The Cryosphere, 9, 357–366, 2015 www.the-cryosphere.net/9/357/2015/ doi:10.5194/tc-9-357-2015 © Author(s) 2015 CC Attribution 3.0 License Comparing C- and L-band SAR images for sea ice motion estimation J Lehtiranta, S Siiriä, and J Karvonen Finnish Meteorological Institute, Marine Research Programme, Helsinki, PB 503, 00101 Finland Correspondence to: J Lehtiranta (jonni.lehtiranta@fmi.fi) Received: 22 April 2014 – Published in The Cryosphere Discuss.: 26 May 2014 Revised: 31 October 2014 – Accepted: November 2014 – Published: 17 February 2015 Abstract Pairs of consecutive C-band synthetic-aperture radar (SAR) images are routinely used for sea ice motion estimation The L-band radar has a fundamentally different character, as its longer wavelength penetrates deeper into sea ice L-band SAR provides information on the seasonal sea ice inner structure in addition to the surface roughness that dominates C-band images This is especially useful in the Baltic Sea, which lacks multiyear ice and icebergs, known to be confusing targets for L-band sea ice classification In this work, L-band SAR images are investigated for sea ice motion estimation using the well-established maximal crosscorrelation (MCC) approach This work provides the first comparison of L-band and C-band SAR images for the purpose of motion estimation The cross-correlation calculations are hardware accelerated using new OpenCL-based source code, which is made available through the author’s web site It is found that L-band images are preferable for motion estimation over C-band images It is also shown that motion estimation is possible between a C-band and an L-band image using the maximal cross-correlation technique Introduction The Baltic Sea gets an ice cover every winter, covering 45 % of its area on an average year In the northern Bay of Bothnia, the typical duration of ice cover is from late October to late May, and the greatest level ice thickness ranges from 50 to 110 cm The bay has an average depth of 41 m and typically has large areas of landfast ice on the eastern and northeastern coasts (Myrberg et al., 2006) Observations of the Baltic sea ice are for winter navigation safety Work has been done to calculate sea ice motion from two consecutive satellite im- ages using different optical flow estimation algorithms (e.g., Fily and Rothrock, 1987; Vesecky et al., 1988; Liu et al., 1997; Karvonen et al., 2007; Thomas et al., 2011), and this approach has provided acceptable results using the C-band synthetic aperture radar, which is regarded as a good compromise for sea ice remote sensing (Dierking and Busche, 2006) This work will compare C-band (38–75 mm wavelength) with L-band (150–300 mm wavelength) for sea ice motion estimation Motion estimation from consecutive satellite images has its limitations Only an average velocity can be determined, and that only if the ice surface remains mostly unchanged Weather conditions can change ice surface properties enough to make feature detection impossible Generally the method only works for image pairs typically less than days apart, naturally depending on the rate of the ice drift and deformation Previous work has also concentrated on sequential images from a single instrument, which places a limitation on the availability of suitable image pairs A satellite might fly over the area of interest only once per day or less For longer time intervals, velocities due to short-duration events such as storms are lost If observations from multiple satellites are used, image pairs mere hours apart are easier to find, but the benefit comes with the added difficulty of comparing images of fundamentally different character To improve the situation, this work will examine the idea of calculating sea ice motion using two pictures from different instruments, namely EnviSAT ASAR (56.2 mm wavelength), RadarSAT-2 SAR (55.5 mm wavelength) and ALOS PALSAR (236 mm wavelength) Published by Copernicus Publications on behalf of the European Geosciences Union J Lehtiranta et al.: Comparing C- and L-band SAR images for sea ice J Lehtiranta et al.: Comparing C- and L-band SAR images for sea ice motion estimation 358 15 RS2cC EScC RS2cC ALOScL 10 EScC ALOScL W −5 temperaturec(oC) SARcimages Figure Satellite images used in this work, normalized for viewing Details given in Table ©MDA, ESA and JAXA 15.03.c00:00 tag satellite time (UTC) t band R1 E1 R2 A1 E2 A2 RadarSAT EnviSAT RadarSAT ALOS EnviSAT ALOS 16 Mar 2009, 04:59 16 Mar 2009, 19:54 17 Mar 2009, 16:00 17 Mar 2009, 20:12 18 Mar 2009, 09:04 18 Mar 2009, 09:36 t0 t0 + 14:55 t0 + 35:01 t0 + 39:13 t0 + 51:05 t0 + 51:37 C C C L C L 100 Data and methods For this work, a set of synthetic-aperture radar (SAR) images from March 2009 were used (see Fig ) C-band images 105 were available from both EnviSAT ASAR and RadarSAT 2, while L-band images were available from ALOS PALSAR A set of six images were chosen for the time period between 16 and 18 March These days were chosen because there were a relatively large number of images available, in110 cluding two L-band images Additionally, two of the images were of different frequency bands and almost simultaneous, with only 32 m between them This is desirable for comparing frequency bands, and a unique occurrence in the set of images that were available The images were resampled to 100 m pixel size, approximately corresponding to the nominal resolution of the employed ScanSAR capturing mode 115 Lots of changes including compaction and lead opening were present during this period Landfast ice and open water areas were seen in visual inspection, as well as different types of drift ice As the ice cover in other parts of the Baltic was sparse, only the seas north of 63◦ N latitude were considered 120 2.1 Weather and ice conditions during the experiment period For the Baltic Sea, the winter 2008–2009 was milder and 125 shorter than average Freezing commenced in the Bay of Bothnia in the second half of November, but the ice cover exThe Cryosphere, 9, 357–366, 2015 16.03.c00:00 17.03.c00:00 18.03.c00:00 19.03.c00:00 Figure Wind and air temperature recorded by the Kemi lighthouse weather station (65.385◦ N, 25.096◦ E) during the experiment Fig Wind and air temperature recorded by the Kemi lighthouse period Timing of SAR images is also marked, red for C-band and weather station (65.385N,25.096E) during the experiment period blue for L-band images Timing of SAR images is also marked, red for C-band and blue for L-band images Table List of satellite images used in this work # Windcspeedc(m/s) Windcdirection tended across the Bay of Bothnia only in the end of January February was a normal winter month, and the maximum ice images During the 16th and 17th March, strong southwestcover, 110 000 km2 , was recorded on 20 February Much of erly winds were pushing the ice pack towards the north this ice was thin, and after a cold period, warmer southwestEventually the wind turned north On the 18th much of the erly winds pushed ice northwards during March On March ice had returned southwards and new leads had formed The 16, only the Bay of Bothnia and northern Gulf of Finland had temperature remained at or below freezing point It is asa significant ice cover (The Baltic Sea Portal, 2009) sumed that no significant melting took place during the exFigure summarizes the weather conditions recorded periment and that melting did not affect the motion estimaby a weather station at the Kemi lighthouse (located at 130 tion results Formation of new ice, however, needs to be taken 65.385◦ N, 25.096◦ E) during the acquisition of the satellite into account images During 16 and 17 March, strong southwesterly winds As reported in ice charts, most of the drift ice in the Bay of were pushing the ice pack towards the north Eventually the Bothnia is deformed, mostly by ridging but also rafting Not wind turned north On 