martí[email protected]) 15 16 Authors’ affiliations 17 18 (1) European Space Agency at ESTEC, Noordwijk, The Netherlands. 19 (2) European Space Agency at ESAC, Villanueva de la Cañada, Spain. 20 (3) Polytechnic University of Catalonia (UPC), Barcelona, Spain. 21 (4) Harp Technologies Ltd., Espoo, Finland. 22 (5) EADS-CASA Espacio, Madrid, Spain. 23 (6) Centre d'Etudes Spatiales de la BIOsphère (CESBIO), Toulouse, France. 24 (7) Institut de Recherche en Astrophysique et Planétologie (IRAP), Toulouse, France. 25 (8) DEIMOS, Lisbon, Portugal. 26 (9) OceanDataLab, Brest, France. 27 (10) IDEAS, ESAC, Villanueva de la Cañada, Spain. 28 (11) VEGA Telespazio, ESAC, Villanueva de la Cañada, Spain. 29 (12) SMOS Barcelona Expert Centre, Barcelona, Spain. 30 (13) European Space Agency at ESRIN, Frascati, Italy. 31 32 Keywords 33 34 Soil Moisture and Ocean Salinity (SMOS) Mission, soil moisture, sea surface salinity, L-Band 35 radiometry, aperture synthesis, MIRAS 36 37 38 39 ©2016. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ DOI http://dx.doi.org/10.1016/j.rse.2016.02.036 REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 2 1 Abstract 40 41 ESA’s Soil Moisture and Ocean Salinity (SMOS) mission, launched 2-Nov-2009, has been in orbit for over 5 42 years, and its Microwave Imaging Radiometer with Aperture Synthesis (MIRAS) in two dimensions keeps 43 working well. The calibration strategy remains overall as established after the commissioning phase, with a few 44 improvements. The data for this whole period has been reprocessed with a new fully polarimetric version of the 45 Level-1 processor which includes a refined calibration schema for the antenna losses. This reprocessing has 46 allowed the assessment of an improved performance benchmark. An overview of the results and the progress 47 achieved in both calibration and image reconstruction is presented in this contribution. 48 2 INTRODUCTION 49 50 With an experience of over 5 years of in-orbit operation, much has been learnt on how MIRAS works 51 and how its observations can be improved through better calibration and image reconstruction 52 techniques. The purpose of this paper is to update the reader with the latest results on the payload 53 performance and data processing of the SMOS mission (Mecklenburg et al., 2012). SMOS is currently 54 delivering several products, some of them used by operational systems, others only for scientific 55 research (Mecklenburg et al., in press). MIRAS is a Microwave Imaging Radiometer with two-56 dimensional Aperture Synthesis, which remains being the first and so far, the only one of its kind, in 57 space. The main feature of MIRAS is that it obtains two-dimensional images at every snapshot without 58 needing any mechanical scanning of its antenna, a very distinct capability when compared with 59 traditional scanners or push-broom radiometers. A detailed description of the instrumental aspects of 60 MIRAS can be found in (McMullan et al., 2008) while the on-board Calibration System and respective 61 in-flight calibration strategy are described in (Brown et al., 2008) and (Martín-Neira et al., 2008). One 62 year after launch the calibration approach was slightly modified with the initial flight experience, and 63 the first SMOS instrument in-orbit performance was assessed in (Oliva et al., 2013), including the 64 effect of the unexpectedly severe Radio Frequency Interference from ground transmitters (Oliva et al., 65 2012). The present paper will then follow the same structure as (Oliva et al., 2013), with important 66 REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 3 additions brought by the accumulated experience of over 5 years: Section 3 provides an overview of the 67 main sources of error and the current mitigation strategies used to overcome them; Section 4 68 summarizes the current status of calibration activities, including all latest modifications to the initial 69 calibration plan; Section 5 presents the in-orbit behaviour of the most critical instrument parameters; 70 Section 6 gives the performance obtained with the latest version of the Level-1 processor, through the 71 spatial and temporal analysis of brightness temperature images, and finally, Section 7 includes a view 72 on the current investigations that should lead to the next version of the Level-1 processor with a hint on 73 the expected improvements. 74 It is worth mentioning that, at the time of the writing of this paper, the running version of the 75 operational SMOS Level-1 data processor is V620, that a new version, V700, has been delivered and is 76 under assessment, and that the entire data record of the SMOS mission (from January 2010 onwards) 77 has been reprocessed with V620 and is available to the whole SMOS user community. 78 79 3 ERROR SOURCES AND MITIGATION TECHNIQUES 80 3.1 Error Sources 81 Different error sources cause different effects on the SMOS brightness temperature images. Therefore 82 in this section the error sources will be presented according to the effect they produce in the images. 83 3.1.1 Systematic Spatial Ripple 84 85 Figure 1 presents the deviation, with respect to a forward model, of an image of the brightness 86 temperature measured by SMOS over a portion of the South-Eastern Pacific Ocean in X-polarization 87 (X-polarization refers to the image formed with the signal collected by the horizontal probe of MIRAS 88 antenna elements). The comparison is performed after averaging many snapshots so that random errors 89 induced by the radiometric resolution can be neglected, and only systematic errors remain. The most 90 prominent features of such deviation image are a +0.96 K bias and a 1.5 K rms spatial ripple, both 91 statistics evaluated within the dashed circle shown in Figure 1. Similar statistics can be computed for 92 REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 4 the Y-polarization as well as for the Stokes-3 and Stokes-4 parameters, obtaining, in general, different 93 values for the different parameters, values which, in turn, depend on the particular image reconstruction 94 approach being applied, that is, on the Level-1 processor version. Furthermore, and although it is not as 95 easy to show as with measurements of the relatively uniform ocean, bias and ripples also appear in 96 images acquired over any region of the Earth, be it land, ice or coastlines, and over the Cold Sky, 97 exhibiting a magnitude which is scene-dependent. Interpreting the bias as a spatial ripple of an infinite 98 spatial wavelength, both bias and spatial ripple shall be understood as comprised within the ‘spatial 99 ripple’ referred to in what follows. 