Spring melt pond fraction in the Canadian Arctic Archipelago predicted from RADARSAT-2

Melt ponds form on the surface of Arctic sea ice during spring, influencing how much solar radiation is absorbed into the sea ice–ocean system, which in turn impacts the ablation of sea ice during the melt season. Accordingly, melt pond fraction (f_p) has been shown to be a useful predictor of sea ice area during the summer months. Sea ice dynamic and thermodynamic processes operating within the narrow channels and inlets of the Canadian Arctic Archipelago (CAA) during the summer months are difficult for model simulations to accurately resolve. Additional information on f_p variability in advance of the melt season within the CAA could help constrain model simulations and/or provide useful information in advance of the shipping season. Here, we use RADARSAT-2 imagery to predict and analyze peak melt pond fraction (f_pk) and evaluate its utility to provide predictive information with respect to sea ice area during the melt season within the CAA from 2009–2018. The temporal variability of RADARSAT-2 f_pk over the 10-year record was found to be strongly linked to the variability of mean April multi-year ice area with a statistically significant detrended correlation (R) of $R=-$0.89. The spatial distribution of RADARSAT-2 f_pk was found to be in excellent agreement with the sea ice stage of development prior to the melt season. RADARSAT-2 f_pk values were in good agreement with f_pk observed from in situ observations but were found to be ∼ 0.05 larger compared to MODIS f_pk observations. Dynamically stable sea ice regions within the CAA exhibited higher detrended correlations between RADARSAT-2 f_pk and summer sea ice area. Our results show that RADARSAT-2 f_pk can be used to provide predictive information about summer sea ice area for a key shipping region of the Northwest Passage.

1 Introduction

Arctic sea ice extent during the summer months has declined considerably over the satellite record (Serreze et al., 2007; Stroeve et al., 2012; Peng and Meier, 2017). Surface melt ponds, which form on sea ice during the spring, play an important role in the decay of sea ice and seasonal reduction in ice extent because they influence how much solar radiation is absorbed into the sea ice–ocean system (Eicken et al., 2004). Specifically, the accumulation of meltwater on the surface of the sea ice lowers the albedo from ∼ 0.8 to between 0.2–0.4 and enhances melt (Perovich et al., 2002). The topographical constraints over multi-year ice (MYI) imposed by hummocks typically result in MYI exhibiting a lower melt pond fraction (f_p) compared to seasonal first-year ice (FYI) (Grenfell and Perovich, 2004; Polashenski et al., 2012; Landy et al., 2015). With Arctic sea ice transitioning from a MYI- to FYI-dominated icescape (Maslanik et el., 2011), the lower f_p of MYI will gradually be replaced with the higher f_p of FYI, facilitating even more sea ice energy absorption and further enhancing sea ice melt (Perovich and Polashenski, 2012).

Predicting the state of Arctic sea ice several months in advance is challenging, and recently the sea ice prediction community has focused efforts on the development and utilization of dynamical forecast models (e.g. Chevallier et al., 2013; Sigmond et al., 2013; Guemas et al., 2016). Despite these recent efforts, rapidly changing Arctic sea ice conditions will continue to necessitate improved sea ice forecasting capabilities (Eicken, 2013). Accordingly, prognostic f_p schemes have been integrated in climate models and have shown to exert a strong influence on summer sea ice area and extent (Flocco et al., 2010, 2012). Schröder et al. (2014) found a strong correlation between model-simulated May f_p and the observed September sea ice extent. Observed f_p has also demonstrated significant predictive skill for September ice extent from late July onwards (Liu et al., 2015). However, while f_p estimates for the entire Arctic can be provided by model simulations, more representative and higher spatial resolution observational estimates at regional and pan-Arctic scales are much more difficult to obtain.

