Large-scale sea ice motion from Sentinel-1 and the RADARSAT Constellation Mission

As Arctic sea ice extent continues to decline, remote sensing observations are becoming even more vital for the monitoring and understanding of sea ice. Recently, the sea ice community has entered a new era of synthetic aperture radar (SAR) satellites operating at C-band with the launch of Sentinel-1A in 2014, Sentinel-1B in 2016 and the RADARSAT Constellation Mission (RCM) in 2019. These missions represent 5 spaceborne SAR sensors, that together routinely cover the 10 pan-Arctic sea ice domain. Here, we utilized over 60,000 SAR images from Sentinel-1AB (S1) and RCM to generate largescale sea ice motion (SIM) estimates over the pan-Arctic domain from March to December, 2020. On average, 4.5 million SIM vectors from S1 and RCM were automatically detected per week for 2020 and when combined (S1+RCM) they facilitated the generation of 7-day, 25 km SIM products across the pan-Arctic domain. S1+RCM SIM provided more coverage in Hudson Bay, Davis Strait, Beaufort Sea, Bering Sea, and over the North Pole compared to SIM from S1 alone. S1+RCM SIM was able 15 to be resolved within the narrow channels and inlets across the pan-Arctic alleviating the main limitation of coarser resolution sensors. S1+RCM SIM provided larger ice speeds with a mean difference (MD) of 1.3 km/day compared to the National Snow and Ice Data Center (NSIDC) SIM product and a MD of 0.76 km/day compared to Ocean and Sea Ice-Satellite Application Facility (OSI-SAF) SIM product. S1+RCM was also able to better resolve SIM in the marginal ice zone compared to the NSIDC and OSA-SAF SIM products. Overall, our results demonstrate that combining SIM from multiple spaceborne SAR 20 satellites allows for large-scale SIM to be routinely generated across the pan-Arctic domain.


Introduction
As Arctic sea ice extent continues to decline in concert with increases in carbon dioxide (CO2) emissions (Notz and Stroeve, 2016), remote sensing observations are becoming even more vital for the monitoring and understanding of Arctic sea ice. Recently, the sea ice community has entered a new era of synthetic aperture radar (SAR) satellites operating at C-band 25 (wavelength,  = 5.5 cm) with the launch of Sentinel-1A in 2014, Sentinel-1B in 2016Tores et al., 2012) and the RADARSAT Constellation Mission (RCM) in 2019 (Thompson, 2015). Together these missions represent 5 spaceborne SAR sensors that when combined offer the opportunity to retrieve large-scale sea ice geophysical variables with high spatiotemporal resolution. Small et al. (2021) demonstrated that combining SAR images from S1 and RADARSAT-2 allowed for the https://doi.org/10.5194/tc-2021-223 Preprint. Discussion started: 22 July 2021 c Author(s) 2021. CC BY 4.0 License. production of high spatiotemporal resolution analysis-ready composite products for large regions. Howell et al. (2019) used 30 analysis-ready composite products generated from S1 and RADARSAT-2 based on the approach described by Small et al., (2021) to provide high spatial resolution estimates of melt onset over a large region in the northern Canadian Arctic.
An important sea ice geophysical variable that could also benefit from large-scale SAR estimates across the Arctic is sea ice motion (SIM). SIM is controlled by the exchange of momentum due to turbulent process primarily from atmospheric and oceanic forcing. Away from the coast, winds explain 70% or more of the variance in Arctic sea ice motion (Thorndike and 35 Colony, 1982) and as a result, monitoring changes in SIM is important for understanding how sea ice responds to changes in atmospheric circulation (Rigor et al. 2002). SIM convergent and divergent processes impact the overall thickness of Arctic sea ice (Kwok, 2015) and the dynamic component of the Arctic sea ice area and volume balance is also impacted by SIM (Kwok, 2004;Kwok, 2009). The long-term record of SIM in the Arctic indicates the ice speed is increasing which are associated with thinner ice being more susceptible to wind forcing (Rampal et al., 2007;Kwok et al., 2013;Moore et al., 2019). 40 Techniques for estimating SIM from satellite observations have a long history dating back to the late 1980s and early 1990s that are primarily based on the maximum cross-correlation coefficient between overlapping images (e.g. Fily and Rothrock, 1987;Kwok et al., 1990;Emery et al., 1991). The maximum cross-correlation approach to estimate SIM can be applied to virtually any overlapping pair of satellite imagery separated by a relatively short time interval of ~1-3 days. For large-scale SIM, passive microwave imagery is typically the most widely used because of its large swath and daily coverage 45 (e.g. Agnew et al., 1997;Kwok et al., 1998;Lavergne et al., 2010;Tschudi et al., 2020). Enhanced resolution SIM products with spatial resolutions of ~2 km have also been generated (e.g. Haarpaintner, 2006;Agnew et al., 2008) although they have not been widely utilized. The trade-off with respect to large-scale SIM estimated from passive microwave imagery, however, is a low spatial resolution (12-25 km). As a result, SIM is more difficult to track with lower spatial resolution passive microwave sensors (Kwok et al., 1998) especially, within narrow channels and inlets (e.g. the Canadian Arctic Archipelago; 50 CAA) compared to SAR. However, SIM estimates from SAR are typically regionally based because of image availability across the Arctic.