100 101 102 103 Figure 1: Example of bias and spatial ripples of SMOS images when compared to a radiative transfer 104 ocean model. The axes are the director cosines and the colour scale is in Kelvin. 105 106 The cause and existence of the systematic spatial ripple underlying all SMOS images was already 107 studied before SMOS launch (Camps et al., 2005)(Anterrieu, 2007). Thanks to the investigations 108 conducted during the last 5 years in several directions, using flight data, an important conclusion has 109 been consolidated: the spatial ripple results mostly from the combination of having different antenna 110 element patterns and imaging in alias conditions (that is, using a spatial sampling which leads to having 111 aliased images in some parts of the real space). This is illustrated in Figure 2, which shows a similar 112 deviation image to Figure 1 obtained simulating different conditions: the left and right columns assume 113 identical and different –perfectly known– antenna patterns respectively, while the rows correspond to 114 different antenna element spacings, the one of MIRAS (0.875λ) in the top, and another one which is 115 alias-free (0.55λ) in the bottom. As it is evident, the spatial ripple appears only in the top right plot, that 116 REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 5 is, for alias condition and different antenna patterns. If there is no aliasing, or/and if the antenna 117 patterns are all identical, then there is no significant spatial ripple. 118 As a consequence of this, two further results have now been well established: first, even in the ideal 119 case where the antenna pattern of every antenna element were known perfectly well, a non-negligible 120 systematic spatial error would still be present in SMOS images, dubbed ‘noise floor’; second, the part 121 of the scene outside the alias-free area does have an impact on the scene recovered in the alias-free 122 region, or in other words, the spatial ripple in the alias-free area depends on the scene in the aliased 123 regions (Corbella et al., 2014). 124 The first result might be the most striking one, since for long, it had been expected that the accurate 125 knowledge of the antenna patterns, very carefully characterized on ground and used in the image 126 reconstruction, would have enabled the acquisition of images with negligible ripple. It also emphasizes 127 the need of having the interferometric instrument with either an alias-free element spacing or as similar 128 antenna patterns as possible, to suppress the spatial ripple from the images. The second result, on the 129 other hand, has been the basis to build ripple reduction methods to improve the current SMOS images, 130 as will be seen in the section devoted to mitigation techniques below. Finally the contribution to the 131 spatial ripple due to the limited knowledge of the antenna patterns and residual calibration errors shall 132 not be forgotten. 133 134 REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 6 135 136 Figure 2: Illustration that the spatial ripple results from the combination of different antenna patterns 137 and alias condition. The image shows the Earth and the sky at a spatially uniform but different 138 brightness temperature, viewed with the nominal SMOS geometry. The axes are the director 139 cosines and the colour bar represents the retrieved brightness temperature in Kelvin. 140 141 3.1.2 Sun and RFI Tails 142 143 Figure 3 is a deviation image as Figure 1 but acquired at a time when the Sun is in front of the antenna. 144 The Sun can be seen as a white circle towards the right side of the unity circle, which represents the 145 front hemisphere of the antenna. The area shaded in blue is the locus of possible positions of the Sun as 146 seen from the SMOS orbit around the year. The Sun locus is sufficiently far away from the extended 147 field of view of SMOS that it would not cause any ripples if it were not because of the aliasing and the 148 side lobes. Indeed, a replica of the Sun is clearly visible inside the field of view. Moreover, the real Sun 149 and its replica appear connected by lines of side lobes which cross the entire field of view, including 150 the alias free region. Extending the lines of side lobes through the black dashed lines it is possible to 151 REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 7 located a second alias in the upper part of the unity circle, outside the extended alias-free field of view. 152 The cause of these side lobes is the same as that of the spatial ripple: the differences across antenna 153 patterns enhance the side lobes joining the Sun aliases, which are generated by the element spacing. 154 Patterns of side lobes are also excited by Radio Frequency Interference (RFI) transmitters, which 155 behave within the image reconstruction process, much in the same way as the Sun. This is illustrated in 156 Figure 4, where the extended alias-free field of view is projected on ground at a location of an RFI 157 source: an hexagonal pattern of side lobes is clearly visible. RFI sources remain being an important 158 error source in SMOS (Oliva et al., in press). 159 160 Figure 3: Sun tails and alias affecting an ocean image 161 162 163 164 165 Figure 4: Hexagonal pattern of side lobes excited by a strong Radio Frequency Interference source 166 167 3.1.3 Land-Sea Contamination 168 169 REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 8 The left plot of Figure 5 shows a global view of the oceans with the accumulated deviation of SMOS 170 measurements from a forward model. Brightness temperature warmer than the model are seen around 171 all continental masses which wrongly lead to fresher water retrievals. This feature of SMOS images is 172 referred to as ‘land-sea contamination’ and is of concern among the Sea Surface Salinity community 173 because it can extend several hundreds of kilometers into the open ocean. Much effort has been devoted 174 to understand the reason behind this land-sea contamination. The current understanding is that it is 175 caused by several contributions. The most important one seems to be a calibration error, of the order of 176 2%, in the amplitude of the correlations, the so-called Gkj correlator efficiency coefficients (Corbella et 177 al., in press). Although this problem with the correlation coefficients has been identified, to date, the 178 root cause has not been found yet and the search within the instrumental details related to it continues. 179 The second contributor is the spatial ripple described above, generated by the warmer land and 180 extending into the ocean. 