Optical remote sensing is the most widely utilized approach to estimate large-scale f_p from space (e.g. Markus et al., 2003; Tschudi et al., 2008; Rösel et al., 2012; Istomina et al., 2015; Webster et al., 2015; Lee et al., 2020) but cloud cover remains a significant problem. Techniques for retrieving f_p using advanced quad-polarization and compact-polarization mode synthetic aperture radar (SAR) imagery, at C- and X-band frequencies, have also been developed (Scharien et al., 2014; Fors et al., 2017; Li et al., 2017), but they are limited in systematic spatial application because the required polarization modes are not always available from wide-swath imagery. However, using the winter backscatter from widely available Sentinel-1 SAR imagery, Scharien et al. (2017) recently demonstrated a technique for predicting spring peak melt pond fraction (f_pk) over the entire Canadian Arctic Archipelago (CAA) 3–4 months in advance of melt pond formation. These f_pk predictions have potential utility in seasonal summer sea ice area and extent forecasts as early as April.

The CAA is a collection of islands located in northern Canada (Fig. 1) whose waterways are sea ice covered between fall and spring. It is an active region for marine shipping and has recently experienced an increase in summer shipping activity (Pizzolato et al., 2014). Model simulations have been utilized to understand the current and predicted future variability of sea ice conditions in the CAA (e.g. Dumas et al., 2006; Sou and Flato, 2009, Howell et al., 2016; Laliberté et al., 2016; Hu et al., 2018; Laliberté et al., 2018). However, modelling the CAA still remains challenging because complex sea ice dynamic and thermodynamic processes are often not accurately resolved in its narrow channels and inlets. In addition, the response of the CAA to climatic change is perhaps counter-intuitive as longer melt seasons are resulting in increased MYI import from the Arctic Ocean during the summer months (Howell and Brady, 2019). Since f_pk is linked to summer sea ice melt processes (e.g. Eicken et al., 2004; Skyllingstad and Polashenski, 2018) additional information on f_pk variability within the CAA could improve our understanding of regional summer melt processes, help constrain model simulations, and facilitate safer shipping activity in upcoming years.

Figure 1 Map of the Canadian Arctic Archipelago region (red shading). The green star indicates the location of the lidar and aerial photograph observations.

In this study, we extend the work of Scharien et al. (2017) and investigate predicted f_pk variability within the CAA over the longer-term record available from RADARSAT-2. Specifically, (i) we estimate f_pk in the CAA using RADARSAT-2; (ii) evaluate the spatiotemporal variability of f_pk in the CAA from 2009–2018; (iii) compare RADARSAT-2 f_pk values to Sentinel-1 f_pk values from Scharien et al. (2017), in situ f_p observations from Landy et al. (2014) and Moderate Resolution Image Spectroradiometer (MODIS) f_p values from Rösel et al. (2012); and (iv) investigate the utility of RADARSAT-2 f_pk to provide predictive information about sea ice area in the CAA during the summer melt season.

2 Methodology

2.1 Data

The primary dataset used in this analysis was 5.405 GHz (wavelength, λ=5.5 cm; C-band) SAR imagery in ScanSAR wide mode at HH polarization from RADARSAT-2 acquired over the CAA (Fig. 1) in April from 2009–2018 (Table 1). RADARSAT-2 ScanSAR wide mode imagery has a spatial resolution of 100 m with an incidence angle range of 20.0 to 49.3^∘. We limited our analysis to only RADARSAT-2 images at HH polarization because Scharien et al. (2017) found HV produced noisy results in addition to there not being sufficient imagery at HV polarization in the early period of the RADARSAT-2 record to cover CAA in April.

Table 1 Number of RADARSAT-2 images acquired over the Canadian Arctic Archipelago in April for 2009–2018.

In situ observations of f_p on landfast FYI were obtained in 2 consecutive years from sites in the CAA using a terrestrial light detection and ranging (lidar) system (Landy et al., 2014) (Fig. 1, green star). In 2011, the site was located in Allen Bay on FYI with relatively rough surface topography, whereas in 2012, the site was located in Resolute Passage on FYI with relatively smooth topography. At each site, time series of f_p observations were collected within the same 100×100 m area of the ice over a 2- to 3-week period following melt onset, covering three of the four stages of melt pond evolution detailed in Eicken et al. (2002). The lidar system produces dense measurements over snow or sea ice, with specular reflection over melt ponds allowing melt pond fractions to be retrieved with an accuracy better than 5 % (Landy et al., 2014). These observations allow us to evaluate how well RADARSAT-2 resolves f_pk of seasonally evolving sea ice coverage.

