Terrestrial Snowmelt as a Precursor to Landfast Sea Ice Break-up in Hudson Bay and James Bay

Abstract

Numerous studies have been conducted to enhance our understanding of how climate change impacts landfast ice and its break-up in spring or summer. Yet, predictions of break-up timing have proven elusive illusive, and dependent on multiple environmental drivers. In this study, we investigate whether/to what extent snow melt on land adjacent to the coast can serve as a precursor to landfast ice break-up. For the study, we used MODIS snowmelt timing products to explore the pattern of snowmelt across the study area. This was compared against landfast ice break-up dates generated mainly from CIS ice-chart, across the Hudson Bay and James Bay coast. Snowmelt timings recorded across the study area followed the same latitudinal gradient as the landfast ice break-up dates, with almost a 50-day difference between the south and the north snowmelt and ice break-up timings. The data shows that the timing of landfast ice break-up across the Hudson Bay and James Bay region showed significantly stronger correlations with the timing of terrestrial snowmelt, as compared to air temperature thresholds, with an average 17-day gap observed between snowmelt to landfast ice break-up. Based on the observations, we believe snowmelt can be a reliable precursor to landfast ice break-up.

Résumé

De nombreuses études ont été menées pour mieux comprendre l’impact des changements climatiques sur la banquise côtière et sa rupture au printemps ou en été. Pourtant, les prédictions du moment de la rupture se sont avérées insaisissables, illusoires et dépendantes de multiples facteurs environnementaux. Dans cette étude, nous cherchons à savoir si, et dans quelle mesure, la fonte des neiges sur les terres adjacentes à la côte peut servir de précurseur à la rupture de la banquise côtière. Pour l’étude, nous avons utilisé les dates de la fonte des neiges dérivées des produits MODIS pour explorer le schéma de la fonte des neiges dans la zone d’étude. Ces données ont été comparées aux dates de rupture de la banquise côtière générées principalement à partir de la carte des glaces du SCG, sur la côte de la baie d’Hudson et de la baie James. Les périodes de fonte des neiges enregistrées dans la zone d’étude ont suivi le même gradient latitudinal que les dates de rupture de la banquise côtière, avec une différence de près de 50 jours entre les dates de fonte des neiges et de rupture au sud et au nord. Des corrélations significativement plus fortes ont été observées entre le moment de la rupture de la banquise côtière de la baie d’Hudson et de la baie James et le moment de la fonte des neiges terrestres, par rapport à des seuils de température de l’air, avec un écart moyen de 17 jours observé entre la fonte des neiges et la rupture de la banquise côtière. D’après ces observations, nous croyons que la fonte des neiges peut être un précurseur fiable de la rupture de la banquise côtière.

Introduction

Changes in the landfast ice regime over the last decades are one of the largest concerns for Arctic and Sub-Arctic coastal communities as it provides a stable platform over coastal waters every year for 3–9 months that is used for transportation, fishing and hunting (Ford and Furgal 2009; Gearheard et al., 2006; Galley et al. 2012: John et al. 2004; Laidler et al. 2009; Lovvorn et al. 2018; Mahoney 2018; Yu et al. 2014). The oil and gas industry relies on ice roads for transport to and from nearshore production and exploration facilities (Masterson 2009; Potter et al. 1981). Furthermore, the stable landfast ice cover protects the coast against erosive actions of waves and tides (Overeem et al. 2011). Over the last two decades the landfast ice regime has shown unpredictability in terms of earlier break-up, delayed ice formation, shortened ice persistence and thinner ice than usual (Fraser et al. 2021; Gupta et al. 2022; Maslanik et al. 2011; Rothrock et al. 1999; Tucker et al. 2001, Yu et al. 2004). This has emerged as a big concern for the coastal communities, increasingly unsafe transportation and loss of opportunities at hunt causes a huge impact on the socioeconomic and the cultural aspects of the communities (Meier et al. 2006; Pearce et al. 2010).

