Landfast ice, also commonly referred to as fast ice or shorefast ice, is sea ice that has become fastened to the coast and remains stationary for a period of time . Landfast sea ice is the most commonly encountered form of sea ice due to its proximity to Arctic coastal communities and its relative safety compared to drifting sea ice. Members of Arctic coastal communities use landfast ice for subsistence hunting and intercommunity travel, among other uses (e.g. ). Landfast ice also serves as a habitat for marine mammals and shorebirds () and can be used for industrial purposes (). As a rigid barrier between the ocean and land, landfast ice can mitigate coastal erosion and modify large-scale circulation patterns . The fundamental property of landfast ice that allows it to perform all these roles in the Earth system is its attachment to the land. proposed that this attachment could be determined from remote sensing data using two criteria: contiguity with the coast and immobility over time. Coastal contiguity can be determined from any single image with sufficient resolution to observe meaningful leads of open water, but determination of immobility requires at least two images of the same area acquired at different times. For example, analyzed triplets of co-located synthetic aperture radar (SAR) data spanning approximately 20 d to identify landfast ice in Alaska, while and derived landfast ice extent from 20 and 30 d composites of cloud-free moderate resolution imaging spectrometer (MODIS) data. Landfast ice is also identified in operational ice charts based on analysis of multiple images utilizing a range of satellite-based sensing methods . These various techniques of remote sensing produce similar information regarding the landfast ice in that the data is presented as a binary presence or absence of landfast ice within a grid cell. While these techniques are useful and valid for mapping landfast ice, the utility of landfast ice does not solely rely on the presence of fast ice but also the safety and stability, a measurement that these techniques and operational products cannot quantify. Interferometric Synthetic Aperture Radar (InSAR) has been shown to not only delineate landfast ice from mobile pack ice, but also measure centimeter- and millimeter-scale deformation within the fast ice , enabling the assessment of the landfast ice stability.

Interferometric coherence (γ^) is a measure of correlation between the two phase signals from a spaceborne radar and is generally only maintained over landfast ice, not pack ice. To perform an interferometric analysis between two SAR images acquired at different times, coherence must be maintained between the signal acquisition dates. Interferometry can be done when the two phase signals are of the same area acquired at different times (temporal baseline) and/or orbit positions (perpendicular baseline). Thresholding interferometric coherence to delineate landfast ice from mobile pack ice has been done before with the L-band (λ=23.6cm) sensor Phased Array type L-band Synthetic Aperture Radar (PALSAR) aboard the Japanese satellite Advanced Land Observing Satellite (ALOS) . The low frequency of PALSAR allowed for landfast ice to maintain coherence over long periods (∼46 d). In addition to the long temporal baseline, the L-band wavelength of PALSAR allows the signal to be more resilient to decorrelating due to changes in the surface. Decorrelation (i.e. reductions of coherence) is likely the result of a combination of:

errors in co-registration of the SAR images (γprocesses)

Our study area was chosen to coincide with that used in the development of the EM2025 landfast ice climatology (see Sect. ) and includes waters of the US Arctic Outer Continental Shelf and adjacent waters in Canada and Russia. The coastlines extend from just west of Neshkan on the Chukotka Peninsula to the easternmost point of the Russian mainland in the Bering Strait and from the westernmost point on mainland Alaska, the Iñupiat village of Wales, to the Mackenzie Delta in Canada. To conform with prior analyses of landfast ice , our study area is divided into two regions (Fig. ). The western region lies entirely within the Chukchi Sea and hereinafter is referred to as the Chukchi region, shaded cyan in Fig. . The eastern region includes a smaller area of the northern Alaskan Chukchi coast but is otherwise contained within the Beaufort Sea and hereinafter is referred to as the Beaufort region, blue shaded region in Fig. . The western and eastern regions meet at a point just west of Wainwright, Alaska.

