The June Outlook for arctic sea ice in September 2010 shows reasonable arguments for either a modest increase or decrease in September 2010 sea ice extent compared to the last two years (5.4 million square kilometers in 2009 and 4.7 million square kilometers in 2008). However, it is important to note that the June 2010 Outlook indicates a continuation of the overall trend in long-term loss of summer arctic sea ice, with no indication that a return to historical levels of the 1980s/1990s will occur.
Reasoning for an increase in sea ice extent from recent years assumes that the current presence of extensive second- and third-year sea ice that we saw in winter 2009/2010 indicates a build-up of multi-year sea ice and a more stable ice pack. Reasoning for a decrease in sea ice extent from recent years, perhaps approaching new record-low minimum, focuses on the below-normal sea ice thickness overall, the thinning of sea ice in coastal seas, rotting of old multi-year sea ice, warm temperatures in April and May 2010, and the rapid loss of sea ice area seen during May.
With 16 responses, the June Outlook reflected both these arguments. The range of June Outlook estimates is 4.2 to 5.7 million square kilometers, with an additional estimate of 1.0 million square kilometers (Figure 1, below).
Six respondents, with estimates of 4.9–5.3 million square kilometers, represent a prediction for persistent conditions from 2008/2009.
The remaining estimates fall into "high" and "low" extent groupings: the low extent group with a range of 4.2 to 4.7 million square kilometers, representing a continued loss of sea ice extent compared to 2008/2009, and the high extent group of 5.4 to 5.7 million square kilometers, suggesting a return to the long-term trend for summer sea ice loss.
Individual responses were based on a range of methods: statistical, numerical models, comparison with previous observations and rates of ice loss, or composites of several approaches.
Pan-Arctic Full Outlook
OVERVIEW OF RESULTS
There are reasonable arguments for either a modest increase or a decrease in September 2010 sea ice extent compared to recent years. In the popular press, reports have spanned the range from a recovery from the recent period of record minimum extents, to predictions of a new record minimum this summer. With 16 responses, the June Outlook reflected a range of thinking about the summer 2010 sea ice season, but not these extremes on each end. Reasoning for a recovery from recent years assumes that the current presence of extensive second- and third-year sea ice in winter 2009/2010 might indicate a build-up of the multi-year (more stable) ice pack. Reasoning for a new record minimum focuses on the below-normal ice thickness overall, the thinning of sea ice in coastal seas, rotting of old multi-year sea ice, and the rapid loss of sea ice area seen during May.
It is important to note for context, however, that all 2010 estimates are well below the 1979–2007 September climatological mean of 6.7 million square kilometers, and all indicate a continuation of a long-term trend in summer sea ice loss.
The range of June Outlook estimates is from 4.2 to 5.7 million square kilometers, with one additional outlook estimate of 1.0 million square kilometers (Figure 1). Six respondents, with estimates of 4.9–5.3 million square kilometers, represent a prediction for persistent conditions from 2008/2009. The remaining estimates fall into "high" and "low" extent groupings: the low extent group with a range of 4.2 to 4.7 million square kilometers, representing a continued loss of sea ice extent compared to 2008/2009, and the high extent group of 5.4 to 5.7 million square kilometers, suggesting a return to the long-term trend for summer sea ice loss. Individual responses were based on a range of methods: statistical, numerical models, comparison with previous observations and rates of ice loss, or composites of several approaches. Of interest is that the three numerical modeling methods (Kauker et al., Wu and Grumbine, and Zhang) represent examples of the high, persistence, and low outlooks. We would also like to draw attention to a recent paper by Holland et al. (see PDF at the bottom of the report) that supports the idea that there is some skill to summer seasonal forecasts.
Where provided, the uncertainty estimates are close to 0.5 million square kilometers. In the 2009 Outlook, most estimates overlapped each other when their uncertainty ranges were considered. This year there is significant separation between the high and low estimates. The goal of the Outlook is not firm forecasts, but to promote a discussion of the physics and factors of summer sea ice loss. In 2010, we are pleased at the range of discussion and thank the contributors for their efforts.