18 much of the ice had returned southmuch level ice remains, the well-defined areas being west wards and new leads had formed The temperature remained 135 of the island of Hailuoto and southwest from Tornio There at or below the freezing point It is assumed that no signifiis no new ice to be found, but large sections of landfast ice cant melting took place during the experiment and that meltlie around the coastline Reported level ice thicknesses range ing did not affect the motion estimation results Formation of from 10 to 50 cm in the drift ice and up to 70 cm in landfast new ice, however, needs to be taken into account ice Six icebreakers were on duty assisting ships As reported in ice charts, most of the drift ice in the Bay of 140 Bothnia is motion deformed, mostly by ridging but also rafting Not 2.2 The estimation approach much level ice remains, the well-defined areas being west of and southwest from Tornio There Forthe thisisland work,ofa Hailuoto straightforward block cross-correlation prois no new ice to be found, but large sections of landfast ice gram was written in the general purpose C++ programlie around the coastline Reported level ice thicknesses range ming language The code works directly in the spatial do- 145 from 50 cmmore in theflexibility drift ice and up to 70 cmthe in landfast main,10 totoallow in fine-tuning compuice Six icebreakers were on duty assisting ships tational parameters (Emery et al., 1991) and to allow easy parallelization Critical parts of the algorithm were imple2.2 Theonmotion estimation approach mented GPU calculation units and programmed using the Open Computing Language (OpenCL) C OpenCL is a 150 For this work, a straightforward block cross-correlation proportable language for writing code that can be run in a pargram was written in the general purpose C++ programming allel fashion on a variety of devices (Stone et al., 2010) language The code works directly in the spatial domain, to This approach cut down the calculation time significantly allow normalized cross-correlation, more flexibility in fineThe OpenCL cross-correlation program can process one pair tuning the computational parameters (Emery et al., 1991) and www.the-cryosphere.net/9/357/2015/ Fig Tru 2009, 10:0 board the of image utes for source c http://jon The m resolution that has t only 48 k the search vectors w inal or 80 ments), a guesses f high-reso lematic v chosen to For the search tradeoff i large eno time sma separatin end of p due to de of discer error frac images to The m SAR images for sea ice motion estimation J Lehtiranta et al.: Comparing C- and L-band SAR images for sea ice motion estimation 359 errors due to deformations, and to concentrate on errors due to lack of discernible patterns within these windows This way the error fractions are maximally useful for comparing C-band images to L-band images The method consists of the following steps: C OScL re-projecting and cropping satellite images using the GDAL toolset, loading the GeoTIFF images, translating 16-bit greyscale values to floating-point numbers, W generating a resolution pyramid for both images, using a 2-D low-pass filter and decimating for every level, running normalized cross-correlation for coarseresolution image windows, 19.03.c00:00 median-filtering the coarse result to produce the average motion field and first guess for next step, lighthouse ent period nd blue for outhwesthe north uch of the med The It is asng the exn estimao be taken 130 he Bay of fting Not eing west io There 135 ndfast ice ses range n landfast 140 ation proprogrampatial do- 145 e compullow easy re implemed using enCL is a 150 in a parl., 2010) nificantly s one pair Figure True colour satellite image of the Bay of Bothnia, 18 Fig 2009, True color imagecaptured of the Bay of Bothnia, 18 March March 10:05satellite UTC Image by the MODIS instrument 2009, 10:05 UTC.satellite, Image captured the MODIS instrument on on board the Terra courtesy by of NASA board the Terra satellite, courtesy of NASA easy parallelization Critical parts of the algorithm were programmed in OpenCL C, which is a portable language for of images in roughly 20 seconds, as opposed to 20 minwriting code that can be run in a parallel fashion on a vautes for a single-core program running on the CPU This riety of devices (Stone et al., 2010) The cross-correlation source code is available through the author’s website at code was run on a Graphics Processing Unit (GPU) prohttp://jonni.lehtiranta.net/ duced by NVIDIA This approach cut down the calculation The motion vectors were calculated using a multitime significantly The OpenCL cross-correlation program resolution approach This is usually done to limit the area can process one pair of images in roughly 20 s, as opposed that has to be processed, but because of the GPU approach, to 20 for a single-core program running on the CPU only 48 kB of fast local memory was available The size of This source code is available through the author’s website the search domain was limited to 96x96 pixels First, motion at http://jonni.lehtiranta.net/ vectors were calculated in a coarse resolution (1/8 of the origThe motion vectors were calculated using a multiinal or 800 m / pixel, which allows almost 40 km displaceresolution approach This is usually done to limit the area ments), and median-filtered result vectors were used as initial that has to be processed, but because of the GPU approach, guesses for the high-resolution matching step Finally, the only 48 kB of fast local memory was available The size of high-resolution result was median-filtered to remove probthe search domain was limited to 96 × 96 pixels First, molematic values For this work, the median filtering radius was tion vectors calculated in et a al., coarse resolution (1/8 of chosen to be were (as in Karvonen 2007) −1 , which theFor original or 800 m pixel allows almostwith 40 km the image windows that were cross-correlated the displacements), and median-filtered result vectors were search domain, a size of 16x16 pixels was chosen Thereused is a as initialinvolved guesses in forchoosing the high-resolution step.toFitradeoff this window matching size, as it has be nally, the high-resolution result was median-filtered relarge enough to contain a discernible pattern, and at theto same move problematic Foritsthis work, the median time small enoughvalues to retain structure in the time filtering interval radius was chosen to be (as in Karvonen et al., separating the pair of images The chosen size is 2007) at the small Forofthe image windows were cross-correlated the end practical options that It was chosen to minimizewith errors search domain, a size of 16 × 16 pixels was chosen There is due to deformations, and to concentrate on errors due to lack aoftradeoff involved in choosing this window size, as it has to discernible patterns within these windows This way the be large enoughareto maximally contain a discernible pattern andC-band at the error fractions useful for comparing same time small enough to retain its structure in the time images to L-band images interval separating the pair of images The chosen size is at The method consists of the following steps: the small end of practical options It was chosen to minimize www.the-cryosphere.net/9/357/2015/ running normalized cross-correlation for the finestresolution image windows, saving this result and a median-filtered version (radius 3) of it in an ASCII text file The results were analysed and plotted using the Matlab and Octave programs 2.3 Performance metrics for motion estimation For this study, no ground truth data was available for comparison It was necessary to define some performance metric that could be calculated from the results alone In this work, the cross-correlation method was not tuned for the image types, and especially between C- and L-band images, low