181 182 183 Figure 5: Stokes-1/2 residual images against a radiative transfer ocean model using present 184 correlation efficiency factors (left), where the land-sea contamination is clearly observed around 185 the continental masses, and using corrected values (right), with significantly reduced errors. The 186 warm areas around Alaska, Greenland, Arabian Sea, Gulf of Bengal and Sea of China are due to 187 Radio Frequency Interference, and the blue rim around Antarctica is due to un-modelled sea ice. 188 189 3.1.4 Seasonal Variations 190 191 The right plot of Figure 6 shows the current deviation of the Stokes-1/2 parameter over the Pacific 192 Ocean, averaged within the alias-free field of view, with respect to an ocean model, along the 193 descending passes. The plot spans 5 years, from 2010 till 2014, and from 60 ° South to 60 ° North in 194 REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 9 latitude, with a brightness temperature scale of ±1.2 K. This Hovmöller plot constitutes a powerful tool 195 to analyze any seasonal (and latitudinal) variations. Besides the red stripe during the Commissioning 196 Phase in early 2010, the variations are contained within ±0.4 K except for the eclipse periods (mid-197 November to mid-February) and a region around October where some warm signatures are observed. 198 During an eclipse, the Sun, which can be as high as 31° above the antenna horizon, is suddenly 199 occulted by the Earth. The antenna skin temperature falls down a couple of tens of degrees, from 200 around 28°C to some 5°C. When the satellite exists the Earth shadow, the Sun warms up the antenna 201 again until it reaches the temperature it would have if there had been no eclipse. This post-eclipse 202 transient causes, in every descending orbit, a warm anomaly in the brightness temperature which 203 extends to latitudes as low as 30° North. On the other hand, the reason for the October warm anomaly 204 has not yet been uncovered. Attempts to correlate it with residual galactic noise or other geophysical 205 signatures have failed and hence, an instrumental origin should be assumed. Furthermore this anomaly 206 seems more intense in 2014. 207 208 209 Figure 6: Latitude-Time Hovmöller plot of the descending pass Stokes-1/2deviation from an ocean 210 forward model, averaged in the alias-free area, with V505 (left) and V620 (right) Level 1 processor 211 versions. Colour bar is given in Kelvin. 212 213 3.1.5 Orbital Variations 214 215 REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 16 1977). An iterative process to estimate the position and brightness temperature of the Sun has 349 shown promising results, but the increase in computational time is critical and is still under 350 evaluation. A simpler method has been implemented in the Level-1 data processor of SMOS, 351 with limited improvement, and has been complemented by flagging. Figure 10 shows an 352 example of the Sun correction. 353 354 355 Figure 10: SMOS image of the Sun in cold sky pointing mode before (left) and after (right) Sun 356 correction. Colour scale is in Kelvin. 357 358 359 In principle, the methods for the correction of the Sun alias and its tails can also be applied to remove 360 RFI sources effects. However, the population of RFI sources is irregular, clustering in some regions of 361 the Earth, with several interferors appearing at the same time inside the field of view of SMOS. In this 362 situation a correction for the RFI sources of the type described for the Sun is difficult. Nonetheless, 363 techniques to better detect, flag and correct for RFI sources keep being developed and assessed 364 (Khazaal et al., 2014). In parallel, a technique called the Nodal Sampling (González-Gambau et al., 365 2015) has been proposed to image RFI polluted areas and is under assessment, showing some 366 promising results. 367 3.2.3 Correction of Land-Sea Contamination 368 369 During the investigation of the land-sea contamination error it has become clear that a mismatch 370 between the amplitude of the visibility at the origin V(0,0) and the rest of the samples, V(k,j), generates 371 this kind of degradation, as shown in the left plot of Figure 5. Empirically it has been proven that 372 REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 17 affecting the correlator efficiency coefficients Gkj of the visibility samples V(k,j) outside the origin by a 373 factor near 0.98 removes significantly the land-sea contamination. The right plot of Figure 5 shows the 374 improvement when this correction is applied. The warm brightness temperature halos surrounding the 375 continents have mostly disappeared. The possibility of correcting the correlation efficiency coefficients 376 Gkj has been implemented in the latest version of the SMOS Level-1 processor (V700) and will be 377 subject of validation before it is accepted to enter into operation. 378 It is worth mentioning that a parallel empirical correction of the land-sea contamination is being 379 prepared at Level-2 based on a mask built with the 5 year long record of SMOS data (SMOS Level-2 380 Ocean Salinity Team, 2015). 381 382 3.2.4 Correction of Seasonal and Orbital Variations 383 384 The seasonal and orbital variations are observed in the right plots of Figures 6 and 7 respectively as 385 warm anomalies around October and in the eclipse season every year. The current strategy to correct 386 for these fluctuations is to simplify the calibration approach of the instrument as much as possible by 387 using the All-LICEF mode (Torres et al., 2006) explained below and then attempt new corrections to 388 mitigate them. 389 390 4 IN FLIGHT CALIBRATION PLAN 391 4.1 The Corbella Equation 392 The Corbella equation, introduced in 2003 (Corbella et al., 2004a), involves a fundamental 393 modification to the formulation of interferometry, as used in radio-astronomy, that is necessary to 394 describe the way an aperture synthesis radiometer of the type of MIRAS works. The calibration of 395 SMOS is based on the Corbella equation, and hence, one of the first and most important tasks 396 undertaken in the frame of the calibration of the instrument was its verification (Martín-Porqueras et 397 al., 2010). Such exercise would ideally involve the imaging of two perfectly uniform black body targets 398 REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 18 at two different physical temperatures. Since the Cold Sky near the galactic pole is the only reasonable 399 realization to such uniform target, the validation of the Corbella equation focused on the Cold Sky. 400 With the help of a simulator, two sets of visibility samples of the Cold Sky were produced using the 401 radio astronomy and the Corbella equations. Then, to improve contrast, the visibility samples at the 402 origin were set to zero, and a simple Fourier Transform was applied to provide the images of the 403 modified brightness temperature of the Cold Sky in each case. The image obtained using the radio 404 astronomy formulation is shown in the left panel of Figure 11, while the one resulting from the use of 405 the Corbella equation is in the right panel. These images were then compared to the one measured by 406 SMOS, shown in the center panel of Figure 11, which was generated in the same way, i.e. through a 407 Fourier Transform of the measured visibility samples setting the one at the origin to zero. The image 408 using the Corbella equation is very similar to that measured by SMOS, while the Cold Sky retrieved 409 with the radio astronomy equation does not capture any of the features present in the measurements. 