Numerous studies have attempted to look deeper into the climatology of landfast ice and investigate the role of various climatological factors responsible for the unexpected ice thickness, melt onset and ice break-up events (Brown and Cote 1992; Dumas et al. 2005; Jones et al. 2016; Persson et al. 2002). The springtime increase in annual air temperature has been traditionally linked with the melt onset and decay in the landfast ice regime (Cooley et al. 2020; Gough et al. 2004; Tivy et al. 2011). Hochheim and Barber (2010) reports an increase of 0.26 °C–0.3 °C per decade increase of spring air temperatures in Hudson Bay, which significantly correlates with the reduction of sea-ice concentration by 15%–20% per decade. However, in some regions changes in air temperature do not have a direct and singular control on landfast ice break-up (Gupta et al. 2022; Jensen et al. 2023). The landfast ice melting process is primarily initiated by temperatures rising and staying above freezing, yet on a larger scale melting corresponds to meteorological changes and ocean heat flux regimes, whereas on a smaller scale melt process is controlled by ice thickness and type due to lower heat flux through thicker ice than thin ice (Perovich et al. 2003). Hochheim and Barber (2014) weighed the role of different climatological forcings and concluded that though air temperature plays a dominant role in determining the sea-ice extent (70%–80%), dynamic forcings like winds have a significant impact on the regime as well. Gupta et al. (2022), showed that the western part of Hudson Bay experiences an earlier landfast ice break-up with ice persisting longer in the east, while air temperature trends showed a rising trend in some areas primarily in the west, south and east. Air temperature trends did not correlate well with the trends of landfast ice break-up, which led to the understanding that factors other than air temperature also had an important role to play in the break-up process.

The snow cover is another crucial factor influencing landfast ice growth and decay (Dumas et al. 2005). A significant part of the Arctic landmass and sea-ice surface experiences a snow cover lasting 5–8 months (Serreze et al. 2003). However, a general consensus of multiple climate models indicated toward a 15%–30% decline in the fall and spring snow cover fraction over the 2020–2050 period in Hudson Bay (Mudryk et al. 2018). Under the influence of the changing climate, snow cover extent in May and June has decreased by 14% and 46% respectively over the Pan-Arctic scale (Brown et al. 2010), whereas the duration of the snow cover is predicted to be decreased by 20%–30% over most of the Arctic by 2050 (Callaghan et al. 2011). Similarly, snow depth on land and on-ice across the Canadian Arctic was also noted to exhibit a declining trend between 1960 and 2020 time period (Lam et al. 2022). In the Arctic and Sub-Arctic, snow is a key climatic variable due to its high reflectance and thermal insulation (Massom 2001). Furthermore, snow is typically a storing unit for precipitation, hence melting of snow has an immense contribution to the ocean (on ice) and terrestrial system (in the form of river discharge), which collectively impacts the vertical structure of the Arctic Ocean and with it the sea ice regime (Serreze et al. 2003). Sustained air temperatures above 0 °C induce snowmelt on the ice surface, which result in the formation of melt ponds on the landfast ice while the interior sea ice remains cold and impermeable. The formation of melt ponds induces a dramatic decline in the surface albedo from dry snow (∼0.8), bare ice (∼0.6) to 0.15–0.3, thus considerably increasing the rate of ice melting (Diaz et al. 2018; Landy et al. 2014; Petrich 2012). Although a study by Ushio (2006), shows thickness of snow cover can play a crucial role in determining the landfast ice break-up frequency. The study shows that frequent break-ups were associated with lower snow depth (<60 cm), while higher snow depths resulted in stable ice conditions. Thinner snow cover allows enhanced absorption of solar radiation thus negating the chance of upward growth of ice by superimposed ice formation, this enhanced absorption thus makes the ice structurally weak and more susceptible to break-up.

On a terrestrial to marine scale, ideally, above 0 °C air temperatures are linked to the initiation of substantial snowmelt on land followed by the onset of melting of the landfast ice, while complete loss of ice pack through break-up happens much later than snowmelt on land. The snowmelt, however contributes to the volume of the terrestrial discharge which often results in the flooding of the landfast ice adjacent to the coast. This coastal flooding expedites the melting process of the nearshore landfast ice (Bareiss et al. 1999). Combined with its contribution to the terrestrial discharge, the loss of snow also results in significantly increased heating of the land surface and air due to solar radiation.