In this study we make use of 5 SAR satellites from the S1 and RCM missions to generate SIM over the large-scale pan-Arctic domain (Fig. 1). To our knowledge this is perhaps the first time such an extensive combining of SAR imagery at the pan-Arctic scale has been undertaken to generate SIM. We first describe our workflow that estimates SIM from S1 and 55 RCM SAR imagery (hereafter, S1+RCM) in close to near-real time and combines the output into a S1+RCM SIM product.
We then discuss the vector quality and SIM uncertainty of large-scale S1+RCM SIM over the annual cycle for the Arctic and its sub-regions. Finally, we compare our S1+RCM SIM estimates to the existing SIM datasets from the National Snow and Ice Data Center (NSIDC) (Tschudi et al., 2020) and Ocean and Sea Ice-Satellite Application Facility (OSI-SAF) SIM (OSI-SAF) (Lavergne et al., 2010).

Data
The primary datasets used in this analysis were Extra Wide Swath imagery at HH polarization from S1 and ScanSAR 50 m (SC50M), ScanSAR 100 m (SC100M), and ScanSAR Low Noise (SCLN) at HH polarization from RCM from March to December 2020 (Table 1). With the recent availability of RCM imagery in 2020, this provided an additional 29,744 images 70 that when combined with S1 (~32,810 images) resulted in 60,000+ images being available to generate SIM across the pan-Arctic for 2020.
We also made use of the 7-day SIM NSIDC Polar Pathfinder dataset and the 2-day OSI-SAF sea ice motion dataset from March to December 2020. Tschudi et al., (2020) provides a complete description of the NSIDC Polar Pathfinder SIM dataset, and Lavergne et al., (2010) provides a complete description of the OSI-SAF SIM dataset. Finally, we used the 2020 75 daily pan-Arctic ice charts from the National Ice Center.

Automated sea ice motion tracking algorithm
We make use of the automated SIM tracking algorithm developed by Komarov and Barber (2014) to estimate largescale SIM across the pan-Arctic domain. A full description is provided by Komarov and Barber (2014), but the main 85 components of the algorithm are briefly described here. To begin with, coarser spatial resolution levels are generated from the original spatial resolution of the SAR image pairs. For example, if the original SAR image pairs have a spatial resolution of 200 m then the additional generated levels would be 400 m and 800 m. A set of control points (i.e., ice features) is automatically generated for each resolution level based on the SAR image local variances. To highlight edges and heterogeneities at each resolution level, a Gaussian filter and the Laplace operator are applied sequentially. Beginning with the lowest resolution level, 90 ice feature matches in the image pairs are identified by combining the phase-correlation and cross-correlation matching techniques that allows for both the translation and rotational components of SIM to be identified. SIM vectors not presented in both forward and backwards image registration passes are filtered out, as well as vectors with low cross-correlation coefficients. In order to refine the SIM vectors, at each consecutive resolution level the algorithm is guided by SIM vectors identified at the previous resolution. An example of the SIM output generated from the algorithm based on two overlapping 95 SAR image is shown in Fig. 2.