Aerial photographs of estimated f_p directly over the lidar site and the adjacent sea ice area away from land and open water were also obtained on 22 June 2012. The aerial photographs have a 0.22 m pixel resolution, covering 750 m by 750 m. In total, 123 aerial photographs of f_p were used, and a complete description of the dataset is provided in Scharien et al. (2014).

Finally, we made use of 8 d composite satellite observations of f_p obtained from the MODIS Arctic melt pond cover fraction dataset that has a spatial resolution of 12.5 km for the period of 2009–2011 (Rösel et al., 2012) and weekly sea ice area and stage of development observations obtained from the Canadian Ice Service Digital Archive (CISDA) regional ice charts for the period of 2009–2018 (Tivy et al., 2011).

2.2 Estimating f_pk from RADARSAT-2

RADARSAT-2 f_pk was determined using a modified approach to that described by Scharien et al. (2017). Their approach determines the second stage of the seasonal melt pond evolution cycle when f_p is at its peak (Eicken et al., 2002; Polashenski et al., 2012) using Sentinel-1 Extra Wide (EW) swath imagery obtained during April within the CAA. April corresponds to late winter sea ice conditions in the CAA, when sea ice growth has reached its maximum and spring warming has yet to begin. Their approach was developed by relating the winter period HH gamma nought (γ^∘) backscatter in decibels (dB) from Sentinel-1 to f_pk observations in 1.7 m spatial resolution GeoEye-1 imagery, from spatially coincident image segments that represented homogeneous FYI and MYI regions. The result was that γ^∘ can be converted to f_pk using the following equation:

$$\begin{array}{}\text{(1)}& f_{\mathrm{pk}}=-\mathrm{0.221}-\mathrm{0.041}\left(\mathit{\gamma }{}^{\circ }\right).\end{array}$$

In Eq. (1), γ^∘ was found to explain 73 % of the variability in f_pk (Scharien et al., 2017).

In this study, all the available HH polarization RADARSAT-2 imagery over the CAA in April from 2009–2018 (Table 1) was first calibrated to γ^∘ which minimizes the influence of incidence angle more so than with sigma nought (σ^∘) (Small, 2011). RADARSAT-2 images were then speckle filtered using a 5×5 Lee filter and spatially registered to a common map projection. Finally, γ^∘ was converted to f_pk by applying Eq. (1) to each RADARSAT-2 image. For each year, the corresponding RADARSAT-2 f_pk images in April were mosaicked together to cover the entire spatial domain of the CAA. Constructing a mosaic over a large region such as the CAA presents certain challenges with SAR imagery, particularly incidence angle variability. Even with the use of γ^∘, Scharien et al. (2017) found that because of varying incidence angles associated with different ScanSAR images that f_pk striping can still occur within the CAA in the mosaicked image. Our approach here was to average out incidence angle variability by taking advantage of a large amount of overlapping RADARSAT-2 imagery within the CAA (i.e. 90 to 159 images; Table 1) together with the fact that the majority of the sea ice in the CAA is landfast (immobile) during April, which results in a temporally stable f_pk for all April images. To produce a RADARSAT-2 f_pk mosaic within the CAA for each year, we calculated the mean f_pk for each overlapping pixel using all of each year's RADARSAT-2 April images that effectively helped to reduce f_pk striping across the CAA.