LATE SPRING 2010 CONDITIONS
Regarding initial conditions for Spring 2010, Figure 2 by Maslanik and others shows maps of sea ice classes derived from sea ice age for April 2010 and 2009. Their approach to determining sea-ice age is based on tracking of sea ice using satellite imagery. Dark blue regions are areas of first-year sea ice. In general, the maps show that there is slightly more third-year sea ice this April than in 2009 (green). An interesting feature in both images is the tongue of old sea ice (red) extending into the southern Beaufort Sea. This feature was observed in the past three years, and airborne and ground-based surveys out of northern Alaska indicate that this is in fact old ice of well over 3 m thickness. Finding such old ice so far south in recent years is consistent with observations for increased mobility of sea ice, even in the central Arctic. The movement of old, thick sea ice to lower latitudes has a two-fold impact. Regionally, it can help delay sea ice loss, but on a pan-arctic scale it enhances overall ice melt and ice volume reduction, as these old floes melt faster at lower latitudes. The other major feature of the 2010 sea ice age map is the extensive areas of second- and third-year sea ice. Will much of this ice remain in the central basin to rebuild the area of old, perennial sea ice over the coming years?
It is unfortunate that maps of multi-year sea ice distribution for 2010 derived from QuikSCAT (provided by Nghiem) are no longer available to compare with sea ice age calculations, as in previous years. A composite product based on active (ASCAT aboard the EUMETSAT METOP satellite) and passive (SSM/I) microwave data with advanced atmospheric correction of raw data (Figure 3) shows a range of sea ice types. Major features are southwestward advection of older sea ice into the Beaufort Gyre and southwestward advection of different sea ice types toward Fram Strait in the Transpolar Drift.
Much has been made of the fact that the April 2010 sea ice extent data released by the National Snow and Ice Data Center (NSIDC) were near the long-term climatological average (Figure 4). This April increase, however, is primarily in the marginal seas of the Arctic, especially the Bering Sea, so it is unclear whether this late winter/early spring advance will have any effect on summer ice conditions. Ground-based observations in the Bering Sea indicate below-normal ice thickness this spring (see the Sea Ice for Walrus Outlook). Further, Figure 4 shows a major loss of sea ice extent through May; contributions to the loss were especially important from the Barents Sea and northern Baffin Bay (Figure 5). Such loss can be related to warm temperatures throughout the Arctic during May (Figure 6). Given the hint of a sea ice-free region near the New Siberian Islands (off the Siberian Coast) in Figure 5 and the temperature maximum in Figure 6, one might suggest an early sea ice melt along the Siberian coast this summer. Overall, the curve shown in Figure 4 is commensurate with the notion that a thinner arctic ice cover that is more mobile can lead to greater seasonal and interannual variability, with a potential loss in predictability.
Key statements from the individual Outlook contributions and PDFs of each contribution are below.
We should have a very interesting season this year, so stay tuned for next month's Outlook in July!
KEY STATEMENTS FROM INDIVIDUAL OUTLOOKS
Name (Organization of First Author); Estimate in Million Square Kilometers; Method
Ordered from Greatest to Least
Definitions of the different types of methods can be found in our Sea Ice Outlook glossary. PDFs of the individual contributions are at the bottom of this page.
Tivy (University of Alaska Fairbanks); 5.7 Million Square Kilometers; Statistical
This method is based on a simple regression where the predictor is the previous summer (May/June/July) sea surface temperature (SST) in the North Atlantic and North Pacific oceans near the marginal ice zone. Warmer than normal SST is associated with a reduction in ice extent; colder than normal SST is associated with an increase in ice extent.
Stroeve et al. (National Snow and Ice Data Center, NSIDC); 5.5 Million Square Kilometers; Statistical
Note: this value is an average of two estimates:
- 5.76 million square km based on the mean age- and latitude- corrected ice survival rates for 2000-2009
- 5.21 million square km based on the mean age- and latitude- corrected ice survival rates for 2005-2009.
We are using the same approach as last year: applying the survival fraction of ice of different ages determined from past seasons to the observed distribution of ice ages at the beginning of the melt season. It appears that a new record low will not be reached this year if the 2010 survival rates are within the range of historical ice survival rates. This is in part because there is more 2nd and 3rd year ice at the start of 2010 than has been seen the last few years. Also, winter extent was larger in 2010 than in previous years. If the 2010 survival rates are similar to 2007, however, the September 2010 extent will rival what was observed in 2007 (4.31 versus 4.13 million square km).