The algorithm has been widely utilized for applications that require robust estimates of SIM in Arctic using SAR at C-band (e.g. Howell et al., 2013;Howell et al., 2018;Komarov and Buehner, 2019;Moore et al., 2021). Moreover, when the SIM vector outputs were collocated and coincided with ice buoy trajectories, it was found to have a root-mean square error (RMSE) of 0.43 km. The limitations that are widely known with respect to estimating SIM from SAR imagery include regions 100 of low ice concentration, melt water on the surface of the sea ice and longer time separation between images also apply to this algorithm.

Generating large-scale gridded sea ice motion
The generalized processing chain for generating large-scale S1+RCM SIM is illustrated in Fig. 3. The approach 110 processes the S1 and RCM image streams separately and then combines the outputs into a S1+RCM SIM product. This parallel approach was chosen because SAR imagery is received by Environment and Climate Change Canada (ECCC) from S1 and RCM in close to near-real time and in order to "keep-up" with the 100's of images coming in per day and the subsequent computational load on automated SIM detection, the processing system is run every hour. As a result, S1 and RCM imagery are currently not mixed together for automated SIM tracking given the different orbit characteristics of the satellites which 115 contribute to differences in terms of when images are acquired compared to when they are received by ECCC. For example, if an S1 image acquired at 1300h UTC is transferred to our system sooner than an RCM image that was acquired at 1100h UTC then the RCM image would be missed.
For both S1 and RCM images streams, the pre-processing steps shown in Fig. 3 first involve calibrating the imagery to the backscatter coefficient of sigma nought (°) using the HH-polarization channel and map-projected to the NSIDC North 120 Pole Stereographic WGS-84, EPSG:3413 coordinate system with a 200 m pixel size. For S1 imagery, pre-processing steps were applied using the Graph Processing Tool (GPT) of the Sentinel Application Platform (SNAP) software, and for RCM imagery, the pre-processing workflow was applied using an in-house pre-processor.
Manual inspection of SAR imagery and subsequent image stack compilation prior to automatic SIM generation, while effective in regional-scale studies (e.g. Howell et al., 2013;Howell et al., 2016;Moore et al., 2021) is not practical for 125 generating large-scale SIM. The main challenges of estimating large-scale SIM across the pan-Arctic domain are (i) handling the large volume and delivery frequency of the imagery, (ii) efficiently selecting image stacks, and (iii) providing more computationally efficient feature tracking from the image stacks.
To address (i) and (ii), an automated approach for determining the suitability of images for inclusion in the automatic SIM generation (i.e. S1 or RCM image stack selection) was developed and depicted in Fig. 4. For the image stack selection, a 130 400 km x 400 km grid of sectors encompassing the pan-Arctic was generated and used to create overlapping stacks of SAR image pairs (Fig.4). The footprint geometry of each SAR image was compared to a given sector's extent and if the overlap <= 30% was achieved, that image was retained for feature tracking. Next, we assess image-to-image overlap within each sector to create a temporally sequential image stack that intersected one another to an acceptable degree (>=32,000 km 2 ). Images within each sector with an overlap of at 32,000 km 2 were retained for feature tracking. The stacks are created every hour using 135 the last-processed image from the previous run and the accumulated new imagery that had arrived in the time preceding.
For (iii), a more computationally efficient application of automatic SIM tracking algorithm (i.e. image stack processing) was developed. Traditionally, image stacks were processed serially which was effective for local-scale studies with limited amounts of imagery, but with significant increases in the SAR image data volume and study area domain size from S1 and RCM, it was necessary to enhance the processing speed of ice feature tracking analysis. The concurrent approach 140 as outlined in Fig. 5 takes advantage of vertical scalability by increasing the number of processes during image pair analysis.