The root-mean-square error (RMSE) of f_pk based on Eq. (1) is 0.085 (Scharien et al., 2017). While calculating the mean f_pk of the overlapping image pixels helps reduce striping across the CAA, it also adds additional uncertainty and its effectiveness depends on the number of overlaps. In order to quantify the additional uncertainty (RMSE_R2), we used the mean and maximum standard deviation of RADARSAT-2 f_pk of all pixels within the CAA calculated from 2009–2018 (f_std) together with a range of pixel overlaps (n) in the following equation:

$$\begin{array}{}\text{(2)}& {\mathrm{RMSE}}_{R\mathrm{2}}=\left[\right(f_{\mathrm{std}}/n^{\mathrm{0.5}}{)}^{\mathrm{2}}+{\mathrm{0.085}}^{\mathrm{2}}{]}^{\mathrm{0.5}}.\end{array}$$

Since RADARSAT-2 imagery is acquired operationally, overlapping images vary interannually but pixel overlaps across the CAA were typically between 6–12. Figure 2 illustrates the RMSE_R2 values for a range of pixel overlaps using the 2009–2018 mean f_std value of 0.08 and the 2009–2018 maximum f_std value of 0.2. For the maximum f_std with pixel overlaps between 6–12 the RMSE_R2 ranges from 0.10–0.12.

Figure 2 The root-mean-square error of RADARSAT-2 peak melt pond fraction values (RMSE_R2) with increasing number of RADARSAT-2 pixel overlaps. The vertical dashed lines indicate the range of typical overlap from 2009–2018.

3 Results and discussion

3.1 RADARSAT-2 f_pk spatial and temporal variability from 2009–2018

The spatial distribution of mosaicked RADARSAT-2 f_pk and pre-melt season (i.e. April) and sea ice stage of developed conditions in the CAA for the 2009–2018 time period are shown in Figs. 3 and 4, respectively. Lower f_pk values are located primarily in the northern regions of the CAA (Queen Elizabeth Islands), Viscount Melville Sound, and the M'Clintock Channel where the majority of the CAA's MYI is typically found. The shallow bays and narrow channels located throughout the CAA exhibit high f_pk values and these regions are typically associated with smooth FYI whereas rougher ice regions (i.e. Gulf of Boothia) are associated with lower f_pk values. We should expect a lower f_pk over MYI regions compared to FYI regions (Grenfell and Perovich, 2004; Perovich and Polashenski, 2012), and indeed the overall spatial distribution of RADARSAT-2 f_pk is in excellent agreement with the spatial distribution of sea ice stage of development prior to the melt season for all years.

Figure 3 Spatial distribution of RADARSAT-2 peak melt pond fraction (f_pk) in the Canadian Arctic Archipelago from 2009–2018 (a–j).

Figure 4 Spatial distribution of sea ice stage of development (type) in the first week of April in the Canadian Arctic Archipelago for 2009–2018 (a–j).

Figure 5 Boxplot time series of RADARSAT-2 peak melt pond fraction (f_pk) and mean April multi-year ice (MYI) area in the Canadian Arctic Archipelago for 2009–2018. The solid blue line represents the mean (a). Spatial distribution of the RADARSAT-2 f_pk standard deviation from 2009–2018 (b).

Figure 5a shows the time series of RADARSAT-2 f_pk variability together with mean April MYI area in the CAA from 2009–2018. Over the 10-year record, the mean RADARSAT-2 f_pk was 0.47 and ranged from a low of 0.43 in 2009 to a high of 0.52 in 2013. The temporal variability in RADARSAT-2 f_pk is reflected in the variability of April MYI area within the CAA over the 10-year record with a statistically significant detrended correlation (R) of $R=-\mathrm{0}$.89. The RADARSAT-2 f_pk linkage with April MYI area is particularly evident from 2011 and 2012 which were very light sea ice years within the CAA whereby a considerable amount of the CAA's MYI area was lost during the summer melt season (Howell et al., 2013), and this resulted in 2012 and 2013 (i.e. the years following extreme melt) being the two highest RADARSAT-2 f_pk years from 2009–2018 (Fig. 3d–e). MYI area within in the CAA then increased following these light ice years and RADARSAT f_pk began to respond accordingly. In fact, there has always been a period of MYI recovery following light ice years with either MYI grown in situ and/or advected from Arctic Ocean into the CAA and gradually migrating to the CAA's southern regions (Howell et al., 2013). Figure 5b illustrates the standard deviation of RADARSAT-2 f_pk from 2009–2018 and spatially reflects the process of MYI flowing southward through the CAA as RADARSAT-2 f_pk was more variable in the MYI regions of the CAA compared to regions where FYI dominates the regional icescape.