410 This test validated the Corbella equation. 411 412 413 Figure 11: Expected modified brightness temperature of the Cold Sky using the radio astronomy (left) 414 and the Corbella (right) equation; center is the SMOS measurement (note: the visibility sample at the 415 origin has been set to zero to improve contrast) 416 4.2 Routine In-orbit Calibration Plan 417 The routine in-orbit calibration plan was established at the end of the Commissioning Phase, in May 418 2010. With the experience of the first year of the operational phase (Oliva et al., 2013), weekly Short 419 Calibrations while flying over Antarctica were added as from March 2011 to track the temporal 420 REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 19 variation of the voltage offset of the receivers, leading to the calibration plan shown in Figure 12, 421 which is the one currently used in SMOS. 422 423 424 425 Figure 12: SMOS routine in-orbit calibration plan 426 427 4.3 In-flight Calibration Improvements 428 Two improvements have been made based on the in-flight experience over these 5 years: the “warm” 429 external calibrations and the addition of an RFI check to validate the external calibrations. 430 4.3.1 Warm Calibrations 431 432 Detailed analysis of the external calibrations revealed that a few LICEF receivers of MIRAS exhibited 433 small and smooth unexpected jumps in their PMS (Power Monitoring System) detector voltages. These 434 jumps seemed to correlate well with the skin temperature of the antenna, happening more frequently for 435 colder skin temperatures, and appeared to be reversible in the sense that, for warmer skin temperatures, 436 the usual values were again obtained. To illustrate this, refer to the left panel in Figure 13, which spans 437 one full orbit flown pointing zenith during the Commissioning Phase, including the transitions from 438 and to Earth pointing at the beginning and end of the plot, respectively. The 3 cyan curves provide 439 the skin temperature of the antenna measured by 3 thermistors (named Tp7). The black and green lines 440 give, respectively, the elevation of the Sun over the antenna plane and its azimuth, in decadegrees as 441 REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 20 read from the scale on the right. The Sun elevation is negative (the Sun is behind the antenna) except 442 for a portion in the right half of the plot, where it reaches an elevation of about 30°. As soon as the Sun 443 appears in front of the antenna the skin temperature (cyan lines) increases from near 0°C till some 444 20°C, to return back to just 1°C or 2°C as the Sun sets behind the antenna horizon. The dark blue and 445 dark red crosses correspond to the detected voltages in the vertical and horizontal polarizations. The 446 vertical polarization shows some fluctuations. The first anomaly appears towards the left of the plot 447 when the physical temperature drops below some 10°C. At this moment the detected voltage jumps up 448 a few millivolts, which is unexpected because as the instrument is pointed towards cold sky the 449 detected voltage is expected to constantly decrease towards a minimum level. The anomalous higher 450 value is maintained until the temperature rises again above some 12°C. Then several fluctuations 451 happen centered around the maximum of temperature in the right half of the plot, showing a high 452 degree of symmetry and correlation with the temperature evolution. The detected voltage attains a right 453 value only in the center of these fluctuations, coinciding with the warmest temperature interval around 454 the peak. It is plausible that these fluctuations could be related to a change in the electrical phase of 455 some Teflon pieces of the antenna at some physical temperature range. Another example is given in the 456 right panel of Figure 13, which corresponds to another of the affected receivers, this time during a 457 typical external calibration manoeuvre: as the antenna cools down and its temperature reaches about 458 2°C, the detected voltages at the two polarizations experience jumps of 10 and 40 mV about. The 459 purple line in the right panel is the estimated physical temperature at which the Teflon parts of the 460 antenna could be. To avoid these voltage jumps, the external calibration manoeuvres are planned, since 461 October 2014, at a modified time to have the Sun at some positive elevation angle over the antenna 462 plane. The Sun illumination on the antenna keeps it warm, avoiding the skin temperature to fall too low 463 and the PMS detector voltage fluctuations. The positive Sun elevation is however kept below a limit of 464 10° to ensure that its presence does not degrade the external calibration acquisitions. Careful checks 465 were carried out to detect any effect from the direct signal of the Sun that could compromise the quality 466 REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 21 of the external calibration. In addition the warm calibrations perform the external calibration with the 467 antenna at a more similar temperature to the measurement mode, which is desirable, and reduce thermal 468 excursion on the antenna, improving reliability. Warm calibrations can be planned any time along the 469 year except around the equinoxes, when the Sun elevation is just too low, in which case, the usual cold 470 calibrations are performed instead. 