This approach allowed for an entire image stack to be efficiently processed with as many computational cores as were available. For example, when three sets of image pairs are processing and process 3 finishes before process 2, process 3 picks up the next sequence of pairs instead of waiting for process 2 (Fig. 5). After stack processing, the last-processed image for the given sector is recorded in a database and processing ends. 145   The final step involved combining the results of the automatic S1 and RCM SIM tracking process into defined spatial 165 and temporal resolution grids to be used for analysis and mapping (Fig. 3). Combining the SIM output from S1 with RCM (i.e., S1+RCM) facilitated the ability to improve the spatial coverage and image density of SAR generated SIM across the Arctic. For example, the spatial distribution of SAR image density per week in 2020 for S1, RCM and S1+RCM is shown in Fig. 6. S1 had a denser coverage compared to RCM for the majority of the Arctic regions and especially the Central Arctic and Greenland Sea. However, RCM coverage was more widely spread across the Arctic compared to S1 and extended into to 170 the Bering Sea, Labrador Sea, and over the North Pole thus filling a gap typically associated with the majority of satellite sensors. An example of the ability of RCM to almost completely cover the North Pole on a single day is shown in Fig. 7. In addition, S1+RCM image density increases with latitude ( Fig. 6) indicating that more consistent coverage of the ice pack will be possible during the melt season which is beneficial considering this is when automated SIM tracking algorithms have more difficulty. Clearly, SAR image density from S1+RCM from March to December 2020 was significant ( Fig. 6) with almost 175 complete coverage every 3-days. However, despite the high image density coverage from S1+RCM, more consistent pan-Arctic SIM coverage can be achieved over 7-days because there are more image overlaps during a longer time span. As a result, a 25 km spatial resolution with a temporal resolution of 7-day was selected to provide the most consistent S1+RCM SIM coverage across the pan-Arctic domain. It should be noted that based on Fig. 6 regional S1+RCM SIM products at higher spatial and temporal resolution are certainly achievable given the image density S1+RCM, and as a result we also briefly 180 demonstrate this capability in the results section. than 75 km/day. In order to control for SIM speed heterogeneity within a 25 km grid cell, the median SIM was used to represent the ice speed for each grid cell. For each grid cell, a series of descriptive statistics were calculated that included the number of S1 and RCM SIM vectors, the median SIM, the standard deviation of SIM, and the mean cross-correlation coefficient. Even after removing anomalously large SIM speeds, the automatic SIM tracking algorithm sometimes detected obviously erroneous 190 SIM vectors far from the marginal ice zone and/or near the coast in sufficient quantity (i.e. 5+) to meet the grid cell criteria.
These grid cells were subsequently filtered out using a threshold distance of 150 km from the marginal ice zone (i.e., ice concentration of at least 18%) using the weekly National Ice Center ice charts.

Quantifying vector quality and uncertainty 205
The automated SIM tracking algorithm utilized in this study has undergone extensive validation against buoy positions and has an uncertainty of 0.43 km derived for SAR image pairs separated by 1-3 days (Komarov and Barber, 2014).
Moreover, SIM output from the tracking algorithm has been found to be in good agreement with other tracking algorithms that includes the RADARSAT Geophysical Processor (e.g. Kwok, 2006;Agnew et al., 2008;Howell et al., 2013). However, considering the application of the tracking algorithm in this study represents considerably larger spatial and temporal domains 210 it is important to assess the quality and uncertainty of the S1+RCM SIM vectors. To provide a quality assessment of the S1+RCM SIM vectors for each grid cell, the cross-correlation coefficient for all S1+RCM vectors in each grid cell were averaged. However, in order estimate the SIM uncertainty of all the S1+RCM vectors in each grid cell, a more structured approach was adopted.
Let us consider a grid cell containing a set of N sea ice velocity vectors ⃗ , where = 1,2, … , . Each vector has the 215 following uncertainty associated with the SIM tracking algorithm deriving the ice motion vector from two consecutive images: where, ∆ is the time interval (in days) separating two SAR images used to derive the considered ice velocity vector ⃗ . = 0.43 is the uncertainty in sea ice displacement (not speed) reported by Komarov and Barber (2013). Note that was derived for the SAR images separated by a variable time interval (1-3 days), so it must be divided by ∆ to come up with the 220 ice velocity uncertainty. The average velocity value assigned to the considered grid cell is the following: The SIM uncertainty of each grid cell, can be estimated as follows: where, 0 is the base uncertainty given as follows: 225 and  is the uncertainty score (varying from 0 to 1) associated with the methodology used to aggregate N individual ice motion vectors derived from pairs of images into the SIM product: where, ̅ is the average cross-correlation coefficient within each grid cell, is the maximum cross-correlation coefficient 230 within each grid cell. = is the fraction of time when SAR imagery are available. Here, represents the time interval when SAR data were available over the entire time interval considered ( ). = is the relative number of ice motion vectors used to create the aggregated mean ice velocity vector ⃗ . Here, is the number of ice motion vectors within the cell, and is the maximum possible number of the ice motion vectors for a grid cell.