What is interesting in Fig. 5a is that the mean RADARSAT-2 f_pk in 2009 was lower than all years from 2014–2018 (with the exception of 2016) despite the CAA containing less MYI area. In addition, 2017 and 2018 exhibited a larger spatial coverage of MYI compared to 2009 (Fig. 4a, i–j). We suggest that higher RADARSAT-2 f_pk in recent years is a result of Arctic Ocean MYI entering the CAA being younger and thinner than in 2009 (Howell and Brady, 2019) with smoother surface topography, thereby having a higher summer melt pond coverage (Landy et al., 2015). This seems to be particularly evident in the Viscount Melville Sound and M'Clintock Channels regions when comparing 2009 (Fig. 3a) with 2017 (Fig. 3i) and 2018 (Fig. 3j). Indeed, several studies have reported considerable decreases in the age and thickness of Arctic Ocean MYI north of the CAA in recent years (e.g. Kwok, 2018; Petty et al., 2020; Tschudi et al., 2020)

3.2 Comparison of RADARSAT-2 f_pk with Sentinel-1, in situ, and MODIS

Frequency distributions of RADARSAT-2 f_pk and Sentinel-1 f_pk from Scharien et al. (2017) in the CAA for 2016 and 2017 are shown in Fig. 6. Sentinel-1 appears to estimate more regions of lower f_pk compared to RADARSAT-2 which are typically associated with MYI. Whereas RADARSAT-2 estimates more regions of higher f_pk which are typically associated with smooth FYI. We consider these subtle differences to be primarily the result of taking the mean of all available April RADARSAT-2 imagery (Table 1) over all incidence angles in the CAA compared to only using images from Sentinel-1 within the CAA constrained to a certain incident angle range. As shown in Fig. 2, the uncertainty in RADARSAT-2 f_pk varies depending on the number of pixel overlaps (images). Overall, the f_pk distributions are in good agreement between both sensors.

Figure 6 Frequency distribution (%) of RADARSAT-2 peak melt pond fraction (f_pk) and Sentinel-1 f_pk from Scharien et al. (2017) in the Canadian Arctic Archipelago for 2016 and 2017.

The in situ evolution of f_p over FYI within the CAA acquired by Landy et al. (2014) allows us to place the RADARSAT-2 f_pk estimates within the melt pond stages of development classification system. Unfortunately, no MODIS f_p observations are located in close proximity to the in situ observations. The evolution of melt ponds on the surface of the sea ice has been classified into four distinct and consecutive stages. A brief description is provided here, and the reader is referred to Eicken et al. (2002) and Polashenski et al. (2012) for a more comprehensive description. In stage I, meltwater from snowmelt fills topographic depressions on the surface of the sea ice until the ponds reach their maximum areal extent. In stage II, melt pond coverage decreases due to horizontal water transport into macroscopic flaws and drainage through the ice. In stage III, the melt ponds typically drain through to the ocean and further changes in melt pond coverage depend on changes in surface topography and freeboard. Finally, in stage IV, melt ponds that survived the melt season refreeze and snow begins to accumulate on their surface.

Figure 7a compares the time series of the entire 100 m lidar melt pond fraction coincident with the f_pk determined from RADARSAT-2 at the coinciding pixels. For 2011, RADARSAT-2 f_pk corresponds to the end of stage I and beginning of stage II, thus providing a very good representation of the seasonal peak of the f_p, when the melt pond control on heat uptake and ice decay, through the ice–albedo feedback, is greatest. For 2012, RADARSAT-2 f_pk also corresponds to the end of stage I and beginning of stage II but is ∼ 0.2 lower than in situ f_p values. This is likely due to the short duration but very high maximum f_p of 0.78 in 2012 as Scharien et al. (2017) found that Eq. (1) sometimes underestimates very high f_p due to the low γ^∘ signal associated with very smooth FYI.