This study investigates the interconnection between snowmelt timing and landfast ice break-up. We hypothesize that snowmelt timing on land is a good predictor for the landfast ice break-up. The objectives for the study were to (1) observe the pattern of snowmelt across the study area and (2) examine how it related to the progression of landfast ice break-up. The observations were evaluated across 14 communities in the Hudson Bay and James Bay region from 2001 to 2018.

Materials and methods

Study area

The Hudson Bay and James Bay together forms the largest inland water body in North America and the second largest bay in the world. Stretched between 51°N–66°N latitude and 95°W–76° W longitude, makes this bay experience a diverse climatic regime throughout its expanse. Considering the resolution of this study, 14 coastal communities namely, Coral Harbor, Chesterfield Inlet, Rankin Inlet, Arviat, Churchill, Cape Tatnum, Fort Severn, Peawanuck, Attawapiskat, Moosonee, Chisasibi, Inukjuak, Akulivik and Ivujivik were selected for the study. The selections were informed from Gupta et al. (2022) such that the observations from these community locations would be representative of an east-west contrast and latitudinal gradients in the landfast ice climatological regime.

Landfast ice break-up dates (BUD)

Detection of landfast ice break-up dates have been described using several methods depending on the type of observational datasets used. In a study by Mahoney et al. (2006) the author provides a classification of stable landfast ice based on a criterion that states that the ice mass attached to the shore stays immobile and exhibit no detectable motion for a 20-day period to be classified as stable landfast ice. In another study by Cooley et al. (2020), the authors used a criterion that states the first day of the year with a > 90% loss of ice mass in a 20 km radius of the study area are defined as break-up dates. For this study we use the method proposed in Gupta et al. (2022) to determine the break-up dates. Here, we defined landfast ice break-up dates or BUD as the day of the year when the complete loss of landfast ice was observed over the selected zone. Monthly and weekly ice charts produced by the Canadian Ice Service were used as the primary dataset for this analysis. Ice charts provide information on the stages of development and concentration of both mobile and landfast ice. Landfast ice is classified and represented in the ice charts as a unique class such that it can be isolated from other ice types. The interval in which ice charts are produced only allow us to identify the week when the break-up took place, based on an ice present to absent scenario. Once we narrowed-down to the week of ice break-up, daily observations from the MODIS and VIIRS were used to determine the exact day of landfast ice break-up. Observations were based on a 50 km segment along the coastlines with 25 km in one direction along the coastline and 25 km in the opposite direction, at the 14 selected study locations. The first daily observation with a complete loss if ice along the 50-km coastline section was identified as BUD. As cloud interference limits the observation in optical imageries, we used the day with first observation of visible ice loss as the BUD. In some cases, when observations over the entire week was obscured by clouds, the day of the year from the ice chart showing ice loss was considered as BUD.

Determination of snowmelt dates (SMD)

The determination of snowmelt dates or SMD was done using version 2 of the snowmelt timing maps (STMs) produced using the MOD10A2 collection 6, which is a standard 8-day composite snow-cover product (O'Leary et al. 2019). These yearly maps provide the date of the timing of snowmelt on land at a spatial resolution of 500 m. This product uses the Normalized Difference Snow Index (NDSI) to estimate the presence of snow in a pixel, further using a 50% snow covered area for a snow present/absent classification threshold. The date with pixels classified as snow free for two consecutive observations were considered as snowmelt dates. The STMs are available for the time period between 2001 and 2018; hence, the observational period for this study was restricted to the same timeline.

For this study, a 50-km long and 10-km wide areas along the landward side of the coastlines were considered for determining the SMD for each selected community. These observational areas were adjacent to the 50 km segments used for the landfast ice observations. The SMD for each year were determined from the averages of the pixel values within the 50 × 10 km areas. These dates were further used to study the chronology of the SMD and BUD and how they vary across the Hudson and James Bay coastline.