Pokrovsky (Main Geophysical Observatory, Russia); 5.5 Million Square Kilometers; Heuristic and Statistical
Future sea ice extent estimates in the Arctic might be obtained by joint analysis of time series of three climate indicators: AMO, PDO, AO for last thirty years. I used a modified regression analysis approach. This year is a cold one in this region (Pacific) due to the north wind domination. That explains that the sea ice extent in the Pacific sector of the Arctic exceeds climate (20th century) magnitudes. Now in the eastern part of the Barents Sea there is a significant area of the sea surface free of ice. The September sea ice extent anomaly should demonstrate tendencies of more ice in the Pacific and less ice in the Atlantic sectors. But, in general, the sea ice extent should attain a higher value than last year.
Kauker et al. (Alfred Wegener Institute for Polar and Marine Research); 5.4 Million Square Kilometers; Numerical Modeling
Note: this value is an average of two estimates:
- The Ensemble I mean value is 5.61 million km2 (bias included). The standard deviation of Ensemble I is 0.41 million km2 (2008: 0.55; 2009: 0.40).
- The Ensemble II mean of 5.19 million km2 is somewhat lower than the mean of Ensemble I (note that the optimization increases the predicted mean by about 0.07 million km2 compared to the uncorrected Ensemble I mean of 5.12 million km2). As for Ensemble I the standard deviation of Ensemble II is 0.41 million km2.
The ensemble prediction of September 2010 looks similar to the situation before 2007. A comparison of the modeled ice thickness on 1 June 2007, 2008, and 2009, and the initial ice thickness on 28 May 2010 reveals considerably larger ice thickness mainly in the East Siberian Sea, north of the East Siberian Sea, and in the vicinity of the North Pole in 2010 compared to 2007–2009.
Rigor et al. (Polar Science Center, University of Washington); 5.4 Million Square Kilometers; Heuristic
This estimate is based on the prior winter Arctic Oscillation (AO) conditions, and the spatial distribution of the sea ice of different ages as estimated from a Drift-age Model (DM), which combines buoy drift and retrievals of sea ice drift from satellites (Rigor and Wallace, 2004, updated). The DM model has been validated using independent estimates of ice type from QuikSCAT (e.g., Nghiem et al. 2007) and in situ observations of ice thickness from submarines, electromagnetic sensors, etc. (e.g., Haas et al. 2008; Rigor 2005).
Untersteiner and Morison (University of Washington); 5.3 Million Square Kilometers; Heuristic
Detailed outlook contribution not provided. (Authors: Norbert Untersteiner and James Morison, University of Washington).
Wellman (Princeton Consultants); 5.1 Million Square Kilometers; Statistical
The May melt has been the fastest in the satellite record. Arctic Oscillation has tended towards a state with lower than average ice export through Fram Strait, but that may be moderating.
Polar Science Weekend; 5.1 Million Square Kilometers; Heuristic
The Polar Science Weekend is an annual public outreach event organized by the Seattle Pacific Science Center and the University of Washington Polar Science Center. Members of the public were invited to make a prediction for this September's sea ice extent. We had a total of N = 60 guesses from about six hours of discussions. The mean was 5.1 million square kilometers and the standard deviation was 2.15 million sq km. The mean is quite near that predicted by the trend line (5.15 +/- 0.57 million sq km), but the spread is greater.
Wu and Grumbine (National Oceanographic and Atmospheric Administration, NOAA); 4.9 Million Square Kilometers; Statistical and Modeling
Note: this value is an average of two estimates:
- Model Prediction for September 2010 average ice extent: 5.13 million km2, standard deviation 0.25 million km2
- Statistical: 4.78 million km2, 0.45 million km2 sdev
Two approaches are used, statistical and modeling. The statistical approach continues from Grumbine in 2009. The approach is to consider the growth of open water as proceeding according to a population growth (positive feedback of more open water leading to more open water) with a constraint. The model prediction is based on the coupled Air-Sea-Ice Climate Forecast System (CFS) at NCEP. These predictions are based on the CFS Reanalysis and Reforecast model. An ensemble of 24 forecasts were made to provide estimates of mean and model variability. At this lead time, the model shows a consistent high bias in its forecasts of September ice extent. We have, therefore, attempted bias correction.
Arbetter et al. (North American Ice Service/National Ice Center); 4.9 Million Square Kilometers; Statistical/Heuristic
Despite the reasonably large current extent (14.665 million km2) and compact concentration (12.461 million km2) in late April, the projected extent for mid-September is another near-record low (4.852 million km2). The most compact ice is on the Canadian side of the Arctic Ocean, while the pack on the Siberian side is diffuse (1-3/10th concentration).