The EM2025 landfast ice climatology (named after the same authors as this study and the year it was released) is a gridded daily dataset of landfast ice extent covering the study area depicted in Fig.  and spanning 1996–2023. A detailed description of the creation and results of the EM2025 dataset has been reported by and . In brief, the EM2025 dataset extends a previous SAR-based dataset of the same study area for the years 1996–2008 using data derived from ice charts produced by the National Ice Center (NIC) and the National Weather Service's Alaska Sea Ice Program (ASIP). Here, we use landfast ice width data derived from the EM2025 dataset using the SLIEalyzer toolbox (https://github.com/armahoney/SLIEalyzer, last access: 14 June 2026), which computes the distance from the coast to the seaward landfast ice edge (SLIE) along a specified set of approximately coast-normal vectors, as illustrated in Fig. . Regions of the coastline which are not captured by the coast vectors are termed “shadow regions” and denoted with shaded yellow regions in Fig.  and all subsequent maps. Shadow regions occur due to complexities of the coastline such as small bays or straits. In these areas the coast-normal vectors intersect the coast mask multiple times. This approach provides a consistent spatial reference frame with which to measure the seasonal and interannual variations in landfast ice almost continuously at 8892 different locations along the coast (6956 in the Chukchi region and 1936 in the Beaufort), spaced approximately every 200 m. It also allows us to calculate various statistics about the locations of the SLIE over different time periods.

In this study, the term InSAR describes a signal processing technique for calculating the phase difference between two radar signals acquired from similar locations in space at separate times. Here, we use pairs of Sentinel-1 interferometric wide (IW) beam images with temporal and perpendicular baselines of 12 d and ≤300 m m respectively. These image pairs were selected using the Short Baseline Subset tool within the Alaska Satellite Facility (ASF) Vertex portal. Thirteen IW reference scenes were chosen exclusively on the basis of coverage of the study region. Six reference scenes were used to cover the Beaufort region while seven were used to cover the Chukchi region. In total, 2084 SAR pairs were identified between the months of November and July for the period from March 2017 to July 2022. These SAR scenes were multi-looked by 20×4 looks, resulting in pixel sizes of 80m×80m. Multi-looking improves the signal-to-noise ratio and increases the interferometric coherence. We used ASF's Hybrid Pluggable Processing Pipeline (HyP3) toolbox to produce interferometric data products in GeoTIFF format.

To obtain a useful ϕ value, the two SAR signals must remain sufficiently coherent between Sentinel-1's repeat orbits. The HyP3 toolbox produces normalized coherence images where values of 0 indicate complete decorrelation and 1 indicates perfect coherence . The magnitude of the reduction of coherence (|γ^|) can be estimated as: 2|γ^|=γprocesses⋅γspatial⋅γthermal⋅γtemporal By limiting our spatial baseline to ≤300m and using constants specific to Sentinel-1 and C-band we can treat γspatial and γthermal as constants. In addition, we assumed that γprocesses is 1, indicating there are no errors in the coregistration of the SAR pairs. These assumptions allowed us to attribute all variability in the reduction of coherence to deformation/displacement of the surface, γtemporal.

In the context of sea ice, reductions of γtemporal are assumed to be caused by ice drift or deformation, which is typically on the order of kmd-1 for mobile pack ice. For drifting pack ice over a 12 d repeat orbit interval, γtemporal is reduced to zero, resulting in decorrelation (|γ^|=0). Fast ice by definition is stationary, but internal deformation still causes reductions of γtemporal. However, for the landfast ice to remain landfast, the magnitude of this deformation needs to be relatively small so to not cause fracturing. The coherence will be reduced in highly dynamic landfast ice; landfast ice will have a higher coherence compared to adjacent pack ice. This principle has previously been used by to delineate landfast ice from pack ice.

demonstrated the ability to delineate landfast ice from mobile pack ice using a coherence threshold and morphological filtering in Alaska. The normalized coherence threshold used by when using L-band PALSAR was 0.1. For this study, we use C-band which is more sensitive to smaller changes in the surface typically resulting in lower coherence values. The coherence threshold used by cannot simply be assumed to be the same for the different wavelengths due to the increased sensitivity of C-band. For that reason we derived a C-band specific coherence threshold to delineate landfast ice from mobile pack ice. A more detailed description of this derivation can be found in Section A1. In short, we used 14 SAR pairs, one from each month between December 2017–June 2018, from the Chukchi and Beaufort region. The Chukchi scene covered most of the northern Seward Peninsula and some of Kotzebue Sound. The Beaufort region scene covered the coastline west of Utqiagvik until approximately the Colville Delta. We applied thresholds ranging from 0.01 through 1.00 to the coherence images then applied morphological closing and opening operations to remove small and noncontiguous areas that exceed the threshold, which are likely noise. We then compared the areas identified by each threshold in each month to the extent of the minimum landfast ice extent identified in that month by the EM2025 dataset. The best threshold was then determined by the highest percentage of pixels that agreed with the corresponding EM2025 monthly minimum extent. The coherence threshold for delineating landfast ice from the mobile pack ice using C-band was determined to be 0.3 where areas which meet or exceeded this threshold were classified as landfast ice. This threshold was then applied to all normalized coherence images and used to measure landfast ice extent and to mask the interferograms, limiting our analysis to areas identified as landfast ice.