Determination of the spring temperature transition day (STT) using air temperature

The spring temperature transition day (STT) defines the day of the year when the 7-day running mean of air temperature reaching or crossing 0 °C and remains so for a period of at least 7 consecutive days. For this we used the 2 m surface air temperature records from the ERA5 climate reanalysis product produced by the ECMWF (Hersbach et al. 2018). The dataset has a spatial resolution of 0.25° × 0.25°, produced at a 1-hour temporal resolution. Further, these hourly dataset was used to generate daily means, which was then used to derive the STT.

Results and discussion

Air temperature and landfast ice break-up

Observations across the selected study locations indicated a general pattern of landfast ice break-up starting in the lower latitudes and gradually progressing toward the north (Figure 1a,b), with almost a 50-day gap between the earliest (Moosonee in mid-May) and last (Coral Harbor in early July) observed mean landfast ice break-up dates (BUD) (Figure 2a). The trends from 2001 to 2018 reveal later than usual break-up at the southern and eastern Hudson Bay and James Bay communities of Fort Severn (4 days/decade), Peawanuck (5 days/decade), Attawapiskat (3.6 days/decade), Moosonee (5 days/decade), Ivujivik (3.3 days/decade) and Akulivik (5 days/decade), whereas BUD at the more northerly Arviat (5 days/decade), Rankin Inlet (3.6 days/decade) and Chesterfield Inlet (5 days/decade) is occurring earlier, while Coral Harbor and Inukjuak showed no trend (Figure 2b). These varied trends across the study area can be associated with changing weather patterns resulting in changes to the sea ice drift patterns (Gupta et al. 2022; Kirillov et al. 2020). Gagnon and Gough (2005, 2006), showcased similar east-west asymmetry in case of mobile sea-ice regime and landfast ice thickness in the Hudson Bay region, aiming at the role of air temperature, snow depth and timing of ice break-up and freeze-up.

Figure 1. Map showing the 2001–2018 average of, (a) snowmelt dates and (b) spring temperature transition days with landfast ice break-up dates observed across Hudson Bay and James Bay. In the figure, COHB: Coral Harbor; CHST: Chesterfield Inlet; RNKN: Rankin Inlet; ARVT: Arviat; CHRL: Churchill; CTAT: Cape Tatnum; FSVN: Fort Severn; PWNK: Peawanuck/Winisk; AWPK: Attawapiskat; MSNE: Moosonee; CHSB: Chisasibi; INJK: Inukjuak; AKVK: Akulivik and IVJK: Ivujivik.

Figure 2. Statistics of spring temperature transition days (STT), snowmelt dates (SMD) and landfast ice break-up dates (BUD) calculated over the 2001–2018 time interval, across the 14 communities. (A) c and e show violin plots representing the 18-year mean and variability of the observed parameters at the study locations. (B) d and f represent the trends in days per decade over the 2001–2018 time period using Mann Kendall’s Test and Sen’s Slope. Locations denoted with “*” represents statistically significant values at a 90% confidence level.

A study by Cooley et al. (2020) has shown a strong relation of landfast ice break-up with springtime air temperature. Results of this study indicate that air temperature is the dominant factor triggering landfast ice break-up, however its influence is shown to vary by region and uncertainties in the study allow for the influence of other contributing factors like winds and ocean temperatures. Interestingly, in this study the 2001–2018 mean spring transition dates (STT), determined from air temperature at the study locations, follow a similar pattern as the progression of BUD. Although the northernmost and southernmost study locations (Coral Harbor and Moosonee, respectively) had a large difference in their mean STTs (DOY 153 and DOY 100, respectively) following the expected latitudinal contrast (Figure 2c). Within the selected study locations, the Inukjuak, Akulivik and Ivujivik locations along the eastern coasts of Hudson Bay experienced STT toward the end of May (DOY 141, DOY 142 and DOY 148, respectively) whereas the west coast locations, Arviat, Rankin Inlet and Chesterfield Inlet, experienced a later STT (DOY 146, DOY 148 and DOY 151, respectively).