471 472 473 Figure 13: Example of PMS detector voltage fluctuations in a zenith pointing orbit (left) and during a 474 typical external calibration (right) that led to the introduction of the external ‘warm’ calibrations 475 476 4.3.2 RFI Check in Validation of External Calibrations 477 478 External calibrations, where SMOS is pointed towards the Cold Sky, are executed only over the Pacific 479 Ocean to avoid picking up signals from strong RFI transmitters on ground through the back lobes of the 480 antennas. However, in one instance, an external calibration carried out 3 June 2015 in the North-481 Eastern Pacific Ocean, near Alaska, appeared contaminated by some ground interference. This caused 482 some disturbance in the data production chain as the calibration file had been ingested before the 483 problem was discovered. To avoid this, since then, every external calibration (these are performed once 484 every 2 weeks) is manually checked for RFI degradation before being accepted for use in the Level-1 485 data processor. An automatic procedure is being built up to replace the manual check. 486 487 REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 22 5 IN FLIGHT INSTRUMENT MONITORING 488 489 The values of some of the key instrument parameters which are monitored or calibrated in flight are 490 presented next. 491 5.1 Physical Temperature 492 5.1.1 Skin Antenna Temperature 493 494 MIRAS carries a thermistor (labeled Tp7) inside the head of the central screw of the antenna of the 3 495 Noise Injection Radiometers whose readings are representative of the physical skin temperature of any 496 of the antennas (Rubiales et al., 2015). This temperature is important because it affects the amount of 497 noise emitted by the front end equivalent resistor. It also provides an indication on how different the 498 thermal conditions of the antenna are between an external calibration and the nominal measurement 499 mode. 500 The evolution of the temperature readings from the 3 Tp7 thermistors is shown in Figure 14. The skin 501 temperature goes through its largest excursion (from about 6°C to 28°C) during every boreal winter 502 solstice, when the Sun reaches maximum elevation above the antenna plane (around 31°) and is 503 eclipsed by the Earth. There is a second period of large thermal excursion (from 8°C to 18°C 504 approximately) around every boreal summer solstice where the Sun elevation reaches up to 15° 505 elevation above the antenna horizon. During the equinoxes the temperature excursion is the smallest 506 (between 5°C to 12°C) and the lowest skin temperatures are recorded, except for the external 507 calibration events. The latter correspond to the individual spikes that drop below 0°C in Figure 14. The 508 Tp7 temperatures went through an initial cooling transient, clearly observed during the first half of 509 2010, to then flatten out into a very small long term residual cooling trend. 510 511 REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 23 512 Figure 14: Evolution of the skin antenna temperature measured by the 3 Tp7 thermistors 513 514 515 5.1.2 Inner Receiver Temperature 516 517 Every one of the 72 LICEFF receivers of MIRAS has a thermistor (labelled Tp6) next to an internal 518 matched load in the front-end electronics used as warm point in the amplitude calibration. This 519 thermistor senses the inner temperature of the receiver. The average value of Tp6 across all LICEF 520 receivers is shown in Figure 15. The physical temperature of the receivers is seen to be quite stable 521 along the mission, centered around 22°C with a global peak to peak variation of about 1°C. As for Tp7, 522 the Tp6 readings present larger excursions during the solstices, and narrower variation around the 523 equinoxes, where its lowest values are attained. 524 525 5.2 Receiver Parameters 526 5.2.1 Antenna Losses 527 528 The antenna has two distinct loss components: one due to the radiating resonant cavity, and the other 529 due to the intermediate layer circuit that combines the signal from the pair of probes of each 530 polarization. The first component is tiny and difficult to measure on ground. It was estimated to be of 531 about L1=0.05 dB, by calculations based on the geometry and materials of the antenna design. On the 532 other hand, the losses of the intermediate layer circuit, of about L2=0.25 dB, was measured on ground. 533 The total antenna losses are then expected to be around 0.30 dB. During the in orbit calibration, the 534 REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 24 antenna loss are directly measured using the Cold Sky and the internal matched load (Corbella et al., 535 2012). The average value across all LICEF receivers for each polarization is shown in Figure 16. The 536 in-flight measured antenna losses are about 0.17 dB larger than their pre-launch estimated value. It is 537 worth noticing the rapid evolution exhibited during the first 6 months of the mission, as well as the 538 seasonal fluctuations, the latter being partly driven by the PMS detector voltage fluctuations described 539 earlier. The antenna losses present a different evolution after October 2014, reflecting the introduction 540 of the warm external calibrations to avoid the mentioned PMS fluctuations. 541 542 543 Figure 15: Average inner LICEF receiver temperature Tp6. The spikes in early Jan’10, May’10 and 544 Jan’11 are due to 3 anomalies occurred in the instrument. 545 546 547 Figure 16: Evolution of the antenna losses as measured in flight 548 549 550 5.2.2 Receiver Detector Gains 551 552 The average detector gain across all LICEF receivers is presented in Figure 17 for each polarization. 553 The absolute gain is shown in the left panel. Similar features to those found in the evolution of the 554 REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 25 antenna losses are repeated here: an initial rapid transient followed by seasonal variations. In addition 555 the receiver gains seem to be undergoing an exponential decay which, according to the relative gain 556 variation shown in the right panel, has accumulated a total decrease of about 1.5%. The reason for this 557 decay is unknown, but could be caused by the overall thermal stabilization over mission life time. 558 559 5.2.3 Receiver Detector Voltage Offets 560 561 The average voltage offset across all LICEF receivers is shown in Figure 18. The behavior is somewhat 562 erratic, without any clear trend, with rapid fluctuations that led, in March 2011, to the introduction of 563 weekly short calibrations as from March 2011 to track them. The voltage offset is therefore well 564 followed with a weekly refresh rate and calibrated out. 565 566 567 Figure 17: Evolution of the end-to-end average receiver gain in mV/K (left) and in percentage variation 568 taking June 2011 as reference value (right) 569 570 571 572 REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 32 6.1.2 Use of the Relative Phase between Polarizations Measured on Ground 699 700 During the Image Validation Test (IVT) of MIRAS that was carried