crosscorrelation coefficients were expected Instead of the crosscorrelation coefficient itself, we consider the ratio of the two highest peaks While a high peak-to-peak ratio is not conclusive evidence of correctness, it is assumed to be a necessary requirement A motion vector is rejected if the margin between two highest cross-correlation peaks is less than 15 %, and otherwise accepted in a “peak margin” sense Additionally, each motion result is evaluated against the expectation of uniformity, flagging as errors all vectors that differ significantly from the median-filtered vector field It is assumed that the median filtering succeeds at removing spurious values and retains real stepwise changes in the ice motion field (Astola et al., 1990), so that the median-filtered motion field represents the real average motion Even when this is not the case, unrealistic vectors will not match it so these cases cannot produce false successes A motion vector is rejected if it differs from the median of its neighbourhood by more than 500 m Otherwise it is considered acceptable in a “regularity” sense Both criteria are arbitrary However, they appear to be sensible choices for this study The Cryosphere, 9, 357–366, 2015 360 J Lehtiranta Lehtiranta et et al.: al.: Comparing Comparing CC- and J and L-band L-band SAR SAR images images for for sea sea ice ice motion motion estimation estimation Figure Screenshot of the motion estimation program written for this work (a) Zoom-in of the first image with some detected motion Fig Screenshot of the motion estimation program thisrepresents work a) zoom-in of the first image withblack somerepresents detected motion vectors vectors (b) The cross-correlation result for the circledwritten vector.for White maximum cross-correlation, zero correlation b) the cross-correlation result for the circled vector White represents maximum cross-correlation, black represents zero correlation andpair the and the area left outside of the calculation Red represents negative cross-correlation (c) Aligned zoom-in of the second image of the area left the calculation Red represents negative c) aligned zoom-in of thecursor’s secondlocation image of the pair Notice the Notice theoutside newly of formed NW–SE aligned leads The thin redcross-correlation lines are rulers that highlight the mouse newly formed NW-SE aligned leads The thin red lines are rulers that highlight the mouse cursor’s location 155 160 165 170 175 2.4 Satellite image processing reprojecting and cropping satellite images using the 180 GDALused toolset Algorithms for operational satellite image analysis are often tuned to the specific instruments As the objective of loading the GeoTIFF images, translating 16-bit this study is to compare different instruments, no instrumentgreyscale values to floating-point numbers specific tuning was done The images still need georectification, typically a landmaskpyramid is used.for both images, using 185 and generating a resolution Forathis SAR images are rectified the level Mercator 2-d work, low-pass filter and decimating forto every projection with a reference latitude of 61◦ 40 This projection chosen runningascross-correlation for used coarse resolution image was it matches the one in both the nautical windows charts for this area and previous ice motion estimation work 190 for5.the Baltic Sea (Karvonen, There stillthe remains median-filtering the coarse2012) result to produce averagea slight motion error after It could fieldthis andprojection first guessstep for next stepbe corrected by matching static features between the images An incidence angle correction not performed was running cross-correlation forwas the finest resolution It image deemed unnecessary, as the method calculates normalized windows cross-correlations for small image windows No speckle fil- 195 saving this result and a median-filtered version (radius tering was applied 3) of it in an ASCII text file 2.5 Masking land points The results were analyzed and plotted using the Matlab and Octave programs For sea ice motion estimation in the narrow basins of the Baltic Sea, land points are for sometimes masked out before 2.3 Performance metrics motion estimation analysis (Karvonen, 2012) In this work, motion detection 200 was using unmasked images Result vectors for For performed this study, no ground truth data was available for comland and sea areas were then analysed separately As a drawparison It was necessary to define some performance metback, image windows that include generate ric that could be calculated fromthe thecoastline results alone In two this valid cross-correlation peaks Land points and shallow arwork, the cross-correlation method was not tuned for the imeas were distinguished by topographical data produced by 205 age types, and especially between C- and L-band images low the Leibniz Institute for Baltic Sea Research (Seifert et al., cross-correlation coefficients were expected Instead of the 2001) cross-correlation coefficient itself, we consider the ratio of were found suffer from ratio a spatially theThe twosatellite highestimages peaks While a hightopeak-to-peak is not varying registration error This was corrected using the finestThe Cryosphere, 9, 357–366, 2015 resolution motion estimates for land points These were interconclusive evidence of correctness, it is assumed bethe a necpolated in order to generate a seamless estimatetofor imessary requirement A motion vector is rejected if the marage registration error This registration error field was finally gin betweenfrom two the highest cross-correlation peaks is less substracted motion results recorded for the driftthan ice 15%, and otherwise accepted in a “peak margin” sense Additionally, each motion result is evaluated against the of uniformity, flagging errors all vectors that 3expectation Visual comparison between L-asand C-band images differ significantly from the median-filtered vector field It is assumed that L-band the median filtering at removing The PALSAR images havesucceeds been compared to spurious values SAR and retains stepwise the ice RADARSAT-1 by the real Canadian Icechanges Service.inThey remotion al., 1990), so that thesuperior median-filtered port thatfield the(Astola L-bandetimages contain a far amount motion represents the real average motion when of ridgefield information compared to C-band LargeEven ridges are this is not the case, vectors so clearly defined, and unrealistic detail remains wellwill intonot thematch springit,melt these cases produce that falsePALSAR successes.allows A motion vector season It iscannot also reported clearer deis rejectedbetween if it differs median of its allows neighborhood lineation icefrom floes.the PALSAR also thin iceby to more thandistinguished 500 meters Otherwise is considered acceptable be easily from thickitice, while C-band images in a “regularity” sense.thin ice with thicker ice types (Arkett could confuse rough Both2008) criteria are arbitrary However, they appear to be senet al., sible this6 study Aschoices images for and (see Table and Fig ) are separated by only 32 min, they are assumed to represent the same ice situation in C- and L-bands No ice-related change can be dis2.4 Satellite image processing tinguished visually, so all differences are taken to result from differences the imagingsatellite instruments a general Algorithms between used for operational imageAs analysis are difference, the L-band image (f) has more contrast within often tuned to the specific instruments As the objective the of sea also more easy to distinguish, while this area studyThe is tocoastline compareisdifferent instruments, no instrumentin the C-band the coastline disappears some, espespecific tuningimage, was done The images still needingeorectificacially northern, locations Below, specific differences in these tion, and typically a landmask is used twoFor