Results and Discussion
4.1 S1+RCM sea ice motion Table 2 shows the number of SIM vectors detected for the pan-Arctic and each sub-region ( Fig. 1) based on over 60,000+ S1 and RCM images available over the period of March to December 2020. On average, there were 4,555,186 SIM 240 vectors detected each week for 2020. The majority (~60%) were located in the Central Arctic sub-region that contains the perennial Arctic sea ice pack and also has a very high weekly S1+RCM image density (Fig. 6). Fig. 8 illustrates the time series of 7-day ice speed from S1+RCM averaged over the entire pan-Arctic domain from March to December, 2020. The ice speed seasonal cycle as detected by S1+RCM was clear with ice speed decreasing during the melt season and increasing into the fall and winter. 245

Pan-Arctic
An example of the spatial distribution of S1+RCM SIM on March 11-17, 2020 and December 16-22, 2020 is shown in Fig. 9. Notable features for March include the Transpolar Drift, counter-clockwise SIM characteristic of a Beaufort Gyre reversal, and landfast (no ice motion) ice conditions within the majority of the CAA. The most notable feature for December was the clockwise SIM characteristic of the Beaufort Gyre. For the March 11-17 example, the spatial coverage was extensive with the exception of a gap within the Laptev Sea which is to be expected based the weekly image density (Fig.6). For the 250 December 16-22 example, the spatial coverage was also extensive and included in the Laptev Sea. Overall, the high density image coverage achieved with S1+RCM was able to provide weekly SAR derived estimates of SIM across the Arctic for 2020.

Canadian Arctic Archipelago (CAA)
Although not the primary focus of this study, Fig. 6 underscores that in addition to pan-Arctic S1+RCM SIM products, high spatial and temporal resolution regional S1+RCM products are certainly achievable for 2020. To that end, we generated 12.5 km, 7-day S1+RCM SIM for the CAA from March to December, 2020. The CAA is a region where SIM is not typically well resolved from coarser resolution satellites because of its narrow channels and inlets which makes automated SIM tracking landfast from November to July as ice speed is typically slow and confined to the periphery regions (Agnew et al., 2008). The spike in June was associated with an ice fracture in eastern periphery of the CAA (not shown). An example of the S1+RCM 275 SIM spatial distribution from August 12-18 at 12.5 km is shown in Fig. 11 and illustrates the considerable spatial variability of SIM within the CAA.
Given the considerable weekly image density available from S1 and RCM, further enhances in spatial and temporal resolution for more local scale SIM studies are also possible (e.g. Moore et al., 2021). Although the focus of this study is primarily large-scale SIM the latter point is important to demonstrate. Figure 12 illustrates and example of how SIM can be 280 resolved from resolutions of 6.25 km, 12.5 km, and 25 km for a 3-day temporal resolution within the middle of CAA. The level of detail that can be resolved at even 6.25 km is striking and even further increases are possible given the original 200 m spatial resolution.    Figure 13 shows the time series of the S1+RCM weekly average of the cross-correlation coefficients and across the pan-Arctic. Note that for automated SIM tracking algorithm used in this study, the cross-correlation coefficients are calculated for the second order derivatives (Laplacians) of the images, and not the original images; therefore, the crosscorrelation coefficients may appear lower than reported in the literature by other studies. Both the cross-correlation coefficient and exhibited the expected variability associated with the seasonal cycle of sea ice and remained relatively high and 305 stable during the dry winter conditions, decreased during the melt season and then returned to stability following the melt season (Fig. 13).