out on ground before launch, a set 701 of 4 probes was placed in the ceiling of the Maxwell Electromagnetic Compatibility chamber of 702 ESTEC to measure the relative phase between all LICEF receivers (Corbella et al., 2009). The 703 instrument was operated in both dual and full polarization. For version V505 of the Level-1 processor, 704 2 separate sets of relative phases were retrieved from the IVT test: one with the relative phases for the 705 horizontal polarization and another one for the vertical polarization. When preparing the next version, 706 V620, of the processor, it became clear that the two sets of relative phases could have an offset between 707 them which had to be corrected. Such phase bias across the two polarizations was in fact causing 708 distortions in the Stokes-3 and, most clearly, Stokes-4 parameters. Therefore, the IVT data set was re-709 analyzed to determine the missing phase offset between the two polarizations, which was found to be of 710 −6.8°. This value was verified using Stokes-4 images over the ocean: as shown in Figure 24, the error 711 (sigma displayed in the lower left corner) was indeed minimized for a phase offset close to the retrieved 712 one. All phases corresponding to vertical polarization were then reduced by that amount in version 713 V620 of the processor. 714 715 716 Figure 24: Validation of the relative phase between polarizations using Stokes-4 over ocean. 717 718 6.1.3 Use of Average Antenna Patterns across 3 Frequencies Measured on Ground 719 720 The antenna patterns of all and every element of MIRAS embedded in the array were carefully 721 measured in an antenna test range. The measurements were performed at the center frequency, 1413.5 722 MHz, as well as at the band edges, 1404 and 1423 MHz. In version V505 of the Level-1 processor only 723 REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 33 the patterns at the center frequency were used. In version V620, the average antenna pattern across the 724 3 measured frequencies is used, which reduces slightly the spatial ripple, as shown in the left lower 725 corner of each panel of Figure 25. 726 727 Figure 25: Deviation images from an ocean model in X-polarization using 1 pattern (left) or the 728 average of the 3 measured patterns on ground (right) –similar results were obtained in Y-polarization–. 729 730 6.1.4 Replacement of Antenna Pattern of Hinge Elements by their Neighbour’s 731 732 Analyzing in detail the measured antenna patterns, it was noticed that those corresponding to elements 733 next to one of the hinges of the deployed arms (upper left panel of Figure 26), presented some ripples 734 which the elements in the center of the arm segments did not exhibit (upper center panel). Some 735 research led to the conclusion that these ripples were caused by travelling waves between the arm and 736 the supporting structure that was used to hold it during the measurements, as shown in the lower left 737 panel of Figure 26, which would leak out through the next hinge and segment contour causing a typical 738 interference pattern. Since the instrument in flight configuration does not have any supporting structure 739 in the back, the real patterns of the hinge elements should be free of fringes and similar to those of the 740 central elements in each arm segment. For this reason, in the Level-1 processor version V620, the 741 antenna patterns of the hinge elements was replaced by that of its inner neighbours, which reduces 742 slightly the spatial ripple of the images, as illustrated in the right panel of Figure 26, to be compared to 743 the left panel of Figure 25. 744 6.1.5 Use of Only the Most Stable NIR Unit 745 746 REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 34 From Figure 22, it is seen that the NIR-CA unit (bottom panels) is the most stable of the 3 Noise 747 Injection Radiometers of MIRAS. To assure the best temporal stability, it was decided to use, in V620, 748 only the antenna temperature measured by NIR-CA, and not those from NIR-AB and NIR-BC, for the 749 visibility sample at the origin V(0,0). All 3 NIR units are still employed in the measurements of other 750 visibility samples outside the origin as well as in the measurement of the noise diodes of the on-board 751 Calibration System. 752 753 754 Figure 26: Antenna patterns of elements next to a hinge (top left) and away from it (top center); 755 electromagnetic simulation showing leakage between arm and back supporting structure (left bottom); 756 image obtained by replacing hinge patterns by their neighbour’s (right). 757 758 6.1.6 Use of In-Orbit Antenna Loss 759 760 As mentioned earlier, antenna losses L1+L2 could not be accurately measured on ground due to set up 761 uncertainties, and instead, were characterized in orbit, for V620 of the Level-1 processor, thanks to the 762 more favourable external calibration manoeuvres (refer to Figure 16). The final estimation of the 763 antenna loss was carried out after removing those instances with detector voltage fluctuations as the 764 examples in Figure 13. Once the total loss had been measured, a second step, critical for the temporal 765 stability, was performed, that of determining the L1 and L2 separately for the Noise Injection 766 Radiometers. As the external calibration could only provide the ensemble loss, the split into its two 767 REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 35 contributions was done by minimizing the orbital variations, taking as metrics the descending minus 768 ascending deviations over the ocean with respect to a forward model. Such exercise resulted in a 769 minimum of orbital variation for L1=0.15 dB as illustrated in Figure 27, the remaining of the total loss 770 measured in orbit being assigned to L2. The values used in V620 for the L2 antenna loss of each 771 receiver and polarization are constant (do not change over time) and equal to their average value 772 obtained across all external calibrations, using the optimized L1=0.15 dB. This procedure to split the 773 antenna losses in its two contributors L1 and L2 was based on the much stronger relationship of 774 L1 with orbital variations than L2, L1 being tightly influenced by the skin temperature of the 775 antenna. 