images are evaluated in detail this work, SAR images are rectified to the Mercator ◦ To summarize, types inlatitude the drift appear simprojection with a ice reference ofice 61 region 40 This projecilarly in images of both frequency bands Sometimes the Ction was chosen, as it matches the one used in both the nauband image is better at distinguishing the edge of an ice floe, tical charts for this area, and previous ice motion estimation and theSea L-band shows features not visible in the worksometimes for the Baltic (Karvonen, 2012) There still remains C-band image (see east edge of Fig 9), but for most features, www.the-cryosphere.net/9/357/2015/ 325 ences are dark lines in the open water in the L-band image, and slightly better contrast in the C-band image However, these formations appear fragile and susceptible for changes, which makes tracking them rather demanding 330 SAR images for sea ice motion estimation J Lehtiranta et al.: Comparing C- and L-band SAR images for sea ice motion estimation 361 rrected by ed It was ormalized peckle fil- ns of the ut before detection ectors for As a drawnerate two hallow arduced by fert et al., 335 340 Figure Detail of landfast ice in northern Bay of Bothnia around 345 Figure Detail of landfast ice in northern Bay of Bothnia on 18 Detail of landfast ice in northern J Lehtiranta et al.: Comparing CL-band SAR images sea ice motion estimation Fig.and Detail of landfast ice on in for northern of Bothnia around Fig BaytoofTornio Bothnia 18 Hailuoto, offshore from Oulu, 18 MarchBay 2009 March 2009 White tracks are shipping lanes andon Kemi, Hailuoto, offshore from Oulu, on 18 March 2009 March 2009 White tracks are shipping lanes to Tornio and Kemi, which appear very bright in SAR images which appear very bright SAR images freezing period Here,intoo, early-season deformations could 295 spatially 260 the finest were inter- 300 or the imwas finally e drift ice 265 mages 305 mpared to They rer amount 270 ridges are pring melt 310 learer dehin ice to nd images 275 es (Arkett 315 separated the same hange can 280 ken to reents As a e contrast to distin- 320 appears in 285 fic differ- pear simmes the C- 325 n ice floe, 290 in the Ct features, r contrast be masked by smoothing surface processes The bright feature north image of Hailuoto appears similarcontrast in both the L-band simplyisland, seemswhich to provide stronger images, is probably a field of broken ice, often called a rubble On appear difOn the other hand, many features in landfast ice appear diffield, analogous to a very wide pressure ridge ferently relatively ferently in C- and L-band images Perhaps a long, relatively Comparing these it can produces be concluded that landpeaceful evolution of images, an ice surface roughpeaceful surface roughfast ice can be a tricky substance for matching windows ness wavelengths of ness in length scales comparable to the radar wavelengths SAR images of different bands Some features will appear similar but at different intensities, and some areas will look 3.1 Landfast ice 3.1 Landfast ice completely different Landfast ice isis immobile immobileand andnon-dynamic non-dynamicbyby definition Landfast ice definition It isIt is assumed that no recent deformation took place in the land3.2 Level ice assumed that no recent deformation took place in the landfast fast Discernible features are assumed be either old zone.zone Discernible features are assumed to betoeither old dedeformations or weather-related As can be seen in Fig 5, the Some ice classified as level iceAscan southwest formations or weather-related canbebeseen seenininthe figure 5, the archipelago looks more homogenous and dark in the L-band corner of figure southwest from Tornio in figure and in archipelago looks6,more homogenous and dark in the1,L-band image Conversely, the image shows large hazy feathe dark ovals in figure These areas showaa large up as hazy relatively image Conversely, the C-band C-band image shows feature, conspicuously framed by the shipping lanes dark areas, presumably because of relatively low specular reture, conspicuously framed by the shipping lanes flection, in SAR images of both wavelengths In general, CThe linear or web-like features visible in the L-band image The linear or web-like features visible in the L-band image band shows these features darker than L-band, as L-band will but missing from the C-band image are probably due to the but missing from the C-band image are probably due to the cause more scattering from beneathThe the level surface (Dierkgreater volume scattering in The surface scattering is greater volume scattering in L-band L-band surface scattering is ing and Busche, 2006) In some areas level ice is relatively weaker and less extended, perhaps due to snowfall or melt– weaker and less extended, perhaps due to snowfall or meltfeatureless freeze events freeze events.and in others rather detailed Some of the areas look identical in C- from and L-bands, others show more contrast Features the L-band image but visible on Features missing missing from L-band image but visible on the in L-band However, based on visual inspection, correlating the C-band image, on the other hand, are probably caused C-band image, on the other hand, are probably caused by image windows insmaller level ice seems feasible Thiswavelength analysis by surface roughness smaller than the L-band surface roughness than the L-band wavelength (23.6is limited by the small amount of level ice (23.6 cm) The shipping lanes that constrict the bright haze in cm) The shipping lanes that constrict the bright haze in the the C-band image, provide a hint of its formation This was C-band image, provide a hint of its formation This was pos3.3 mobile Open ice possibly mobile broken slush, which to form a rough sibly broken slush, which frozefroze to form a rough sursurface on the northern side of the shipping lanes face on the nothern side of the shipping lanes Sea areas less than 60% ice cover classified asband open Near the southwest corner, there is aaare brighter gray band Near thewith southwest corner, there’s brighter gray ice In open ice, separate ice floes drift freely among waves without clear features This is the shear zone at the landfast without clear features This is the shear zone at the landfast Using both frequency bands, ice forms by similar grayforces curls, ice boundary, experiencing deformation by external forces ice boundary, experiencing deformation external visible in figure 8, that should allow motion detection using but still attached to the landfast ice, islands, or the shallow but still attached to the landfast ice, islands, or the shallow cross-correlation to work well Most notable visible differsea floor The dark feature under it is open water or thin ice sea floor The dark feature under it is open water or thin ice ences are dark lines in the open water in the L-band image, in a lead, and we also see some drift ice in the corner of the in a lead, and we also see some drift ice in the corner of the and slightly contrast in the in C-band image However, image These better features look similar both images image these formations appear fragile and susceptible for features changes, In Fig 6, the L-band image has ill-defined features In figure 6, the L-band image has ill-defined bright which makes tracking them rather demanding in the landfast ice zone while the C-band shows little scatter- 330 in scattering To know the evolutionary history of these features, one ing features, one would need to track their formation from the beginning of the would beginning of the freezing period Here, too, early-season deformations could 335 www.the-cryosphere.net/9/357/2015/ Figure Elliptic dark area classified as level ice near Raahe on 18 Fig 7.2009 Elliptic dark area classified as level ice near Raahe on 18 March March 2009 be masked by smoothing surface