After masking the interferograms with the coherence threshold, we used the magnitude of the interferometric phase gradient to derive the apparent strain in each interferogram. For short-baseline 12 d SAR pairs over sea ice, we can assume that variations in interferometric phase ϕ are dominated by variations in line-of-sight surface motion rather than by topography. ϕ is insensitive to motion perpendicular to the line of sight, but has shown that it is possible to estimate two-dimensional horizontal strain of sea ice from the phase gradient, provided the phase slope is largely planar and the mode of deformation (e.g. radial divergence or rotation) is known. However, in general, derivation of two- or three-dimensional surface motion requires phase information from multiple look directions and the known deformation type. Instead, we define the term apparent strain (ϵa) to describe the magnitude of the horizontal gradient in line-of-sight displacement. Apparent strain can be calculated using the following: 3ϵa=λ4π(Δx)∇ϕ Where λ is the C-band SAR wavelength (5.64 cm) of Sentinel-1 and |∇ϕ| is the magnitude of the phase gradient, given by: 4|∇ϕ|=∂ϕ∂x2+∂ϕ∂y2(kernel length-1) Since ϕ is a cyclic quantity that wraps over an interval of 2π, we calculate ∇ϕ following the approach described by whereby ϕ is first converted to a complex quantity, ϕ*, with continuous real and imaginary components: 5ϕ*=eiϕ=cos⁡ϕ+isin⁡ϕ6∂ϕ∂x=∠∂ϕ*∂x7∂ϕ∂z=∠∂ϕ*∂z where ∠ indicates the argument of the complex exponent.

We approximate the partial derivatives in Eqs. () and () using finite differences across a 4 pixel window and insert them into Eq. () to derive the magnitude of the phase gradient. Apparent strain, ϵa, is then calculated using Eq. () and should be interpreted as the minimum net strain that occurred between SAR image acquisitions, since it only measures deformation along the satellite's line of sight. However, since it only represents displacement between two snapshots in time, it may underestimate the maximum strain that occurred during the 12 d period. As with the coherence images, we grouped ϵa results by month, according to the date of the primary image in each SAR pair, and mosaicked results to create images of monthly average ϵa for each of the Chukchi and Beaufort regions.

The Sentinel-1 SAR pairs used by along the Alaskan Beaufort coastline do not follow the same strict temporal baseline requirement of 12 d between pairs, prohibiting the direct use of these SAR pairs during the derivation of apparent strain thresholds for each stability class, but will be used for later comparison. To account for this, we interpreted the landfast ice regimes associated with each stability zone and created masks that reflect the characteristics of landfast ice in each zone, but are usable in all years (Fig. ). The spatial extents for each of these masks were determined independently of apparent strain values but instead by bathymetry, geomorphological characteristics, and features of the nearshore environment. These masks, which are shown in Fig. , represent areas that fit the characteristics of each landfast ice regime associated with each stability zone defined by . These stability zones were defined using the following criteria:

Bottomfast Landfast Ice (BFLFI) – Areas of the nearshore environment with a water depth shallower than 1.5 m

Application of the coherence threshold described in Sect.  to 2084 SAR image pairs allowed us to produce a total of 30 monthly mosaics of landfast ice. These monthly mosaics represent the apparent strain for each month from December and May during the 2017–2021 landfast ice seasons. We used the SLIEalyzer toolbox (Sect. ) to derive the average position of the SLIE over all seasons for each month for comparison with the equivalent results derived from the EM2025 dataset. For example, Fig.  compares positions of the InSAR-derived mean landfast ice extent during April (blue line) with that derived from the EM2025 dataset (cyan line) for the period 2017–2021. The SLIE positions show generally good agreement, with no consistent bias in the study region for this month. However, when averaging the difference between methods, the coherence mosaics tended to underestimate the landfast ice extent by between 1.5–5 km on average each month.