Gauthier et al. (Canadian Ice Service); 4.9 Million Square Kilometers; Heuristic/Empirical
The Canadian Ice Service (CIS) is predicting the minimum arctic sea ice extent to be less than 5 million square kilometres in September 2010. A value equal to or slightly greater than the average extent observed in September 2008 is expected. This value (4.7 ≤ x
|Stroeve et al.||175.61 KB|
|Kauker et al.||1.96 MB|
|Rigor et al.||4.33 MB|
|Polar Science Weekend||1.84 MB|
|Wu and Grumbine||38.49 KB|
|Arbetter et al.||315.74 KB|
|Gauthier et al.||176.73 KB|
|Kaleschke and Spreen||371.78 KB|
|Lindsay and Zhang||283.14 KB|
|Reference Article - Holland et al.||687.4 KB|
INTRODUCTION AND OVERVIEW
We received 9 contributions to the June Regional Outlook report. The regional contributions this month suggest that ice conditions will be below normal relative to the past two to three decades, and that they may potentially rival the 2007 record minimum if multi-year ice at lower latitudes melts back early in the season.
Regional Outlook contributions can help shed light on the uncertainties associated with the estimates in the Pan-Arctic Outlook by providing more detail at the regional scale. In addition, the persistence of regional features (such as above-normal multi-year ice concentrations compared to recent years in the southern Beaufort and Chukchi Seas) and geographic constraints (such as the confinement of ice in the Canadian Arctic Archipelago) can lend predictive skill to forecasts.
In contrast with the June Pan-Arctic Outlook estimates, which show a greater range of predicted values, the Regional Outlook contributions are more in line with each other than last year. For example, in the Northwest Passage, empirical, statistical, and numerical modeling approaches yield a consistent picture of remnant ice and lack of completely open conditions, in contrast with last year when two methods diverged in the predictions.
Forecasts based on the statistics of past summer ice for the different regions provide us with a framework to evaluate whether the ice cover behaves in a similar fashion as in the past or whether the present year is anomalous. For example, lingering multi-year ice not accounted for in some of the models may delay the retreat of the ice edge towards the North. Ice retreat depends strongly on the amount of solar heating of the surface ice and ocean; clear sky conditions early in the melt season can go a long ways towards helping ice retreat later in the season, so this is one of the variables that will be monitored closely (e.g., in the Barrow coastal sea ice break-up forecast, see below). Ship observations will also help evaluate this important factor this season.
Definitions of sea ice terminology can be found in our Sea Ice Outlook Glossary.
SUMMARY LIST OF REGIONAL FORECASTS
- Howell and Agnew – closed
- Zhang – closed
- Tivy – closed
- Tivy – early opening
Beaufort / Chukchi Seas:
- Maslanik, Tivy, Pokrovsky – below-normal ice area overall, but high concentration may remain along the coast in the southern Beaufort and Chukchi Seas
- Zhang – ice edge further north than in 2009
- Petrich and others – too early in the season to use 16-day weather forecasts to predict fast ice break-up around Barrow
- Lindsay and Zhang – below-normal ice cover, ice characteristics governing navigation comparable to 2007 and 2009
East Siberian / Laptev Seas:
- Zhang, Tivy – below-normal ice cover but greater than 2009
Barents / Kara / Greenland Seas:
- Zhang – ice edge further north than in 2009
- Gerland – less ice around Svalbard compared to 2009
- Pokrovsky – below-normal ice cover
- Tivy (Greenland Sea) – below-normal ice cover and less than 2009
- Tivy (Kara and Barents Seas) – below-normal ice cover but comparable to 2009
Canadian Arctic Archipelago:
- Tivy – below-normal ice cover, slight increase from last year
- Gudmandsen - unstable ice barrier in Nares Strait, likely more inflow of old ice into Baffin Bay than last year
BEAUFORT / CHUKCHI SEAS
An estimate of ice age at the end of April is shown in Figure 1. This image was provided by Maslanik and is a University of Colorado satellite-derived sea ice age product. The lobe of old ice extending through the Beaufort Sea and into the Chukchi Sea is estimated to be >5 years old. Aerial thickness surveys and ground-based sampling in the western Beaufort and eastern Chukchi Sea in April this year indicate that level ice in the region is well over 3 m thick and strong (due to low salt content; www.sizonet.org). The outlook submitted by Maslanik based on this distribution of multi-year ice predicts that the lobe of multi-year ice will persist later into the melt season and the first-year ice to the north will melt out earlier, yielding a “semi polynya” of open water/low concentration ice partially surrounded by multi-year ice into late summer.