processes The bright feature north of Hailuoto island, which appears similar in both images, is probably a field of broken ice, often called a rubble field, analogous to a very wide pressure ridge Comparing these images, it can be concluded that landfast ice can be a tricky substance for matching windows of SAR images of different bands Some features will appear similar but at different intensities, and some areas will look completely different 3.2 Level ice Some ice classified as level ice can be seen in the southwest corner of Fig 6, southwest from Tornio in Fig , and Fig Openovals ice between theThese Swedish coast and up theas compact ice in the8.dark in Fig areas show relatively pack in North Kvarken on 18 March 2009 dark areas, presumably because of relatively low specular reflection, in SAR images of both wavelengths In general, Cband shows these features darker than L-band, as L-band will cause more scattering from beneath the level surface (Dierk3.4 Compact drift ice ing and Busche, 2006) In some areas, level ice is relatively featureless and in others rather Some of the comareas Drift ice, classified in finnish ice detailed maps as consolidated, look identical in Cand L-bands, others show more contrast pact or very close ice, often covers the central Bay of Bothnia in L-band However, oncontinuum, visual inspection, correlating during winters It is abased mobile it deforms readily, image windows in level iceforces seemsover feasible This analysis is and transmits compressive large distances limited by the amount of levelpacked ice floes of compact In figure 9, small separate but closely drift ice can be seen, sometimes separated by leads or other open water features Many distinct ice floes are recognizable The Cryosphere, 357–366, in both images, but the fainter floes near the9,east edge are2015 not visible in the L-band image despite standing out very clearly in the true-color image The L-band image seems less able to distinguish the edge between a lead and a smooth ice floe 3.4 Com Drift ice, pact or ve during wi and transm In figur drift ice c open wate in both im visible in in the true to distingu Occasiona such as th the edge frequency motion es In figur is seen in features, L Fig Southern tip of the compact drift ice on the Bay of Bothnia on the 18 March 2009 Encircled the area of faint, barely distinguishable ice floes Fig Elliptic dark area classified as level ice near Raahe on 18 March 2009 southwest 1, and in relatively ecular reeneral, Cband will ce (Dierkrelatively the areas e contrast orrelating nalysis is 362 Fig 11 Le J Lehtiranta et al.: Comparing C- and L-band SAR images for sea ice motion estimation 370 375 J Lehtiranta et al.: Comparing C- coast and L-band SAR images for sea ice estimation Figure 10 motion Drift ice on the western Bay of Bothnia, 18 March 2009 Figure Open ice between the Swedish and the compact ice J Lehtiranta etbetween al.: on Comparing C-coast and and L-band SAR images for sea10.ice motion Fig Drift ice onestimation the western Bay of Bothnia, 18 March 2009.7 380 Fig Open ice theMarch Swedish the compact ice pack in North Kvarken 18 2009 pack in North Kvarken on 18 March 2009 d as open ng waves ray curls, ion using ble differnd image, However, changes, 3.4 Compact drift ice 330 335 340 345 nia around 350 Drift ice, classified in finnish ice maps as consolidated, compact or very close ice, often covers the central Bay of Bothnia during winters It is a mobile continuum, it deforms readily, and transmits compressive forces over large distances In figure 9, separate but closely packed floes of compact drift ice can be seen, sometimes separated by leads or other open water features Many distinct ice floes are recognizable in both images, but the fainter floes near the east edge are not 355 Fig Southern tiptip of of thethe compact drift iceice onon thethe Bay of of Bothnia Figure Southern compact drift Bay Bothvisible in the L-band image despite standing out very clearly Fig Southern tip2009 of the compact drift ice on the Bay of Bothnia on the 18.18 March Encircled the of barely distinnia on9.the March 2009 Encircled thearea area of faint, faint, in the image The L-band image seems less able on the true-color 18.ice March guishable floes.2009 Encircled the area of faint, barely distinto distinguish the guishable ice floes edge between a lead and a smooth ice floe Occasionally there is texture not present in the C-band image, such as the bright features in the southeast corner However, 360 3.3 Open ice the edge of open water is well visible and similar on both frequency bands, enough for Sea areas with lessand thanmost 60 %ice icefloes coverarearesimilar classified as open 370 motion estimation ice In open ice, separate ice floes drift freely among waves 370 In figure a compact andice mostly ice curls, pack Using both 10, frequency bands, formscontinuous similar gray 365 is seen in both Cand L-band Both images reveal the same visible in Fig 8, that should allow motion detection using features, L-band in cross-correlation tobetter work contrast well Most notable visible differ- ences are dark lines in the open water in the L-band image, 375 and slightly better contrast in the C-band image However, 375 these formations appear fragile and susceptible to changes, which makes tracking them rather demanding 3.4 Compact drift Fig 10 Drift ice on theice western Bay of Bothnia, 18 March 2009 Fig 10 Drift ice on the western Bay of Bothnia, 18 March 2009 350 350 355 355 380 380 Drift ice, classified in Finnish ice maps as consolidated, compact close ice,figures often covers Bay of Bothnia It or is very evident from 10 andthe 11central that sometimes leads It is winters evident from 10continuum, and In 11general thatitsometimes leads during ItinisL-band afigures mobile deforms readily appear very dark images however, leads appear veryindark inkinds L-band general and it transmits compressive forces In over largehowever, distances are visible both ofimages images, and should pose noleads spe385 are both kinds of images, should pose no drift speInvisible Fig 9,inseparate but closely packed of compact cial problem for motion estimation inand a floes mixed-frequency im- 385 cial problem for sometimes motion estimation in abymixed-frequency imice be seen, separated leads or other open agecan pair age pair water features Many distinct ice floes are recognizable in both images, but the fainter floes near the eastern edge are not Results discussion visible in theand L-band image despite standing out very clearly 390 Results and discussion in the true-colour Fig The L-band image seems less able 390 4.1 Motion estimates 4.1 Motion estimates To summarize, the9, motion estimates The Cryosphere, 357–366, 2015 calculated for image To thesame motion calculated pairssummarize, covering the timeestimates interval are similar infor all image cases pairs the same interval in all For acovering C-C or L-L bandtime image pair, are the similar matching is cases better 395 For C-C or L-L band image pair, matching is better and amotion results may be found for the a larger area than in a 395 and motion results may be found for a larger area than in a It is evident from figures 10 and 11 that sometimes leads appear very dark in L-band images In general however, leads are visible in both kinds of images, and should pose no spe385 cial problem for motion estimation in a mixed-frequency image pair Results and discussion 390 4.1 Motion estimates To summarize, thedrift motion estimates calculated image Figure 11 Leads in ice, Bay of Bothnia, 18 Marchfor 2009 Fig Leads inthe driftsame ice, Bay Bothnia,are 18.similar March 2009 pairs11 covering timeofinterval in all cases Fig LeadsorinL-L drift band ice, Bay of Bothnia, 18 matching March 2009 For 11 a C-C image pair, the is better 395 and motion results maybetween be found for aand larger area