To compare the coherence-derived landfast ice extent against the EM2025 dataset for all months from December to May, we calculated the difference in monthly mean landfast ice width for each coast vector measured by the coherence-based method and the EM2025 dataset (Fig. ). Differences were computed by subtracting the EM2025-derived widths from the coherence-derived widths; negative values therefore indicate underestimation by the InSAR-based approach. Figure  represents the difference in widths along the entire coastline where coast vectors from the Chukchi region are plotted in cyan while the Beaufort region is in blue. During December (Fig. A), mean landfast ice width was zero throughout much of the study region and both datasets generally agree where landfast ice had not yet formed. However, where landfast ice was present (primarily in the Beaufort region), the InSAR-derived results tended to underestimate the width. On average in the Beaufort region during December, the coherence method measured 5.1 km less landfast ice at each coast vector. This underestimation is reflected in the predominance of negative differences, with the blue line lying below the zero-difference line in Fig. A.

There is a similar pattern for the month of January (Fig. B), with the InSAR-based approach still tending to underestimate the width of landfast ice as it expands throughout both study regions. The tendency to underestimate persists until mid-winter, February, March, and April, during which the relationship becomes more variable along the coast, with the InSAR-based approach indicating more than 10km more landfast ice than the EM2025 dataset at certain coast vectors (Fig. C–E). Although the average differences remain negative during the mid-winter months, with values of -5.1, -4.6, and -3.4 km from February to April in the Beaufort region, and -2.5, -3.7, and -1.5 km from February to April in the Chukchi region, these differences constitute only small percentages of the total landfast ice during this period. There is also notable spatial variability within the subregion, Inner Kotzebue Sound, during these months, with the difference in landfast ice width changing from <-10 km to >10 km and back again over the span of just a few coast vectors. The cause of this variability is not certain, but examination of the primary SAR imagery and interferograms over Kotzebue Sound indicate that the surface of the ice loses coherence temporarily without any substantial horizontal motion. This coherence loss could be caused by surface flooding, which has been observed to be caused by heavy snow load in this region . Kotzebue Sound also has extensive areas of shallow water (∼≤2 m), in which the ice can repeatedly interact with the seafloor as the water level rises and falls under the influence of winds and tides. This process can result in flexure fracturing of the ice surface, which can also lead to coherence loss. Similar observations of poor coherence using C-band SAR over Kotzebue Sound were made by .

In the month of May (Fig. F), the landfast ice width difference between the two datasets is more consistently negative. In the Beaufort region, the coherence-based method measured on average 10.7 km less landfast ice extent during May. Similarly, in the Chukchi region the coherence method measured 5.4 km less landfast ice than the EM2025 dataset. The differences are due to extensive areas where no landfast ice was identified by the coherence-based method, but landfast ice was still present in both the Chukchi and Beaufort regions in the EM2025 dataset. In particular, Kotzebue Sound never met the coherence threshold to be considered landfast ice during May of any season from 2017–2021. Surface melting is the most likely cause of this underestimate and highlights the limitations of using coherence to identify landfast ice outside of the winter months.

Although there is considerable variability in monthly mean ϵa at the scale of adjacent pixels, there was an overall tendency for lower ϵa values to be found near the coast and higher values to occur closer to the SLIE (Figs.  and ). This spatial distribution becomes more evident as the landfast ice season progresses and another tendency emerges in which the apparent strain in landfast ice tends to decrease over time (Fig. ). Both Figs.  and are representative of the mean seasonal progression of landfast ice extent and apparent strain values while Fig.  depicts how the monthly distribution evolves within the whole study region. The inset of each figure, denoted by a lowercase A, is a zoomed-in view near Utqiagvik, Alaska. Dark blue areas indicate very stable landfast ice low apparent strain, while yellow indicates relatively less stable landfast ice high apparent strain (Figs.  and ). The transition from yellow to green to blue, high to low apparent strain, within Elson lagoon throughout the winter months indicates that stability increases as landfast ice persists (Figs. Aa–Ca and Aa–Ca). By calculating the probability distribution of monthly ϵa values, the modal value of ϵa decreases from December to May (Fig. ). In addition, the distributions are right-skewed and the proportion of high apparent strain pixels decreases in the later months. Both of these changes result in a greater proportion of the landfast ice pixels having a lower apparent strain (i.e. more stable) during the later months of the season. During the month of May, very few pixels within the Chukchi region met the coherence threshold (Fig. C). In both regions, areas of landfast ice did not meet the coherence threshold to be considered landfast ice during the month of June, thus the distributions were omitted. The method performs robustly during the winter months, when landfast ice is most extensive and stable, and prior to the onset of surface melting.