Spatiotemporal variability of sea ice motion vector quality and uncertainty 300
was initially high in early-March because of lower amounts RCM imagery when it first became operational (Fig. 13). As found in previous studies, higher and lower quality vectors are more apparent during the shoulder seasons (i.e. melt-freeze transitions) as a result of water on the surface of the ice and low ice concentration making automated feature tracking more difficult (e.g., Agnew et al., 2008;Lavergne et al., 2010). Moreover, there are also fewer 310 vectors detected during the shoulder seasons compared to dry winter conditions which contributes to higher . Figure 14 illustrates the spatially for selected weekly periods during the 2020 annual cycle. For all cases, values are typically found in the central Arctic and gradual increases outwards (Fig. 14). The observed spatial variability of is in part related to weekly SAR image density that decreases away from the central Arctic because they are primarily only covered by RCM (Fig.6). For example, higher was observed in the periphery regions of the Beaufort Sea, Hudson 315 Bay, and the Bering Sea during March 11-17 (Fig. 14a) and in Lapev Sea for December 16-22 (Fig. 14d).
was lower during the summer months but the weekly image density of S1+RCM (Fig. 6), provided considerably more images over the marginal ice zones during the melt season. As a result, there are more image data to better resolve challenging ice conditions during the shoulder as shown from August 11-17 (Fig. 14b) and September 30-October 7 (Fig. 14c).
https://doi.org/10.5194/tc-2021-223 Preprint. Discussion started: 22 July 2021 c Author(s) 2021. CC BY 4.0 License. Figure 15 summarizes and the cross-correlation coefficient using box-plots for the each Arctic sub-region from 320 March to December 2020. The interquartile range for most sub-regions were between 0.35-0.45 for the cross-correlation coefficient (Fig. 15a) and between 0.4-0.6 km/day for (Fig.15b). The largest was found in Bering Sea and the lowest in the Central Arctic (Fig. 15b). Lower cross-correlation coefficients were more apparent for regions that contain a significant portion marginal ice zone in 2020 (e.g. East Siberian Sea, Chukchi Sea, and Greenland Sea). Based on Fig.15, the association between higher (lower) and lower (higher) cross-correlation coefficients was not always apparent at the sub-325 region scale because of the variability of each regions sea ice physical characteristics over the annual cycle. Depending on how ice conditions evolve during the melt season, the distribution of regions with low ice concentrations may vary accordingly.
Sub-region may also change regionally in subsequent years depending if the weekly pan-Arctic S1+RCM image density changes. To that end, the regional and cross-correlation coefficients values presented in Fig. 15 should be interpreted as initial or baseline values since we are only considering the year 2020. 330

Comparison of S1+RCM against NSIDC and OSI-SAF
Comparing our S1+RCM SIM results with existing NSIDC and OSI-SAF SIM products provides additional 350 quantitative confidence metrics. In order to facilitate a representative 1-to-1 comparison between S1+RCM SIM and both the NSIDC and OSI-SAF SIM products, the spatial and temporal resolution of the S1+RCM were matched with the NSIDC and OSI-SAF SIM products for 2020. For OSI-SAF, S1+RCM was generated with a 2-day at 62.5 km and for NSIDC, S1+RCM was generated with 7-day temporal resolution and 25 km spatial resolution. For each product's temporal resolution (i.e. 7-day for NSIDC and 2-day for OSI-SAF), all the S1+RCM SIM vectors within each products grid cells (i.e. 25 km for NSIDC and 355 62.5 km for OSI-SAF) were averaged. The average of all the grid cells within each region across the Arctic (Fig.1) was then determined for all products. This resulted in 343 weekly averages for the S1+RCM and NSIDC comparison and 1957 2-day averages for the S1+RCM and OSI-SAF comparison. More samples were available from OSI-SAF because of the difference in temporal resolution of these two products (i.e., 2-day vs 7-day) Scatterplots of S1+RCM versus NSIDC and OSI-SAF are shown in Figs. 16a and 16b, respectively. Both existing 360 SIM products are in good agreement with S1+RCM with correlation coefficients greater than 0.85. Larger speeds are more apparent for S1+RCM with the mean difference (MD) 1.3 km/day for the NSIDC and 0.76 km/day for OSI-SAF for 2020. The root-mean square difference (RMSD) was lower for the NSIDC compared to OSI-SAF at 2.58 km/day and 3.25 km/day, respectively. We note better agreement with NSIDC because of its higher spatial resolution compared to OSI-SAF and the absence of larger ice speeds in NSIDC compared to OSI-SAF due to lower temporal resolution. The overall larger speed 365 associated with S1+RCM is most likely the result of higher spatial resolution compared to lower resolution satellite data used in NSIDC and OSI-SAF that is more difficult to track at lower spatial resolution because of temporal decorrelation. Kwok et al. (1998) also noted this problem when comparing SIM from passive microwave with SAR and found it also applies to regions of low ice concentration.