than in a to distinguish the edge a lead a smooth ice floe mixed pair Based on the metrics defined in chapter 2.3, an Occasionally there is texture not present in the C-band image, L-L image pair is superior for motion estimates compared to tion is well supported by the southwesterly winds that turned such as the bright features in the southeast corner However, tion is well supported by the southwesterly winds that turned C-C pairs, while mixed pairs are still feasible despite them north towards the end of period It is notable though, that the edge of open water is well visible and similar in both north towards the end of the motion period It are isthe notable though, presenting thepair most problematic case neither image produces for southern tip ofthat the frequency bands, and most ice floes similar enough for 400 neither image pair produces motion for the southern tip of the The average motion for the whole experiment period is drift ice area This is probably because the ice edge changed motion estimation drift ice area This is probably because the ice edge changed shown in figure 12 Both a C-C pair and a mixed L-C pair shape completely, and the numerous ice floes were too small In Fig 10, a compact and mostly continuous ice pack is shape completely, and the numerous ice floes were too small produce an acceptable result forBoth mostimages ofestimates the drift ice The moto be in distinguished These two parallel correspond seen both C- and L-band reveal the same to be distinguished These two parallel estimates correspond tion fields are almost identical, and the average eastward moto the first row of table 2.3 17.6 % of the motion vectors features, though L-band in better contrast to theisR1-A2 first row of table 17.6 of sometimes thecross-correlation motionleads vectors in It the image pair 2.3 had an acceptable evident from Figs 10 and 11%that apin the R1-A2 image pair had an acceptable cross-correlation peak very margin, and % ofimages the vectors were close to the leads local pear dark in14.0 L-band In general however, peak margin, and 14.0 % of vectors were close to the median Forinthe concurrent image pair R1-E2, both C-band, are visible both kinds ofthe images, and should pose nolocal spemedian For the concurrent image pair R1-E2, both C-band, an additional %motion of the estimation motion vectors passed both criteria cial problem for in a mixed-frequency iman additional 2% thean motion vectors passedmovement both criteria figure 13, weofsee average southward for ageIn pair figure3613,hours we see southward movement for theInlatter of an theaverage experiment This is in line with the 36 hours experiment This is in line with the latter prevailing windsofasthe well, as the northward transport of the prevailing winds as well, as the northward transport of ice had stopped before the winds turned north This time, had Results and discussion ice stopped before turned north.is This time, for the C-band pair, alsothe thewinds southern ice edge successful for the C-band pair, also the southern ice edge is successful but 13a shows no motion where 13b finds realistic vectors 4.1 13a Motion estimates but shows no motion where 13b finds realistic These two parallel estimates correspond to the secondvectors row of These two parallel estimates correspond to the second row of table 2.3 Again, the C-band pair produces more acceptable To summarize, the motion estimates calculated for image table 2.3.some Again, C-band produces acceptable vectors, of the which mustpair be located in more the southern ice pairs covering the same time interval are similar in all cases vectors, some of which must locatedtime in the edge, less deformed during thebeshorter spansouthern coveredice by For a C–C or L–L band image pair, the matching is better and edge, less deformed these image pairs during the shorter time span covered by motion results may be found for a larger area than in a mixed these Theimage four pairs latter motion estimates, represented on the two pair Based on the motion metrics estimates, defined in represented Sect 2.3, an on L–L image The four the two bottom rowslatter of table 2.3, appear very much like 13b This is pair is superior for motion estimates compared to C–C pairs, bottom rows of table 2.3, appear very much like 13b This is because each of these image pairs cover the whole period of because of these image pairs cover the whole period of northerlyeach winds northerly winds Comparing the performance of parallel image pairs, some www.the-cryosphere.net/9/357/2015/ Comparingwere the performance of parallel some observations made As expected, the image motionpairs, estimation observations werebetter made.for Asshorter expected, the motion algorithm works timescales, as estimation less deforalgorithm bettertofor shorterFor timescales, less deformation hasworks had time happen all imageaspairs, largemation has had time to happen For all image pairs, large- tion is we north tow neither im drift ice a shape com to be dist to the firs in the R1 peak marg median F an additio In figu the latter the preva ice had s for the Cbut 13a s These two table 2.3 vectors, s edge, less these ima The fo bottom ro because e northerly Compa observatio algorithm mation scale mot contained dius me smooth m gorithm w correct T ing the la a mixed i J Lehtiranta et al.: Comparing C- and L-band SAR images for sea ice motion estimation J Lehtiranta et al.: Comparing C- and L-band SAR images for sea ice motion estimation J Lehtiranta et al.: Comparing C- and L-band SAR images for sea ice motion estimation 363 66oN 66oN 66oN 66oN 30’ 30’ 30’ 30’ 65oN 65oN 65oN 65oN 30’ 30’ 30’ 30’ 64oN 64oN 64oN 64oN 30’ 30’ 30’ 63oN o 63oN 30’ 10 km 63oN o 20 E 10 km 22 E 20oE 22oE o 24 E o 26 E 63oN 24oE 26oE 10 km 20oE 10 km 22oE 24oE 26oE 20oE 22oE 24oE 26oE Fig 12 a) motion vectors from combining images and 6, of C- and L-band, respectively b) motion vectors from combining images and Figure 12 (a) Motion vectors from combining images and 6, of C- and L-band, respectively (b) Motion vectors from combining images 5, both C-band anda) 5, motion both C-band Fig.1 12 vectors from combining images and 6, of C- and L-band, respectively b) motion vectors from combining images and 5, both C-band 66oN 66oN 66oN 66oN 30’ 30’ 30’ 30’ 65oN 65oN 65oN 65oN 30’ 30’ 30’ 30’ 64oN 64oN 64oN 64oN 30’ 30’ 30’ 63oN o 63oN 30’ 10 km 20 E 10 km 63oN o 22 E o 24 E o 26 E 63oN 10 km 20oE 10 km 22oE 24oE 26oE o 20oE 22oE 24oE and 6, of 26oCE and L-band, respectively 20oE E 24oE from combining 26oE Figure 13 (a) Motion vectors from combining images (b)22Motion vectors images C-band Fig.and 13.5,a)both motion vectors from combining images and 6, of C- and L-band, respectively b) motion vectors from combining images and 5, both C-band Fig 13 a) motion vectors from combining images and 6, of C- and L-band, respectively b) motion vectors from combining images and 5, both C-band while mixed pairs are still feasible despite them presenting motion vectors in the R1–A2 image pair, 17.6 % had an ac- 405 405 410 410 415 415 the most problematic case Same-band image pairs (C-C, L-L) are found better than The average motion for the whole experiment period mixed-band (C-L) pairs Further, the L-band is found moreis Same-band image pairs (C-C,pair L-L) are foundL–C better than shown in Fig 12 Both a C–C and a mixed pair prosuitable for motion estimation in this data set than C-band mixed-band (C-L) pairs Further, the L-band is found more duce an acceptable result for most of the drift ice The moUnfortunately, it seems that a large peak margin in crosssuitable for motion estimation inand this data set than C-band tion fields are almost identical, the average eastward mocorrelation is not sufficient as an indicator of correctness Unfortunately, it seemsby that asouthwesterly large peak margin in crosstion motion is well supported