Correlations between STT and BUD for all 14 locations were positive and statistically significant at 90% confidence interval (Table 1), with some locations showing a rather strong correlation namely, Rankin Inlet, Fort Severn, Inukjuak and Ivujivik. Statistically it showed that the timing of landfast ice break-up was related primarily with the timing of spring transition, indicating a dominant role of air temperature as a controlling factor for landfast ice break-up. Similar observation was reported by Cooley et al. (2020). However, comparing the trends of BUD and STT for the 2001–2018-time period over the study area suggests a different argument. Trends of landfast ice break-up at Chesterfield Inlet and Rankin Inlet show an earlier than usual break-up in these locations, yet the STTs occur later than usual (Figure 2d). In the Hudson Bay and James Bay region landfast ice break-up occurred an average of 31 days (±9 days) after the spring transition. In general, the timing of the spring transition should be directly related to the timimg of landfast ice break-up and snowmelt timing, yet the weaker correlations with the prior and inconsistency in the signs of these trends in some locations indicate toward a possible influence of other climatological controls in the region. Hochheim and Barber (2014) suggested an important contribution of winds leading to ice break-up even though temperature acted as a primary control. Similarly, Gupta et al. (2022) indicated the role of geophysical controls like coastal orientation and bathymetry greatly influencing landfast ice stability and persistence. Hence, further in this study we investigated the role of terrestrial snowmelt which is similarly impacted by temperature yet has an impact on the ice regime. However, it is important to consider that this inconsistency in trends does not account for the measurement errors in calculating STT, BUD and SMD.

Table 1. Spearman’s Correlation coefficient of landfast ice break-up dates (BUD) with snowmelt days (SMD) and spring transition days (STT) in all selected study locations.

Locations	BUD-SMD			BUD-STT
	r value	p value	coef dter	r value	p value	coef dter
Chesterfield Inlet	0.682	.002	0.465	0.434	.073	0.188
Rankin Inlet	0.741	.001	0.548	0.877	.0001	0.769
Arviat	0.767	0	0.588	0.682	.002	0.466
Churchill	0.844	.0001	0.712	0.792	0	0.628
Cape Tatnum	0.764	0	0.583	0.704	.002	0.496
Fort Severn	0.656	.004	0.43	0.816	.0001	0.665
Peawanuck	0.868	.0001	0.753	0.665	.003	0.443
Attawapiskat	0.756	0	0.572	0.604	.009	0.365
Moosoonee	0.713	.001	0.508	0.64	.005	0.41
Chisasibi	0.775	0	0.6	0.716	.001	0.513
Inukjuak	0.602	.01	0.363	0.827	.0001	0.684
Akulivik	0.765	0	0.585	0.544	.021	0.296
Ivujivik	0.213	.396	0.045	0.807	.0001	0.651
Coral Harbor	0.514	.031	0.264	0.775	0	0.601

Bold values represent stronger “r” values.

Role of snowmelt on land as a precursor of landfast ice break-up

Snowmelt dates (SMD) estimated across the selected study locations reveal a pattern similar to that of landfast ice break-up and spring temperature transition (Figure 1a). The south-to-north timing difference of 46-days of the snowmelt was observed to be very similar to the 50-day interval observed for landfast ice break-up. Considering a bay-wide scenario, loss of snow cover shows a latitudinal progression with Moosonee experiencing SMD as early as the beginning of May (DOY 121) and finally Coral Harbor and Ivujivik where the SMD was around the middle of June (DOY 164 and 167 respectively) (Figure 2e). This pattern of SMDs across the 14 locations aligned with the mean landfast ice BUDs observed (Figure 2a). Correlations between the SMD and BUD calculated for all selected study locations revealed a rather strong correlation (Table 1), with the strongest value observed at Peawanuck (r = 0.87, p = .0001) and the lowest significant correlation at Coral Harbor (r = 0.51, p = .03).