In the Beaufort Sea, landfast ice typically reaches its seasonal maximum width during April . This is also the month for which qualitatively defined landfast ice stability throughout the Arctic based on interferometric phase gradient. Hence, we use the monthly mean ϵa values for the month of April to derive the probability distributions of ϵa within the three stability zones identified in Fig. . The distributions of apparent strain in each stability zone bottomfast (orange), sheltered (blue), and not-sheltered (black) are shown in Fig. . The bottomfast region has a bimodal distribution while the sheltered and not-sheltered zones have well-defined and distinct modes. The modal apparent strain value in the bottomfast ice zone is the lowest while the not-sheltered zone has the highest. The bimodal distribution of apparent strain values within the bottomfast stability zone is likely associated with tide cracks, which form at the oceanward boundary of bottomfast ice. The distributions for sheltered and not-sheltered zones are approximately lognormal with considerable overlap which is expected, as we assume the not-sheltered zone contains landfast ice which is stabilized by grounded ridges. The overlap between the distribution of ϵa values within the SHLFI and NSHLFI reinforces the idea that landfast ice which is shoreward of a barrier island, sheltered, has a similar ϵa as landfast ice shoreward of a grounded ridge (stabilized). In addition, the wider distribution apparent strain of the NSHLFI compared to the SHLFI is further evidence of multiple stability classes being present within the NSHLFI area.

To separate these stability classes objectively, we used Otsu's algorithm to determine ϵa thresholds for each class. Physically, the stability classes represent landfast ice which has similar regimes to the stability zones defined by . Bottomfast ice is broadly in contact with the seafloor (resting or frozen), stabilized ice is shoreward of a point feature (either barrier island or grounded ridge), and nonstabilized represents the floating extension which only has a point feature shoreward of its location. The apparent strain threshold for the bottomfast ice stability class was determined to be ϵa=8.6×10-6, marked by a yellow dashed line in Fig. . For the sheltered and not-sheltered classes, the thresholds were determined to be ϵa=1.8×10-5 and ϵa=2.4×10-5 marked by blue and black dashed lines, respectively (Fig. ). We chose the threshold of the sheltered class to differentiate the sheltered class from the not-sheltered class. This was done since we assumed that stabilized landfast ice exists within the not-sheltered region, landfast ice that is stabilized by grounded ridges. Furthermore, we assumed the apparent strain values of landfast ice which is stabilized by barrier islands and grounded ridges are similar. Determining thresholds using Otsu's method yielded two objectively determined thresholds that define the apparent strain bounds for our three stability classes:

Bottomfast: ϵa≤8.6×10-6

To test our apparent strain thresholds, we calculated the apparent strain of the interferograms used by along the Beaufort coastline and compared the extents of our quantitatively defined stability classes to their qualitative stability zones. Apart from the method of stability classification, qualitative vs. quantitative, did not enforce a strict temporal baseline for the SAR pairs, and they used a non-zero coherence value to identify the seaward landfast ice edge (SLIE). The three Sentinel-1 SAR pairs used by were: 8 April 2017–2 May 2017; 15 April 2017–9 May 2017; and 17–29 April 2017. The interferograms produced from these SAR pairs and the resulting apparent strain maps can be seen in Fig. A and B. The interferograms show the extents of the stability zones defined by with the areas outlined in red and identified as bottomfast, purple stabilized and cyan nonstabilized (Fig. A). The apparent strain values were sorted into the stability classes such that we could directly compare the extent of the qualitative zones and the quantitative classes (Fig. C). The main difference between the methods is the extent of the nonstabilized landfast ice. The higher coherence threshold used in this study, γ≥0.3, compared to the non-zero requirement by , resulted in an underestimate of landfast ice extent. However, the transition from stabilized to nonstabilized is captured well (Fig. Ca). In addition, the areas that are classed as bottomfast are separated from nearby stabilized landfast ice by a thin area of nonstabilized ice. It is likely these areas of high apparent strain are associated with tide cracks due to the geometry and location.