thefound winds that turned Many vectors were to be nonsensical even 420 correlation is notthe sufficient as an indicator of correctness north towards end of the period It is notable though, when they were produced by a unique cross-correlation peak Many vectors were foundmotion to be for nonsensical even that motion neither image pair produces the southern tip This can happen e.g when the ice surface pattern is lost be- 420 when theydrift wereiceproduced by is a unique cross-correlation of the area This probably because the icepeak edge tween images In closer investigations it was found that a mochanged shape e.g completely, numerous ice is floes This can happen when theand icethe surface pattern lostwere betion estimate using the highest peak is often correct even if too images small toInbecloser distinguished Theseit two estimates tween investigations was parallel found that a mothecorrespond second-highest is just lower.in Table Of the to thepeak R1-A2 andbarely R1-E2 tion estimate using the highest peak isrows often correct even if 425 the second-highest peak is just barely lower 425 www.the-cryosphere.net/9/357/2015/ ceptable cross-correlation peak ofmargin, 14.0 % of the 4.2 Statistical performance image and pairs vectors were close to the local median For the concurrent 4.2 performance pairs imageStatistical pair R1–E2, both C-band, of animage additional % of the moOverall, both Cand L-band image pairs and mixed image tion vectors passed both criteria pairs similar statistical properties in themovement motion results In show Fig.both 13, we and see an average southward for Overall, CL-band image pairs and mixed image The maximal normalized cross-correlation coefficient the latter 36 h of the experiment This is in line with the pairs show similar statistical properties in the motion results found is mostly between 0.6, with some matches prevailing winds as well, as0.2theand northward transport of ice The maximal normalized cross-correlation coefficient reaching up to 0.95 As can be seen in figure 14, for C-band had stopped before the winds turned north This time, for found is mostly between 0.2 and 0.6, with some matches pairs the worst is around This is closer to 0.4 in the C-band pair,match the southern ice0.2 edge is also successful butthe reaching up to 0.95 As can be seen in figure 14, for C-band L-band of figure 15, which higher correlation Fig 13apair shows no motion wherehas Fig.overall 13b finds realistic vecpairs the worst match is around 0.2 This is closer to 0.4 in the tors These two parallel estimates correspond to the E1-A2 coefficients L-band pair of figure 15, which has overall higher correlation and E1-E2 rows in Table Again, the C-band pairmost produces The ice conditions and2 their change are the imporcoefficients tant factors of success This is evident from 15b The A1-E2 The ice conditions and their change are the most important factors of success This is evident from 15b The A1-E2 The Cryosphere, 9, 357–366, 2015 364 J Lehtiranta et al.: Comparing C- and L-band SAR images for sea ice motion estimation 08 08 Hi s t ogr am of peak c r os s - c or r el at i on c oef f i c i ent Hi s t ogr am of peak c r oss- cor r el at i on coef f i c i ent 1 07 07 08 08 05 05 r ac i on f r fac t i ton r ac i on f r fac t i ton 06 06 04 04 03 03 02 02 06 06 04 04 02 02 01 01 00 Hi s t ogr am of peak c r os s - c or r el at i on c oef f i c i ent Hi s t ogr am of peak c r oss - c or r el at i on coef f i c i ent 2c r os s - c or 0.r4el at i on 0.c oef f i c i ent c r oss- cor r el at i on c oef f i ci ent 00 1 2c r os s - c or 0.r 4el at i on f i c i ent c oef c r os s- cor r el at i on c oef f i ci ent 1 Figure 14 Maximum cross-correlation for matched windows in the R2–A2 image (C–L, R2–E2 image (C–C, right) Fig 14 Maximum cross-correlation for matched windows in the R2-A2 image pairpair (C-L, left)left) andand thethe R2-E2 image pairpair (C-C, right) Fig 14 Maximum cross-correlation for matched windows in the R2-A2 image pair (C-L, left) and the R2-E2 image pair (C-C, right) Hi s t ogr am of peak c r os s - c or r el at i on c oef f i c i ent Hi st ogr am of peak c r os s- cor r el at i on coef f i c i ent 08 08 07 07 07 07 06 06 06 06 0 0 0 0 0 05 05 04 04 03 03 02 02 01 01 00 05 05 04 04 03 03 02 02 01 01 00 r ac i on f r fac t i ton r ac i on f r fac t i ton 08 08 2c r oss - cor 0.r4el at i on 0.c oef f i c i ent cr os s- cor r el at i on c oef f i ci ent 1 Hi s t ogr am of peak c r os s - c or r el at i on c oef f i c i ent Hi s t ogr am of peak c r os s- cor r el at i on c oef f i c i ent 2c r os s - c or 0.r 4el at i on f i c i ent c oef c r oss - c or r el at i on coef f i c i ent 1 Figure 15 Maximum cross-correlation coefficient histogram for the A1–A2 image pair (L–L), left, and the A1–E2 image pair (L–C), right Fig 15 Maximum cross-correlation coefficient histogram for the A1-A2 image pair (L-L), left, and the A1-E2 image pair (L-C), right Fig 15 Maximum cross-correlation coefficient histogram for the A1-A2 image pair (L-L), left, and the A1-E2 image pair (L-C), right more acceptable vectors, some of which must be located in the southern ice edge, less deformed during the shorter time span covered by these image pairs The four latter motion estimates, represented on the two bottom rows of Table 2.3, appear very much like Fig 13b This is because each of these image pairs cover the whole period of northerly winds Comparing the performance of parallel image pairs, some observations were made As expected, the motion estimation algorithm works better for shorter timescales, as less deformation has had time to occur For all image pairs, large-scale motion estimation was successful All motion estimates contained a large number of spurious vectors too, but a radius median filtering was found to produce a realistic and smooth motion field Due to the median filtering, the algorithm works even if only 10–20 % of motion vectors are correct This success rate is thus found sufficient for detecting the large-scale motion However, as evident in Fig 13, a mixed image pair can fail in details in some sub-regions Same-band image pairs (C–C, L–L) are found better than mixed-band (C–L) pairs Further, the L-band is found more The Cryosphere, 9, 357–366, 2015 suitable for motion estimation in this data set than C-band Unfortunately, it seems that a large peak margin in crosscorrelation is not sufficient as an indicator of correctness Many motion vectors were found to be nonsensical even when they were produced by a unique cross-correlation peak This can happen, for example, when the ice surface pattern is lost between images Upon closer investigation, it was found that a motion estimate using the highest peak is often correct even if the second-highest peak is just barely lower 4.2 Statistical performance of image pairs Overall, both C- and L-band image pairs and mixed image pairs show similar statistical properties in the motion results For most image windows, the highest found normalized cross-correlation coefficient was between 0.2 and 0.6 The best matches had a cross-correlation coefficient up to 0.95 As can be seen in Fig 14, for C-band pairs the worst match is around 0.2 This is closer to 0.4 in the L-band pair of Fig 15, which has overall higher correlation coefficients www.the-cryosphere.net/9/357/2015/ J Lehtiranta et al.: Comparing C- and L-band SAR images for sea ice motion estimation 365 Fig 16 Geographical distributions of errors, (a) pair R2-A2 (CL), (b) R2-E2 (CC), (c) A1-A2 (LL) and (d) A1-E2 (LC) Figure 16 Geographical distributions of errors, (a) pair R2–A2 (C–L), (b) R2–E2 (C–C), (c) A1–A2 (L–L) and (d) A1–E2 (L–C) Table Performance values for parallel image pairs, as the percentage of motion vectors that are accepted based on the peak margin -criterion (pm-good) and regularity-criterion (reg-good), both defined in Sect 2.3 