A statistical forecast, based on canonical correlation analysis and using fall surface air temperature anomalies over the Beaufort Sea as the predictor for September sea ice concentration (Figure 2) submitted by Tivy, shows a high probability of above-normal ice concentrations in the southern Beaufort Sea and a high probability of below-normal ice concentrations slightly north of the region of first-year ice in Figure 1. This is to some degree consistent with the Maslanik outlook based on the distribution of multi-year ice.
The outlook submitted by Pokrovsky also points to more ice in the southern Beaufort Sea compared to previous years. A positive winter Pacific Decadal Oscillation (PDO) index and associated low sea-surface temperature (SST) anomalies in the North Pacific are expected to slow summer sea ice melt in the coastal Beaufort and Chukchi Seas due to colder than normal Pacific water inflow.
In contrast, the ensemble forecast from a coupled ice-ocean model submitted by Zhang shows the September ice edge further north than in 2009 (Figure 1, East Siberian / Laptev Sea region). The ice-ocean model is initialized with satellite estimates of ice concentration and model-simulated ice thickness and ocean fields, and is forced by the atmospheric fields from 2003 to 2009. Hence, as a probabilistic estimate, some of the empirical evidence entering into the other outlooks may not impact ensemble simulations to the same degree.
A prediction for the date of break-up of fast ice along the coast of Barrow was submitted by Petrich and others. Their prediction is based on the quantity of incoming solar radiation and uses 16-day forecasts from a numerical weather prediction model (WRF). The most recent 16-day weather forecast does not yet reach far enough into the future to predict break-up, but the forecast will be updated for the July Outlook. However, observations and forecasts suggest a slightly earlier break-up than normal.
Lindsay and Zhang (see also their website at http://psc.apl.washington.edu/lindsay/Alaskan_summer_ice.html) submitted a forecast for the 11 shipping-specific sea ice parameters near Barrow, Alaska that are included in the "Seasonal Outlook for North American Arctic Waters'" issued by the North American Ice Service. Their prediction uses output from the Zhang coupled ice-ocean model as predictors in a statistical regression-based model. All 11 predictions point to below-normal ice conditions. There is a 50% or greater chance of record minimum ice conditions for 7 of the 11 parameters, so a new record minimum in some of them is likely. They predict a July 3rd opening of the sea route into Prudhoe Bay (less than 5-tenths ice) and 95 days of ice-free conditions along the shipping route.
EAST SIBERIAN / LAPTEV SEAS
Sea ice in the East Siberian and Laptev Seas is predominantly first-year ice (Beaufort Sea Region, Figure 1). The ensemble prediction from a coupled ice-ocean model submitted by Zhang shows considerably more ice in the East Siberian Sea compared to 2009. The ice-ocean model is initialized with satellite estimates of ice concentration, model-simulated ice thickness, and ocean temperature/salinity fields and is forced by the atmospheric fields from 2003 to 2009. This is somewhat consistent with the two statistical forecasts submitted by Tivy. The predicted September sea ice area in the East Siberian and Laptev Seas, from a simple regression model using summer (Aug-Sep-Oct) sea surface temperatures in the North Atlantic as the predictor, is below normal but greater than in 2009. The probabilistic forecast shown in Figure 5, using winter (Jan-Feb-Mar) North Atlantic sea surface temperature as the predictor, predicts a high probability of above-normal ice concentrations along the coast.
BARENTS / KARA / GREENLAND SEAS
Barents and Kara Seas are predominantly first-year ice, whereas significant quantities of multi-year ice are present along the coast of the Greenland Sea as it is being exported from the Arctic Ocean (Figure 6, Beaufort / Chukchi Sea region). The May 2010 ice edge shown in red in Figure 6 is compared against the location of the ice edge in the past three years. This year, the ice edge is the furthest south in the south Greenland Sea and the furthest north in the north Greenland Sea, as compared to the past three years. In the Barents Sea, between Svalbard and Novaya Zemlya, the ice edge this year is the furthest, while more ice is visible in the southeastern Barents Sea.
Gerland and others speculate that it is the influence of the Atlantic water that controls the position of the ice edge. This is in agreement with Pokrovsky who predicts below-normal ice cover in the Atlantic sector of the Arctic Ocean, based on observed warm spring sea surface temperature anomalies in the northeast Atlantic Ocean.