Although SAR has difficultly tracking sea ice in the vicinity of the marginal ice zone and regions of low concentration, 370 passive microwave has more difficultly. With the high S1+RCM image density available over the marginal ice zone during the melt season (Fig. 6) there was sufficient number of images available during the summer months for S1+RCM to be able to better resolve SIM compared to these existing SIM products with a lower nominal spatial resolution. For example, the spatial distribution of SIM for S1+RCM and NSIDC product in the Beaufort Sea is shown for September 9-15, 2020 in Fig. 17. While larger ice speeds are apparent with S1+RCM, there are also many regions along the marginal ice zone with sea ice concentration 375 (SIC) below 18% that are not detected by NSIDC SIM product.  385 Figure 17. The spatial distribution of a) S1+RCM sea ice motion and b) National Snow and Ice Data Center (NSIDC) sea ice motion in the Beaufort Sea region from September 9-15, 2020. Also shown are the sea ice concentration (SIC) polygons for 18% (black) and 81% (magenta). Note that the white areas in the figure indicate either zero ice motion for the landfast ice or no ice motion information extracted (because of no SAR data, no ice, or no stable ice features).

Conclusions
In this study we made use of 5 SAR satellites from S1 and the RCM with over 60,000 images to estimate SIM over the large-scale pan-Arctic domain from March to December 2020. The higher density image coverage of S1+RCM as oppose to just S1 and/or RCM provided more available SAR image pairs over Hudson Bay, Davis Strait, Beaufort Sea, Bering Sea, 395 and the North Pole. Results indicated that on average, 4.5 million SIM vectors from S1 and RCM were automatically detected https://doi.org/10.5194/tc-2021-223 Preprint. Discussion started: 22 July 2021 c Author(s) 2021. CC BY 4.0 License.
per week for 2020 facilitating the generation of large-scale S1+RCM SIM products. Notable spatial features were apparent (i.e. the Transpolar Drift and the Beaufort Gyre) and the seasonal cycle of sea ice speed exhibited the expected variability with decreases during the melt season and increases into the fall and winter. Moreover, by using an input spatial resolution of 200 m, more spatial heterogeneity in large-scale SIM was preserved as well as SIM was able to be resolved within the narrow 400 channels and inlets across the Arctic. The corresponding S1+RCM SIM vector quality (cross-correlation coefficients) and also reflected the seasonal cycle with lower quality vectors and higher during the shoulder seasons. Comparing the S1+RCM SIM estimates to the existing SIM datasets of NSIDC and OSI-SAF revealed that S1+RCM provides larger ice speeds and detects more vectors in the marginal ice zone. The advantages of detecting SIM from SAR as opposed to passive microwave were ultimately confirmed from a large-scale comparison. 405 S1+RCM SIM covered the majority of the pan-Arctic domain from March to December using a spatial resolution of 25 km and temporal resolution of 7-days. However, more consistent spatial coverage was achieved during the melt season given the weekly image density of S1+RCM decreases with decreasing latitude. Continued coordinated efforts by working groups like the Polar Space Task Group are encouraged to improve or refine SAR coverage across the pan-Arctic over the annual cycle. Although, covering the majority of pan-Arctic domain with SIM generated from S1+RCM at spatial resolutions 410 of less than 25 km and temporal resolutions less than 7-days was not consistently possible for 2020, we demonstrated that regional S1+RCM SIM products at higher spatial and temporal resolution are achievable given the weekly image density of S1+RCM.
The unique nature of the approach to automatically estimate SIM from S1+RCM described in this paper is that future refinements are possible. For instance, we anticipate adding HV channel to complement SIM estimated from HH polarization. 415 Should the timing of when S1 and RCM images acquired and received by ECCC becomes more consistent, we can explore if mixing S1 and RCM images provides most robust pan-Arctic SIM estimation. The anticipated launch of the NASA-ISRO (NISAR) L&S-band SAR satellite also provides an opportunity to add L-band into our SIM processing chain. L-band SAR would be able to provide improved SIM estimates during the melt season compared to C-band (Howell et al., 2018). Even without refining our approach, the upcoming launch of Sentinel-1C and Sentinel-1D will continue to facilitate large-scale SIM 420 from C-band SAR for many years to come.