However, the trends of landfast ice and snowmelt timing across the Hudson Bay and James Bay region over the 2001–2018 period did not align and most trends were not statistically significant at a 90% confidence level (Figure 2b,f). Combining the results of STT, SMD and BUD estimations revealed a uniform chronology in the timing of these events across most of the selected study sites. Ideally, springtime air temperature rising and staying above 0 °C initiates the melt of landfast ice as well as the snow cover. Snow cover experiences a more rapid warming and melt compared to ice (Ledley 1991). As the snow cover undergoes melting and this meltwater is released into the system, which then eventually floods the landfast ice. Bareiss et al. (1999) showed nearshore landfast ice break-up to be a function of flooding of the landfast ice cover. This flooding often carries sediments which is deposited on the ice surface thus decreasing the surface albedo and eventually promoting increased surface heating. Apart from sediment deposition, flooding also results in the transfer of heat from the terrestrial to marine environment (Dean et al. 1994; O’Brien et al. 2006). With the combined action of warmer terrestrial input and increased surface heating due to low albedo, localized melting is enhanced and finally results in the complete decay of the ice pack adjacent to the coast, thus making the ‘fast’ ice pack now mobile and hence classified as landfast ice break-up.

To explore if SMD can function as a proxy for landfast ice melt onset, we compared the DOY of SMD and BUD estimated cross all the 14 locations over the 2001–2018 time interval (Figure 3a). The timings of these two events had a positive relationship and strong correlation (r = 0.717, p = .0001) of inter-annual variability. Interestingly, in a bay-wide perspective correlations of spring transition with BUD and SMD were also noted to be positive and significant (r = 0.689 and 0.680 respectively), but weaker to that of the correlation of BUD with SMD (Figure 3b,c). Even though both STT and SMD were well correlated to BUD, snowmelt had a better statistical relationship with the timing of landfast ice break-up. Similarly, at the individual locations, timing of break-up in 9 out of 14 locations exhibited a better relationship with snowmelt timing compared to spring transition (Table 1). These observations thus suggest that snowmelt timings have substantial contribution to the landfast ice break-up process and can be used as a more reliable and immediate proxy for predicting landfast ice break-up.

Figure 3. Relationship between spring temperature transition day (STT*), snowmelt day (SMD*) and landfast ice break-up day (BUD*) anomalies. This was calculated over 14 communities over an 18-year period (from 2001 to 2018). The anomalies were obtained by removing the location-specific mean values from the SMD, STT and BUD data to avoid any impact of the latitudinal gradient in the distribution and further inferring toward the role of complex climatological controls on these events.

Considering a bay-wide scenario, a mean gap between SMD and BUD of 17 days (±7 days) was recorded. The largest time difference between the two events was noted for the Peawanuck area (24 days) and the shortest gap was noted at Ivujivik (4 days) (Figure 4). However, in this 18-year time period we have also noticed ample variations across the study locations and some noteworthy extreme events. While the highest documented gap between SMD and BUD was 60 days recorded at Inukjuak in 2009, some gaps were negative (24 and 13 days in Ivujivik in 2001 and 2009 respectively) meaning landfast ice break-up event took place earlier than the mean SMD. Even though the 18-year bay-wide mean of snowmelt to ice break-up gap was 17 days, such extreme events pose a risk to the communities to carry out travel or hunting activities on the ice. Such events where the ice breaks-up before the complete loss of snow or very shortly after the loss of snow can be attributed to extreme weather events like winds, waves, tidal amplitudes and storms. A recent study by Jensen et al. (2023) indicated that a strong and high frequency winter storms can create substantial stress on the landfast ice cover which combined with ocean circulation can impact break-out frequency of ice. Similarly, tidal amplitudes have been observed to enhance break-out frequency by inducing stress on the anchor strength of the grounded pressure ridges in low tide conditions (Jones et al. 2016). Hence, careful consideration of such weather anomalies and geophysical forcings is imperative for a better and more precise prediction of ice conditions.

Figure 4. Variation in the snowmelt to break-up gaps estimated across all 14 communities over the years 2001 to 2018. The ‘x’ in the plot designates the mean of the distribution. The points on the plot area represent extreme events when the gap period was unusually high and/or low.