The ensemble model prediction submitted by Zhang shows a complete retreat of the ice edge in the Barents, Kara, and Greenland Seas (Figure 1, East Siberian / Laptev Sea region) and the ice extent is considerably less than in 2009. The ice-ocean model is initialized with satellite estimates of ice concentration and model-simulated ice thickness and ocean fields, and is forced by the atmospheric fields from 2003 to 2009.
Gerland and others also point to a reduction in ice extent compared to last year due, in part, to above-normal sea surface temperature anomalies in the North Atlantic. A regression-based forecast for September ice extent around Svalbard (an area extending from 72–85N and 0–40E), which uses May sea surface temperatures, the March index of the Arctic Oscillation, and April ice conditions as predictors, yielded a mean ice extent in September 2010 of 255,788 square kilometers around Svalbard. This forecast estimate ranks 9th in the record of September ice extents (1967–present).
The regression-based forecast submitted by Tivy for the Greenland Sea (Figure 7) also shows a reduction in September sea ice area compared to 2009 based on fall (Sep-Oct-Nov) sea surface temperature anomalies in the North Atlantic. However, the regression-based forecast for the Barents / Kara Seas is not consistent with the other outlooks (Figure 7). The September ice area is predicted to be comparable to 2009 based on winter (Jan-Feb-Mar) sea level pressure anomalies over the Kara and Laptev Seas.
CANADIAN ARCTIC ARCHIPELAGO (CAA)
In the pre-season outlook submitted by Tivy, a simple regression model is used to predict the September sea ice area in the Canadian Arctic Archipelago (Figure 8). The predictor is multi-year ice anomalies in the Beaufort Sea along the coast of the islands the previous summer (14-month lead), which is an area of ice import into the CAA. The model predicts below-normal ice area compared to the long-term mean and only a slight increase compared to 2009. Spatial forecasts from both a statistical model submitted by Tivy and a dynamical model submitted by Zhang are discussed in the following section on the Northwest Passage.
NORTHWEST PASSAGE (NWP)
At the beginning of May, multi-year ice concentrations in the Western Parry Channel region of the Northwest Passage (Figure 9) were only slightly above the record low in 1999 and less than in 2007, when the region cleared for the first time during the satellite era. However, all three outlooks for this region indicate that some ice will remain in the NWP in September.
Howell and Agnew make the point that the ice in the NWP and the surrounding regions is mobile during the summer, and as a result the spatial distribution of multi-year ice at the beginning of the season can be used to predict whether or not the NWP could be ice-free in September. This year, Howell and Agnew suggest that high concentrations of multi-year ice in M’Clintock Channel (Figure 9) will likely delay breakup in this region and keep multi-year ice concentrations high in Western Parry Channel as ice is transported southward during the summer.
In an ensemble prediction from an ice-ocean model, Zhang shows ice remaining in the Eastern Parry Channel in September (Figure 10). The ice-ocean model is initialized with satellite estimates of ice concentration and model simulated ice thickness and ocean fields and is forced by the atmospheric fields from 2003 to 2009.
Using a statistical model based on winter near-global sea surface temperatures, Tivy shows high concentrations of ice remaining throughout the NWP region. The model, which uses Canonical Correlation Analysis to describe linear relationships between predictors and predictands, is used in both Canada and the United States for seasonal predictions of temperature and precipitation.
Gudmandsen provides a more detailed analysis of ice export from the Lincoln Sea north of the Canadian Arctic Archipelago into Baffin Bay, through Nares Strait between Greenland and Canada. In contrast with last year he notes that episodes of ice export, stopped by formation of temporary ice bridges, led to an influx of multiyear ice into Nares Strait, which will slow down melt and ice retreat later in the season.
HUDSON BAY AND THE ARCTIC BRIDGE
By September, the Hudson Bay region is generally ice-free. Tivy submitted an outlook for July ice concentration anomalies in Hudson Bay. July is the main transition month between ice-covered and ice-free conditions and it is a month of particular interest, since the mean opening date for the shipping route to Churchill is July 25th. Using a statistical model based on canonical correlation analysis with fall sea surface temperature anomalies in the North Atlantic as the main predictor, Tivy shows below-normal ice concentrations throughout most of the region (Figure 12), which suggests an earlier-than-normal opening of the shipping season.