Image pair pm-good reg-good R1–A2 (C–L) E1–A2 (C–L) R2–A2 (C–L) A1–A2 (L–L) 17.6 % 20.1 % 24.7 % 45.6 % 14.0 % 14.2 % 15.8 % 28.4 % R1–E2 (C–C) E1–E2 (C–C) R2–E2 (C–C) A1–E2 (L–C) 19.6 % 22.7 % 27.9 % 30.7 % 16.2 % 16.7 % 18.6 % 18.7 % The ice conditions and their change are the most important factors of success This is evident from Fig 15b The A1–E2 image pair boasts large cross-correlation coefficients despite mixing two different wavelengths The histograms for motion estimation error magnitude, as estimated by the difference in metres between each motion vector and the local median, are all rather similar The histograms of error show a strong peak for no or very small error and a distribution characteristic of this problem This www.the-cryosphere.net/9/357/2015/ distribution roughly corresponds to the idealized theoretical distribution of the distance of a random point This distribution arises from the fact that the search window is square and it allows at most 40 pixels of movement in each dimension It is concluded that there are no systematic errors in the motion estimation algorithm Considering the margin between the two highest correlation peaks, it was found that a C–C pair is better than a mixed C–L pair at finding unique peaks The difference is small though, and very often the highest cross-correlation peak stands only slightly above the second contender This was expected, as the maximal cross-correlation (MCC) method is known to often produce multiple cross-correlation peaks for noisy signals To improve performance, the algorithm should consider multiple cross-correlation peaks, not just the highest one 4.3 Geographical distribution of errors The geographical distribution of errors was calculated for the test cases with smallest time difference in order to evaluate problems stemming from local effects and not changes that occur over longer time intervals Figures 16a and 16b correspond to the same time interval and show that a C–C pair is stronger than a C–L pair in all localities, but the mixed-band The Cryosphere, 9, 357–366, 2015 366 J Lehtiranta et al.: Comparing C- and L-band SAR images for sea ice motion estimation pair also succeeds to some extent everywhere the C–C pair does Figure 16c and d correspond to another time interval and shows that an L–L pair is much better than a mixed pair, again without any clear difference in the areas of successful motion estimation To summarize, all image combinations have trouble with the northwesterly lead opening near the northeast edge of landfast ice, and all combinations behave better in the central ice pack It is clear that a single-frequency pair is desirable, but also that for most regions, a mixed-frequency pair performs reasonably well No image pair finds more than an occasional good motion vector in open ice of less than 30 % coverage It seems that the C-band is better than L-band for matching image patterns on land While this is of no concern for perfectly georeferenced images, this might mean that georectifying L–L image pairs might be more problematic Conclusions We show that it is possible to calculate sea ice motion using an L-band SAR image together with a C-band image The program written for this purpose works and produces convincing results, so the chosen algorithm of maximal crosscorrelation suits this purpose L-band images are fundamentally different than C-band images as the ratio of surface and volume scattering is different and some C-band scatterers are invisible to L-band radar This difference manifests itself primarily in landfast ice, possibly because long periods of thermodynamical changes create different surface features near the length scales of the employed wavelengths Fortunately, the motion estimation largely succeeds for landfast ice, and most features in drift ice appear much easier targets for motion detection The different frequency bands complement each other when plentiful data is available, but they are somewhat poorer for backup purposes as each band has distinct strengths and weaknesses On C-band, ice floe edges appear in a more reliable manner, while the L-band distinguishes the coastline better and generally shows more features and better contrast For motion estimation, a pair of two L-band SAR images is found to be desirable among the compared options A pair of two C-band images also performs well, and a mixed pair performs adequately The introduction of L-band SAR instruments can thus present both more reliable motion estimates by using L–L pairs and better time resolution, albeit at a cost of increased uncertainty, by using mixed L–C pairs This work provides a new tool for motion estimation It also provides insights into the usage of L-band SAR images, both alone and in combination with C-band images Thus, it is good preparation for the future launch of the ALOS2 satellite and for the handling of its L-band images, and utilizing the GPGPU computational framework was both a strength in this work and a valuable lesson for the future The Cryosphere, 9, 357–366, 2015 Acknowledgements This study was supported by the Finnish Meteorological Institute and the Polar View project The authors wish to thank Eero Rinne, Lars Kaleschke and Lang Wenhui for their detailed and insightful comments on the original manuscript Edited by: L Kaleschke References Arkett, M., Flett, D., De Abreu, R., Clemente-Colon, P., Woods, J., and Melchior, B.: Evaluating ALOS-PALSAR for Ice Monitoring-What Can L-band for the North American Ice Service?, in: Geoscience and Remote Sensing Symposium, IGARSS 2008, IEEE International, 5, V–188, 2008 Astola, J., Haavisto, P., and Neuvo, Y.: Vector median filters, P IEEE, 78, 678–689, 1990 The Baltic Sea Portal: Ice winter 2008–2009, available at: http: //www.itameriportaali.fi/en/tietoa/jaa/jaatalvi/en_GB/2009/ (last access: 18 December 2014), 2009 Dierking, W and Busche, T.: Sea Ice Monitoring by L-Band SAR: An Assessment Based on Literature and Comparisons of JERS1 and ERS-1 Imagery, IEEE T Geosci Remote, 44, 957–970, 2006 Emery, W., Fowler, C., Hawkins, J., and Preller, R.: Fram Strait satellite image-derived ice motions, J Geophys Res., 96, 4751– 4768, 1991 Fily, M and Rothrock, D.: Sea ice tracking by nested correlations, IEEE T Geosci Remote, GE-25, 570–580, 1987 Karvonen, J.: Operational SAR-based sea ice drift monitoring over the Baltic Sea, Ocean Sci., 8, 473–483, doi:10.5194/os-8-4732012, 2012 Karvonen, J., Similä, M., and Lehtiranta, J.: Sar-based estimation of the baltic sea ice motion, in: Geoscience and Remote Sensing Symposium, IGARSS 2007, IEEE International, 2605–2608, 2007 Liu, A K., Martin, S., and Kwok, R.: Tracking of Ice Edges and Ice Floes by Wavelet Analysis of SAR Images, J Atmos Ocean Tech., 14, 1187–1198, 1997 Myrberg, K., Leppäranta, M., and Kuosa, H.: Itämeren fysiikka, tila ja tulevaisuus, Palmenia, Helsinki University Press, 2006 Seifert, T., Tauber, F., and Kayser, B.: A high resolution spherical grid topography of the Baltic Sea, 2nd Edn., Baltic Sea Science Congress, Stockholm, 25–29 November 2001, Poster# 147, available at: www.io-warnemuende.de/iowtopo (last access: 18 December 2014), 2001 Stone, J E., Gohara, D., and Shi, G.: OpenCL: A parallel programming standard for heterogeneous computing systems, Comput Sci Eng., 12, 66–72, 2010 Thomas, M., Kambhamettu, C., and Geiger, C.: Motion Tracking of Discontinuous Sea Ice, IEEE T Geosci Remote, 49, 5064–5079, 2011 Vesecky, J F., Samadani, R., Smith, M P., Daida, J M., and Bracewell, R N.: Observation of sea-ice dynamics using synthetic aperture radar images: Automated analysis, IEEE T Geosci Remote, 26, 38–48, 1988 www.the-cryosphere.net/9/357/2015/ Copyright of Cryosphere is the property of Copernicus Gesellschaft mbH and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission However, users may print, download, or email articles for individual use