To further investigate the performance of the apparent strain thresholds, we produced a confusion matrix of the classification types by vs. this study's apparent strain classes (Table ). The nonstabilized class is the only class where both methods agreed on the majority of pixels which met the ≥0.3 coherence threshold (Table ). It is likely the classification disagreement, where the apparent strain indicated bottomfast ice while concluded that stabilized is the result of multiple grounded ridges in a small area. Both regions of large misclassifications, west of the Colville Delta and north of Prudhoe Bay (Fig. C), have been identified as common areas for grounded ridges . Thus we do not view the disagreement of classification as wrong, just that there are various processes that can reduce apparent strain, and in the right conditions, produce very stable landfast ice.

In general, landfast ice extent is captured well using interferometric coherence during the winter months of the landfast ice season but under-identifies landfast ice extent at the beginning or end of the season. Despite the typical presence of landfast ice in both the Chukchi and Beaufort regions during November and June , we did not find any pixels outside the land mask with coherence values ≥0.3 until December or after May. The coherence-based method does not capture landfast ice extent during the beginning and end of the landfast ice season, two important parts of the season when discussing changes between seasons. In the early winter, the loss of coherence over immobile landfast ice is likely caused by the rapid changes in the dielectric properties of the ice surface that occurred during the early stages of growth (e.g. ). During the months of December and January, the average monthly width of landfast ice identified by our coherence-based method under represents relative to the EM2025 dataset from 2017–2021 (-5.1 km in the Beaufort and -0.5 km in the Chukchi). Much of the difference in the Beaufort region occurs between Point Barrow and Kaktovik where our InSAR-derived results show the landfast ice is consistently 6–10 km narrower. In the Chukchi region, differences between the InSAR-derived width and EM2025 monthly mean width occur primarily in Kotzebue Sound. Similar to the area between Point Barrow and Kaktovik in the Beaufort region, the InSAR-derived width was on average 8.2 km lower than the EM2025 width in Kotzebue Sound during January.

Agreement between InSAR-derived landfast ice extent and the EM2025 dataset is best from February through April. The InSAR-based method still slightly underestimated landfast ice width compared to the EM2025 dataset by an average of 2.5 and 4.4 km, for the combined Chukchi and Beaufort regions from February through April. During these months, there are areas where the coherence-based method overestimated the landfast ice extent. Specifically, during April north of the Colville Delta (Fig. ) the coherence-based SLIE is 1–2 km further offshore than the EM2025 dataset. This is an area identified as a node and in proximity to grounded ridges . Finally, in May, the InSAR-based method underestimated the landfast ice width consistently across the study region. The consistent under measuring of landfast ice in May is attributed to surface melting causing a loss of coherence between acquisitions. Overall, the coherence-based identification of landfast ice measures the landfast ice width well, however there are certain areas where the combination of 12 d between acquisitions and C-band SAR prevents the methods from identifying landfast ice. On this basis, we find that 12 d repeat Sentinel-1 InSAR may be a useful tool for helping discriminate landfast ice extent during the coldest months of the year when the dielectric properties of the ice surface are most stable. With the imminent launch of the NISAR satellite, the combination of a 12 d repeat orbit with a L-Band sensor will likely improve the retention of coherence over landfast ice in the Arctic. In addition, NISAR might allow for coherence to be maintained over Antarctic landfast ice, something C-band is unable to do on a 12 d repeat orbit. showed that L-band coherence could be maintained for 45 d over landfast ice and so we anticipate that NISAR expands the useful window of this method for a greater portion of the landfast ice season.