776 777 778 Figure 27: Split of NIR antenna losses achieved by minimizing the descending minus ascending pass 779 difference of the deviations of the first Stokes parameter (divided by 2) from a model of the brightness 780 temperature of the ocean (in Kelvin) 781 782 6.1.7 Improved Gibbs-1 Image Reconstruction 783 784 In addition to being fully polarimetric, as commented earlier, the image reconstruction of V620, 785 depicted in Figure 28, brings two other main improvements over V505. The first one is the 786 implementation of the whole processing on hexagonal grids, as opposed to rectangular ones, to avoid 787 interpolation errors. The second consists of a new Gibbs-1 ‘model approach’ by which the image 788 reconstruction is performed over residual visibilities resulting from subtracting the estimated 789 contribution of the Corbella term, the sky and the Earth, from the measured visibilities. 790 REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 36 791 792 Figure 28: Improved Gibbs-1 model approach for image reconstruction as implemented in V620. 793 794 6.1.8 Improved Removal and Flagging of Sun and RFI sources 795 796 Since version V505, the algorithm for the Sun removal has been continually improved. In V620 the 797 Sun brightness temperature is estimated through a 4-point spatial interpolation instead of taking the 798 single closest neighbour point, this allowing a much finer positioning of the energy and thus, a 799 reduction of the Sun tails. Several enhancements were also done to the flagging of the pixels affected 800 by the Sun brightness temperature, in particular, the dynamic adjustment of the width of the tails of the 801 real Sun and its aliases for every snapshot, and the correct flagging of all tails (one of the tails was not 802 properly flagged in the previous version), as shown in Figure 29. As for the flagging and treatment of 803 RFI sources and their impact, the improvements are reported in (Oliva et al., in press) (Khazaal et al., 804 2014) (Daganzo-Eusebio et al., 2013). 805 REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 37 806 807 Figure 29: SMOS extended alias-free field of view showing, in grey, the pixels which are flagged 808 due to the Sun tails (real Sun is on the right, with 2 alias on left top and left bottom). The width of 809 the tails with flagged pixels has been increased with the improved Sun tails dynamic flagging of 810 V620, and all tails are now flagged (note the lower right corner tail in V6). 811 812 6.2 Performance of Level-1 Processor Version V620 813 This section is devoted to present the performance of the currently operational version of the Level-1 814 processor, V620, in comparison with the earlier version V505. The performance has been assessed over 815 the entire data set of the two reprocessing campaigns, one with each processor version. The 816 quantification of the performance has been carried out following some defined metrics, as depicted in 817 Figure 30, comprising: calibration parameters, temporal stability (orbital, seasonal and yearly) of both 818 the antenna and the brightness temperatures, systematic spatial errors in the images (bias and ripple), 819 Sun correction, land-sea contamination and random noise. The Cold Sky, the Pacific Ocean and 820 Antarctica are the targets where the metrics have been evaluated, by comparing observations to a 821 surveyed map of the sky, a radiative transfer model of the ocean (Ocean Target Transformation 822 or OTT) or simply their average value over Antarctica (Ice Target Transformation or ITT) 823 respectively. 824 REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 38 825 Figure 30: Overview of the metrics used to evaluated the performance of the Level-1 processor. 826 827 6.2.1 Removed Negative Slope at High Elevation Angles 828 829 V505 suffered from a negative slope in the dependence of the brightness temperature with elevation 830 angle, in a way that measurements at low incidence were too cold when compared against modeled 831 values. This is illustrated in the left panel of Figure 31 which shows the brightness temperature 832 measured by SMOS, in horizontal and vertical polarizations, as a function of incidence angle over 833 Dome-C station in Antarctica (Macelloni et al., 2013). As reference, the red dashed line represents a 834 model, the diamonds are Aquarius observations and the solid circles ground measurements. Besides the 835 discrepancy at high incidence angles which is expected due to the unavoidable sky contamination in the 836 ground observations (through the part of the main lobe above the ice horizon), the SMOS 837 brightness temperatures are clearly colder than the model towards 0° incidence. The enhancements 838 implemented in V620 described earlier (in particular in Section 6.1.7) resulted in a substantial 839 reduction of such cold trend at low incidence, as can be seen in the right panel of Figure 31. It is also 840 worth noting that, for V620, the ripples along incidence angle are smaller and the match with Aquarius 841 and Dome-C ground radiometer is better. 842 REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 39 843 844 Figure 31: Performance of V505 (left) and V620 (right) with incidence angle over Antarctica. 845 846 6.2.2 Lower Spatial Ripple 847 848 Although there is a limit to how much the spatial ripple can be removed as explained before, Figure 32 849 shows that V620 achieves about 0.2 K lower spatial ripple in both polarizations than V505, thanks to 850 the improvements in Sections 6.1.1 through 6.1.4 and 6.1.7. The root mean square spatial ripple of 851 V620 over most of the Extended Alias-Free Field of View is therefore of about 1.5 and 2.0 K for X and 852 Y polarizations respectively, evaluated over the ocean. It has to be noted that the bias of V620 is 853 warmer than that of V505, overshooting almost 1 K in X polarization above the forward ocean model. 854 855 Figure 32: Spatial ripple and bias of V505 (left) and V620 (right) over Ocean (colour scale in Kelvin) 856 857 6.2.3 Improved Stokes-3 and Stokes-4 Parameters 858 859 REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 40 The fully polarimetry operation of V620 including the cross-polar patterns with corrected sign 860 convention (Section 6.1.1) and relative phase between polarizations (Section 6.1.2) yield significant 861 improvements in the Stokes-3 and Stokes-4 parameters (Lin et al., 2013). This is clearly shown in 862 Figure 33, where more uniform residuals against the ocean model are obtained with V620 (right 863 column) than with V505 (left column). The cleaner polarimetric brightness temperatures provided by 864 V620 allows accurate estimation of the ionospheric total electron content and the Faraday rotation 865 angle directly from SMOS observations (Corbella et al., 2015). 