Figure 7 (a) Temporal evolution of observed melt pond fraction (f_p) and RADARSAT-2 peak melt pond fraction (f_pk) at in situ observation sites for 2011 (74.7229^∘ N, 95.1763^∘ W) and 2012 (74.7264^∘ N, 95.5772^∘ W). (b) Frequency distribution of RADARSAT-2 f_pk and aerial photograph f_p observations in Resolute Passage on 22 June 2012; the pink vertical link represents the mean lidar f_p on 22 June 2012.

To give spatial context beyond the single-point comparison at the lidar site, Fig. 7b shows the distribution of RADARSAT-2 f_pk and the f_p determined from aerial photo observations on 22 June 2012 near Resolute. The aerial photographs were acquired within 1 week of f_pk coverage being observed at the lidar site. The comparison was done by averaging all RADARSAT-2 pixels within each aerial photo (123 photos) which represents ∼ 861 samples. The mean aerial photograph f_p was 0.54 and RADARSAT-2 f_pk was 0.53 with an the RMSE of 0.10 and bias of 0. The distributions are in reasonably good agreement but RADARSAT-2 values are slightly narrower than the distribution of f_p from the aerial photographs. It is likely the RADARSAT-2 distribution is narrow on the left tail because our method captures peak pond coverage, and some of the regions photographed were before or after their seasonal peak. We attribute the narrow right tail to the documented underestimation of Eq. (1) from Scharien et al. (2017). However, it is notable that both RADARSAT-2 and the aerial photograph datasets capture the same bimodal f_p distribution, with the first mode around 0.4–0.5 characterizing rougher sea ice areas and the second mode around 0.7 capturing smooth flooded sea ice.

The seasonal time series of the 8 d composite MODIS f_p, the maximum seasonal MODIS f_p and the predicted RADARSAT-2 f_pk for 2009–2011 is shown in Fig. 8. MODIS f_p observations within the CAA indicate initial pond formation occurred in May for all years with f_pk reached in mid-July for 2009 and in early June for 2010 and 2011. Compared to the RADARSAT-2 f_pk values, the peak MODIS f_p is ∼ 0.20 smaller. RADARSAT-2 f_pk is higher on average than MODIS because the MODIS 8 d product does not represent f_pk. The MODIS f_p observations are determined weekly using 8 d composite image products that would include some melt pond formation and drainage processes prior to, and after, the seasonal peak. Moreover, MODIS f_p values are essentially aggregated from 500 m clear-sky pixels within a 12.5 km × 12.5 km grid cell (Rösel et al., 2012) and the 500 m spatial resolution may limit detection of smaller pond fractions. Furthermore, not all of the 500 m pixels within the 12.5 km × 12.5 km grid cell are likely to be at the same melt pond stage evolution. Finally, MODIS f_p observations give the time series of f_p; therefore even the highest seasonal estimated MODIS f_p is reduced because while some regions of the CAA are at their seasonal peak others are behind or ahead. To that end, we also calculated the maximum f_p from MODIS regardless of timing during the melt season, for each pixel, also shown in Fig. 8. These values more closely compare with the RADARSAT-2 f_pk but are still ∼ 0.05 smaller on average. Even the maximum f_p from MODIS is from an 8 d running mean of daily pond fraction estimates, so will underestimate the f_pk if the duration of peak ponding is < 8 d. However, the top whisker of the box plot of the maximum f_p from MODIS indicates that MODIS does capture some regions at peak during the 8 d time series. Although we are using MODIS f_p product to compare against our RADARSAT-2 f_pk estimates, Rösel et al. (2012) found that the MODIS f_p product also has errors up to ∼ 0.1. Overall, MODIS f_p estimates are more representative of the seasonal mean f_p rather than f_pk within the CAA.