We have shown that, over the study area, the distribution of apparent strain evolves such that the modal value of ϵa decreases monotonically from December to May across the whole study region (Fig. ). This pattern suggests that landfast ice becomes more stable the longer it remains in place over winter. To better understand the processes likely to be responsible for this, we partitioned this analysis among the 11 subregions shown in Fig.  and calculated the distribution of monthly mean apparent strain in each subregion (Fig. ). Similar to the study region as a whole, the mode of the apparent strain distribution decreases monotonically from month to month, with landfast ice transitioning toward more stable categories over time in most subregions (Fig. ). Subregion 9 (Point Barrow to Kaktovik) (Fig. ), is the best representation of the evolution of apparent strain in a subregion throughout the year due to the large extent amount of landfast ice and the variety of coastlines. It should be noted that, from December to April, this increase in stability occurs, while the overall extent of landfast ice also increases, suggesting that the process by which the SLIE advances also increases the stability of the landfast ice. In some subregions, the total landfast ice extent in May is less than that in April, so any increase in the modal ϵa value between these two months likely indicates that the least stable landfast ice is the first to detach. However, not all subregions follow a general decrease in apparent strain throughout the season.

Subregions with extensive lagoon systems have different trends of apparent strain distributions throughout the season. Subregion 1 (Northern Chukotka Peninsula), Subregion 2 (Northern Seward Peninsula), and Subregion 7 (Point Hope to Wainwright) are regions where the coastline is primarily composed of lagoons. The landfast ice within these lagoons forms prior to that formed outside of the lagoons. During the months when fast ice only exists within a lagoon, the distribution indicates a lower apparent strain mode. In this instance, all the landfast ice is being stabilized by the barrier islands, and depending on the depth of the water and the thickness of the ice, it could be bottomfast. The distribution of apparent strain within Subregion 7 (Fig. G), is a prime example of the distribution in a subregion dominated by lagoons. The monthly distributions are concentrated within the bottomfast class in the early months (December–February), then follow a typical seasonal progression in March through May (Fig. G). The high percentage of low apparent strain values in Subregion 7 during the early months is due to the landfast ice not existing outside of shallow lagoons (Fig. A–C). It is not until March when substantial amounts of landfast ice exist outside of the shallow lagoons.

We have shown that the three apparent strain classes we defined in Sect.  for bottomfast, stabilized, and nonstabilized landfast ice align well with the extent of equivalent stability zones identified qualitatively by from single interferograms. Within other single-interferogram apparent strain maps, we commonly observe strong gradients in apparent strain that are not evident in the monthly average maps. These strong gradients are likely associated with the transition from grounded to floating landfast ice. For example, at the seaward edge of the bottomfast ice, there are tide crack zones where the floating landfast ice flexes in response to sea level variations while the bottomfast ice remains stationary. The difference in vertical motion across this narrow region (typically tens of meters wide) leads to a region of high fringe density, like those found at the grounding lines of ice shelves and tidewater glaciers , resulting in high apparent strain. Examples of such features can be seen in two apparent strain maps from 5–17 April 2017 (Fig. A) and 9–21 April 2022 (Fig. B) near Oliktok Point in the Beaufort region. We see abrupt increases of apparent strain in both maps associated with a tide crack, a barrier island, and a grounded ridge (Fig. C). During 2022, the landfast ice was more extensive allowing for the formation of a second grounded ridge (Fig. C). The resemblance of these features to those around barrier islands leads us to identify them as the tide crack zone around grounded ridges. Moreover, they occur at the locations grounded ridges have been identified in this region during the 2021–2022 landfast ice season by . In addition, similar patterns in apparent strain values can be observed within the lagoon and oceanward of the barrier island, providing further evidence the abrupt gradient in apparent strain is associated with a stabilizing feature.

The shoreward feature identified as a grounded ridge occurred in approximately the same location in both 2017 and 2022, but the apparent strain pattern deviates offshore of this location by around 23 km. Both transects indicate a reduction of apparent strain offshore of the seaward grounded ridge (Fig. C). We believe this to be the local stabilized effect of a grounded feature. However, proving and interpreting this is outside the scope of this study. It is also possible the areas of high apparent strain gradients are not the grounded features, but the low apparent strain features surrounding these areas. A direct comparison to the location of the grounded ridges by could provide evidence of the apparent strain pattern related to the grounded features vs. the fast ice surrounding the ridge.