866 867 868 869 Figure 33: Residual Stokes-3 (top) and Stokes-r4 (bottom) of V505 (left) and V620 (right) over Ocean 870 (colour scale in Kelvin) 871 6.2.4 Removed Latitudinal Drift 872 873 The red line in Figure 34 shows the deviation of the Stokes-1/2 parameter (average of the vertical and 874 horizontal brightness temperatures) against a forward model along a strip of the Pacific Ocean from 50° 875 South to 20° North, for version V620. The plot is for one orbit, but the general behaviour is 876 systematic, with some seasonal variations as shown in the Hövmoller plot in the right panel of 877 Figures 6. Besides some variations on each side of the equator which can be related to geophysical 878 signals (unmodeled rain and surface roughness effects), there is no tendency with latitude as 879 expected, thanks mainly to the improvement in Section 6.1.6. This is an important improvement 880 REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 41 over V505 processor, for which the Stokes-1/2 parameter, shown by a blue line on the same figure, did 881 have a significant slope over latitude. 882 883 884 885 Figure 34: Latitudinal trend of the brightness temperature deviation from an ocean model, 886 averaged in the alias-free field of view, of V505 (blue line), and corrected value in V620 (red line). 887 888 6.2.5 Reduced Orbital Variations 889 890 The reduced orbital variations of V620 in the Stokes-1/2 parameter over the ocean have already been 891 introduced and shown in the right panel of Figure 7. What is presented in the left panel of the same 892 figure is the strong orbital variations, of about 2 K peak to peak, that V505 exhibited between 893 ascending and descending passes in the same Stokes-1/2 parameter. These pronounced variations have 894 been reduced in V620 mostly thanks to the optimization and use of the antenna losses measured in orbit 895 (Section 6.1.6) and the use of only the most stable NIR unit (Section 6.1.5). 896 6.2.6 Improved Yearly and Seasonal Stability 897 898 Figure 6 shows the latitudinal, seasonal and yearly variations of the Stokes-1/2 parameter over the 899 ocean. The left plot, for V505, has strong variations with latitude which, as discussed above, have been 900 removed in V620, shown in the right panel. V505 presents also significant seasonal variations, seen as 901 ±2 K alternating blue and red bands in the corresponding Hovmöller plot. These strong seasonal 902 REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 48 Macelloni, G, Brogioni M., Pettinato, S., Zasso, R. Crepaz, A., Zaccaria, J., Drinkwater, M., (2013), “Ground 1103 based L-Band Emission Measurements at DOME-C Antarctica: the DOMEX-2 experiment. IEEE Trans. Geosci. 1104 and Remote Sensing, vol 51, No 9, 4718-4730. 1105 1106 Martín-Neira M., S. Ribó, A. J. Martín-Polegre, (2002), “Polarimetric mode of MIRAS,” IEEE Transactions on 1107 Geoscience and Remote Sensing, vol. 40, no. 8, pp. 1755–1768. 1108 1109 Martín-Neira M., M. Suess, J. Kainulainen, F. Martín-Porqueras, (2008), “The Flat Target Transformation”, 1110 IEEE Transactions On Geoscience And Remote Sensing, Vol. 46, No. 3, 613 – 620. 1111 1112 Martin-Porqueras F., J. Kainulainen, M. Martín-Neira, I. Corbella, R. Oliva, R. Castro, J. Barbosa, A. Gutierrez, 1113 (2010), “Experimental Validation of the Corbella's Visibility Function using HUT-2D and MIRAS”, Proceedings 1114 of IGARSS-2010, Honolulu, Hawaii (US), pp.4286-4289. 1115 1116 McMullan K., M. Brown, M. Martín-Neira, W. Rits, S. Ekholm, J. Marti and J. Lemanczyk, (2008), “SMOS: 1117 The payload”, IEEE Transactions On Geoscience and Remote Sensing, Vol. 46, No. 3, 594 – 605. 1118 1119 Mecklenburg Susanne, Matthias Drusch, Yann H. Kerr, Senior Member, IEEE, Jordi Font, Manuel Martín-Neira, 1120 Steven Delwart, Guillermo Buenadicha, Nicolas Reul, Elena Daganzo-Eusebio, Roger Oliva, and Raffaele 1121 Crapolicchio, (2012), “ESA’s Soil Moisture and Ocean Salinity Mission: Mission Performance and Operations”, 1122 IEEE TGARS, Vol 50, No 5, pages 1354-1366. 1123 1124 Mecklenburg S., M. Drusch, Y. Kerr, J. Font, N. Reul, M. Martín-Neira, R. Oliva, E. Daganzo-Eusebio, J. 1125 Fauste, L. Kaleschke, N. Rodriguez-Fernandez, P. de Rosnay, G. Macelloni, J. Muñoz-Sabater, K. Rautiainen, 1126 J.P. Grant, N. Verhoest, H. Lievens, P. Richaume, R. Sabia, S. Delwart, R. Crapolicchio, A. de la Fuente, M. 1127 Kornberg, (in press), “ESA’s Soil Moisture and Ocean Salinity Mission: from Science to Operational 1128 Applications”. This RSE special issue, submitted. 1129 1130 Oliva R., E. Daganzo, Y. Kerr, S. Mecklenburg, S. Nieto, P. Richaume, C. Gruhier, (2012), "SMOS Radio 1131 Frequency Interference Scenario: Status and Actions Taken to Improve the RFI Environment in the 1400-1427 1132 MHz Passive Band". IEEE Trans Geosci. Remote Sens., Special issue on Soil Moisture and Ocean Salinity 1133 Mission. Vol. 50, No 5, pp. 1427-1439. 1134 1135 Oliva R., M. Martín-Neira, I. Corbella, F. Torres, J. Kainulainen, J.E. Tenerelli, F. Cabot, F. Martin-Porqueras, 1136 (2013), “SMOS Calibration and Instrument Performance after One Year in Orbit”, IEEE Transactions on 1137 Geoscience and Remote Sensing, Vol. 51, No 1, pp. 654-670. 1138 1139 Oliva R., E. Daganzo, Y. Soldo, Y. Kerr, F. Cabot, P. Richaume, E. Anterrieu, A. Gutierrez, J. Barbosa, G. 1140 Lopes, (in press), “Status of RFI in the 1400-1427 MHz passive band: The SMOS perspective”, This RSE 1141 special issue, submitted. 1142 1143 Rubiales P., J. Closa, E. Checa, S. Dolce, M. Kornberg, M. Martín-Neira, (2015), “SMOS Payload Thermal 1144 Control: Review of performances after 5 years in orbit operation”, 45th International Conference on 1145 Environmental Systems, Bellevue, Washington (US). 1146 1147 SMOS Level 2 Ocean Salinity Team, “SMOS L2 Ocean Salinity Algorithm Theoretical Baseline Document”, 1148 ARGANS, SO-TN-ARG-GS-0007_L2OS-ATBD_v3.12_150731, July 2015. 1149 1150 Torres F., I. Corbella, A. Camps, N. Duffo, M. Vall-llossera, S. Beraza, C. Gutierrez, M. Martín-Neira, (2006), 1151 “Denormalization of Visibilities for In-Orbit Calibration of Interferometric Radiometers”, IEEE Transactions on 1152 Geoscience and Remote Sensing. Vol. 44, No. 10, pp.2679-2686. 1153 1154 Torres F., I. Corbella, L. Wu, N. Duffo, J. Gourrion, J. Font, M. Martín-Neira, (2012), “Minimization of Image 1155 Distortion in SMOS Brightness Temperature Maps Over the Ocean”, GeoScience and Remote Sensing Letters, 1156 Vol. 9, No. 1, pp.18-22. 1157 REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 49 1158 Torres F., I. Corbella, I. Duran, W. Lin, N. Duffo, M. Martín-Neira, (2014), “Residual Calibration Error Impact 1159 On SMOS SLL Performance”, IGARSS-2014 Proceedings, pp.2542–2545, Quebec (Canada). 1160 1161 Torres F., I. Corbella, L. Wu, N. Duffo, S. Delwart, M. Martín-Neira, (2015), “In-Orbit Validation of SMOS Full 1162 Polarimetric Equations”, IEEE Geoscience and Remote Sensing Letters, Vol. 12, Issue 3, pp.458-461. 1163 1164