Figure 8 Boxplots of the seasonal time series of MODIS melt pond fraction (f_p), the maximum seasonal MODIS f_p and RADARSAT-2 peak melt pond fraction (f_pk) for (a) 2009, (b) 2010 and (c) 2011. The solid blue line represents the mean.

3.3 Influence of RADARSAT-2 f_pk on summer sea ice area

In order to investigate whether RADARSAT-2 f_pk values can be used to provide predictive information for summer sea ice area within the CAA, we separated the CAA into numerous predefined subregions and then determined the detrended correlations between RADARSAT-2 f_pk and weekly sea ice area from the CISDA regional ice charts in each region over the period of 2009–2018. We tested each week from the start of June to the end of September. The strongest correlation, together with the corresponding week of occurrence are shown in Fig. 9a and b, respectively. All the strongest correlations are negative, indicating – as expected – that years with higher predicted f_pk values are associated with lower sea ice area at a later point in the summer. The higher f_pk the lower the area-averaged albedo of the ice surface leading to accelerated melt and lower sea ice concentrations (e.g. Perovich and Polashenski, 2012). There is considerable spatial variability in the strongest correlation across the CAA with relatively low correlations in the majority of the northern CAA and very low correlations in the eastern regions of the CAA. The regions of Kellet-Crozier ($R=-$0.92), Viscount Melville Sound ($R=-$0.73), M'Clintock Channel ($R=-$0.77), and Norwegian Bay ($R=-$0.78) all exhibit statistically significant correlations above the 95 % confidence level. In terms of timing for the statistically significant regions, RADARSAT-2 f_pk correlated the strongest to weekly sea ice area in August for all regions except Norwegian Bay (Fig. 9b). Compared to previous studies, the primary difference between using f_p values to predict summer sea ice conditions seems to be the timing of when the correlation is the strongest. Using simulated f_p values, Schröder et al. (2014) found the strongest correlation to September sea ice occurred for the May f_p. Liu et al. (2015) used observed MODIS f_p values and reported the strongest correlation to September sea ice in late July. Our findings suggest that methods such as these may be able to predict August sea ice area from f_pk simulations or observations with higher confidence than September ice area, at least in the CAA.

Figure 9 Spatial distribution of the (a) strongest detrended correlation (R) between RADARSAT-2 peak melt pond fraction (f_pk) and weekly sea ice area and (b) week of occurrence.

Why is the relationship stronger in some regions of the CAA and weaker in others? RADARSAT-2 f_pk values are determined from imagery acquired in April when ice conditions in the CAA are landfast (immobile) and do not evolve in concert with sea ice dynamics operating within the CAA. As a result, RADARSAT-2 f_pk values will not be spatially representative of the region's ice conditions when region-specific dynamic breakup processes dominate over thermodynamics (i.e. in situ melt). In other words, the origin of the ice in these regions during the summer melt season will not always be the same as in April (i.e. pre-melt) when the initial RADARSAT-2 f_pk value was determined. The time series of weekly detrended RADARSAT-2 f_pk and weekly sea ice area for selected regions within the CAA is shown in Fig. 10 and provides evidence for this regional dichotomy. In the Viscount Melville Sound and M'Clintock regions the correlations gradually get stronger, reaching a peak in August. These regions are known to be immobile and stagnant (e.g. Melling, 2002) with the majority of breakup taking place in September, which is when the relationship begins to degrade. The Kellet-Crozier is another stagnant region, which supports that in the absence of considerable ice dynamics the relationship between RADARSAT-2 f_pk and sea ice area is strong throughout the melt season. The time series in Penny Strait illustrates how the correlation gradually increases but when the region's dynamic breakup begins in July, ice is advected southward which degrades the correlation. This was also the case for many other regions in the northern CAA (not shown) as the flushing of sea ice southward from the northern CAA is a regular occurrence during the melt season (Melling, 2002; Howell et al., 2006). The low correlations in the southeastern regions of the CAA are also likely a function of ice dynamics as these regions of the CAA are known to be considerably influenced by currents and wind (Prinsenberg and Hamilton, 2005), and sea ice speed in Lancaster Sound and Barrow Strait can reach 10 km d⁻¹ (Agnew et al., 2008).