Conclusions

Gupta et al. (2022) points out that other factors apart from spring temperature may play an important role in determining the timing of landfast ice break-up in Hudson Bay, and further inferred toward a more complex process involving other environmental factors to have a significant impact on the break-up process. Taking a step forward, this study investigated the landfast ice break-up event with emphasis on air temperature patterns and snowmelt timing to have a better understanding of the landfast ice break-up climatology This was achieved by addressing the investigation in two parts, i.e., (1) observing the pattern of snowmelt across the study area and (2) examine how it related to the progression of landfast ice break-up. These observations were evaluated across 14 coastal communities along the Hudson Bay and James Bay coastline from 2001 to 2018.

The patterns of spring air temperature transition dates (STT) and landfast ice break-up dates (BUD) across the selected study locations aligned well and showed a positive correlation. Snowmelt dates (SMD) and BUD across the bay exhibited a latitudinal gradient with loss of snow and landfast ice cover was first observed in the south and gradually progressing toward the higher latitudes SMD and BUD followed a similar pattern and a strong correlation was observed between the two. Even though both SMD and STT were well correlated with landfast ice break-up timing, SMD displayed a stronger statistical relationship to BUD at most locations. For most study locations the time interval between snowmelt and landfast ice break-up remained uniform, a bay-wide average interval of 17 (±7) days was achieved. This meant that decay or complete detachment of the nearshore landfast ice cover can be expected in almost 17 days of the snowmelt event, however the location of the study area is very important in this estimation. The results suggest that SMD, as opposed to spring air temperature patterns can act as a strong precursor to landfast ice break-up and can have a possible future application in ice prediction based studies, provided further investigations are done across the Arctic and Sub-Arctic to examine the robustness of the theory. However, the risk of extreme events where landfast ice break-up taking place before the estimated snowmelt timing or shortly after the snowmelt timing remains a huge concern. Weather events leading to a warmer winter, late snow fall or storm events can result in such extreme scenarios. As climate change intensifies, more extreme weather events may lead to more unpredictable behavior in ice break-up patterns making its further unsafe for the coastal communities. We therefore do not suggest the use of SMD to help guide human activities associated with late season landfast sea-ice.

Authors Contributions

• Contributed to original data acquisition: KG.

• Contributed to analysis and interpretation of the data: All authors

• Drafted the article: KG.

• Revised the article: All authors.

• Approved the submitted version for publication: All authors

Data accessibility statement

The datasets used in this study are available in numerous websites and freely accessible. The sources of the primary datasets used in the study are as follows:

• Weekly ice charts from The Canadian Ice Service.

  (https://iceweb1.cis.ec.gc.ca/Archive/page1.xhtml?lang=en)

• Air temperature reanalysis products from ERA-5.

  (https://cds.climate.copernicus.eu/cdsapp#!/dataset/reanalysis-era5-single-levels?tab=overview)

• Snowmelt timing records from the Oak Ridge National Laboratory Distributed Active Archive Center (ORNL DAAC)

  (https://daac.ornl.gov/cgi-bin/dsviewer.pl?ds_id=1712)

Other datasets include satellite images from Landsat series (https://earthexplorer.usgs.gov/) and daily observations from the online viewing platform NASA Worldview (https://worldview.earthdata.nasa.gov/).

Acknowledgements

The authors thank all collaborators of the BaySys program and Manitoba Hydro. This work is a contribution to the Arctic Science Partnership (ASP, asp-net.org) and ArcticNet.

Disclosure statement

The authors have no competing interests to declare.

Additional information

Funding

This work is a contribution to the Natural Sciences and Engineering Council of Canada (NSERC) Collaborative Research and Development project: BaySys (CRDPJ 470028-14) led by D. Barber (Academic PI) and K. Sydor (Industry PI). Individual support from Canada Research Chair Tier 1 to David Barber, NSERC Discovery Grant to Jens Ehn and Canada Excellence Research Chair to Dortha Dahl-Jensen has been provided to Kaushik Gupta.

Notes on contributors

Kaushik Gupta

Kaushik Gupta (Lead author): Ph.D. Candidate, Center for Earth Observation Science, University of Manitoba, Canada.

Jens K. Ehn

Jens Ehn (Co-author): Assistant Professor, Center for Earth Observation Science, University of Manitoba, Canada.