Figure 10 Time series of detrended correlations between RADARSAT-2 peak melt pond fraction (f_pk) and weekly sea ice area for selected regions in the Canadian Arctic Archipelago from June to September. The dashed black line is statistical significance at the 95 % confidence level.

The strong and statistically significant correlation in the Viscount Melville Sound region is encouraging as it is a key shipping region in the northern route of the Northwest Passage. To that end, we used linear regression to predict mean August sea ice area within Viscount Melville Sound with the detrended RADARSAT-2 f_pk values as a predictor. Figure 11 illustrates the results compared to observations (detrended) from the CISDA ice charts for 2009–2018. There is reasonable agreement between the predicted and observed sea ice area in the region with an RMSE of 18×10³ km² and an R²=0.44. The largest discrepancies occurred for 2013 and 2014 with the RADARSAT-2 f_pk model prediction resulting in too little sea ice area. Overall, within the Viscount Melville Sound region of CAA there is a period for which a significant statistical relationship exists between RADARSAT-2 f_pk and the summer ice area before sea ice dynamics degrades the relationship.

Figure 11 Predicated sea ice area anomalies (detrended) using RADARSAT-2 peak melt pond fraction (f_pk) and observed sea ice area anomalies (detrended) from the Canadian Ice Service Digital Archive (CISDA) ice charts in the Viscount Melville Sound region of the Canadian Arctic Archipelago, 2009–2018.

4 Conclusions

In this study we predicted and analyzed spring f_pk using RADARSAT-2 within the CAA from 2009–2018. The spatial variability in RADARSAT-2 f_pk was found to be excellent agreement with the spatial distribution of sea ice stage of development prior to the melt season as high (low) f_pk values were associated with FYI (MYI) types. The temporal variability of RADARSAT-2 f_pk over the 10-year record was significantly correlated to April MYI area, highlighting the importance of MYI within the CAA. RADARSAT-2 f_pk was found to be in good agreement with the f_pk maximum extent observed in situ for 2011 but was slightly lower than 2012 when peak f_p was very large (> 0.7). Compared to peak MODIS f_p values, RADARSAT-2 f_pk values were larger by ∼ 0.05. Based on our comparative analysis, RADARSAT-2 f_pk is more representative of peak f_p within the CAA compared to the MODIS 8 d product, which on average was found to underestimate f_pk by ∼ 0.2 and is more representative of the seasonal mean f_p. We also found excellent agreement between RADARSAT-2 and Sentinel-1, which suggests that combining both Sentinel-1 and the recently launched RADARSAT Constellation Mission (RCM) could facilitate pan-Arctic f_pk estimates. The RCM will also facilitate continued investigation of additional metrics that when combined with γ^∘ could further improve predicted f_pk.

The results presented in this study indicate that dynamically stable sea ice regions within the CAA exhibit a higher detrended correlation between RADARSAT-2 f_pk and summer sea ice area. Specifically, the strong and statistically significant de-trended correlation in the Viscount Melville Sound region demonstrates that RADARSAT-2 f_pk estimates are useful for providing predictive information about summer sea ice area in the northern route of the Northwest Passage. This information could find utility in constraining regional model simulations (e.g. Lemieux et al., 2016). Alternatively, it could be advantageous to exploit the high spatial resolution of SAR and investigate whether local-scale f_pk estimates could enhanced knowledge of summer ice conditions in northern communities (e.g. Cooley et al., 2020). Ultimately, imagery from RCM will ensure our time series of RADARSAT-2 f_pk estimates in the CAA will continue, gradually building statistics and facilitating the development of more robust statistical relationships in upcoming years.