Improved CMIP6 Projections of Hudson Bay Navigability Using Bias Correction and Route Continuity

Soleymani, Armina; Crawford, Alex; Mudryk, Lawrence; Wang, Feiyue

doi:10.5194/egusphere-2026-4196

Preprints

https://doi.org/10.5194/egusphere-2026-4196

Preprints

23 Jul 2026

| 23 Jul 2026

Improved CMIP6 Projections of Hudson Bay Navigability Using Bias Correction and Route Continuity

Armina Soleymani, Alex Crawford, Lawrence Mudryk, and Feiyue Wang

Abstract. Hudson Bay’s seasonal sea ice limits maritime access to the Port of Churchill. This study improves on past efforts to project future navigability of shipping routes from Churchill by correcting for substantial model biases, better combining SIC and SIT into calculations, and incorporating a novel path-finding approach that requires continuous navigable routes (rather than relying on spatial domain averages). We evaluate future navigability for Type E vessels under two criteria: ice-free and low-risk (applying the Polar Operational Limit Assessment Risk Indexing System (POLARIS)). Additionally, we examine both the global shipping route (Churchill–Labrador Sea) and regional connectivity within northwestern Hudson Bay (Churchill–Chesterfield Inlet). In the 1980s, median navigable periods (NPs) for the Churchill–Labrador route were 102 days (ice-free) and 123 days (low-risk), compared to 106 and 123 days, respectively, for the Churchill–Chesterfield route. By the 2010s, these periods increased by approximately 4 weeks. Various methods of employing POLARIS and climate models yield a range of baseline NPs but consistent trends. Each additional 0.5 °C of global warming extends NPs by 13–18 days up to 3.5 °C, beyond which the rate of increase accelerates. Under 5 °C warming (late 21st century in SSP5-8.5), the Churchill-Labrador Sea route would remain ice-free for up to 265 days (8.8 months), representing and increases of 137 days relative to present conditions. Under low-risk conditions, navigability would reach 304 days (10.1 months). This demonstrates that year-round shipping would require ice-strengthened vessels throughout the 21st century even with a high-end warming scenario.

Received: 14 Jul 2026 – Discussion started: 23 Jul 2026

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Preprint (PDF, 6724 KB)

Supplement (6441 KB)

Download & links

Armina Soleymani, Alex Crawford, Lawrence Mudryk, and Feiyue Wang

Status: final response (author comments only)

CC1:
'Comment on egusphere-2026-4196', Minghu Ding, 19 Aug 2026
This study focuses on the limitations that seasonal sea ice in Hudson Bay imposes on navigable access to the Port of Churchill. By correcting climate model biases and introducing a novel path-finding approach that requires continuous navigable routes, it provides a relatively detailed projection of future shipping navigable periods. The navigability of Type E vessels is evaluated under two criteria. While the study has certain value in terms of methodological improvements and scenario analysis, several issues still require clarification or improvement.
Major comments:
The goal of the paper is to evaluate different navigability assessment methods and simulation bias-correction methods. Therefore, the authors should more clearly and objectively quantify the degree of improvement that different methods bring to the results. Although Figs. 8 and 9 provide information, it is very difficult to precisely read the NP values for specific years or specific warming levels and the differences among methods from the figures. I would suggest the addition of a quantitative table to present the results more clearly and objectively. For example, in the historical period, which method or combination of methods yields NP closest to observations? By how much is the multi-model ensemble bias reduced? What are the projected results of each method?

2.3: “shipping routes were represented as a 200 km-wide zone”. A 200 km resolution is overly coarse, and the choice of this width lacks sufficient justification. Many CMIP6 models have resolutions around 100 km. The 200 km choice sacrifices specific precision, though it remains meaningful for regional accessibility. I would suggest that the models should first be interpolated to a uniform resolution (e.g., the same 25 km as the observations) before calculations are performed, and that the sensitivity of the results to resolution be examined.

When the authors apply sea-ice data from different models to the same 200 km-wide shipping zone for assessment, do they first interpolate all model data onto a uniform spatial grid? If no uniform interpolation is performed, then a considerable portion of the inter-model differences shown in Figs. 5–7 may arise from differences in model spatial resolution. For example, lower-resolution models may be unable to resolve the narrow islands and channels in Hudson Strait, thereby systematically overestimating navigability.

Type E (pleasure craft) vessels are selected as the assessment object. Type E vessels are defined as “non-ice-strengthened” in the POLARIS system. Are Type E vessels the main vessel type currently or in the future operating at the Port of Churchill? Using the Type E vessels as the baseline will systematically underestimate future navigable periods. It is recommended that the authors investigate the main vessel types operating at the port before conducting the assessment, and/or add discussion to clearly state the rationale and limitations of adopting Type E vessels.

Minor comments:
The authors state in the introduction that “the southern route is currently the preferred operational route”. But in the manuscript, they set up two routes. These two routes differ completely in destination distance, so a direct comparison between them lacks persuasiveness. And, what is the purpose of presenting this comparison? This should be clearly explained in the text. If there is no special objective, results for routes within the region are suggested to be merged for analysis as SI.

Line 190:

2.6: The definitions of OD, CD and NP are very complex and obscurely worded. How is the start day of the “sea ice year dynamically” determined? Specifically, which day is it? The text mentions “zones where the timing of maximum ice extent”, but provides no operable algorithm. The determination of OD and CD involves concepts such as “the median day for which a route is open/close”, “last day the route was closed”, and “the first closed day”, making it difficult for readers to intuitively understand their physical meaning. In particular, when NP exceeds 365 days, OD and CD become “undefined” and are specially treated, introducing an artificial discontinuity. Since the assessment is oriented toward the full year, it is recommended to define OD, CD and NP based on the simplest and most intuitive calendar year.

Line 275: “Model skill and projections are also evaluated using percentiles (median, 10th, 90th) against a reference dataset.” How exactly is this done?

Line 285: Are the weights independent of time? Then are the observations and simulations using climatological means? This should be clarified.

Line 350: “the opening day exhibits greater variability…2000s.” The authors attribute this to internal variability. In reality there is also the cause of fragmented drift ice, see Cook et al. (2024): Cook, A.J., Dawson, J., Howell, S.E.L. et al. Sea ice choke points reduce the length of the shipping season in the Northwest Passage. Commun Earth Environ 5, 362 (2024). https://doi.org/10.1038/s43247-024-01477-6
Citation: https://doi.org/10.5194/egusphere-2026-4196-CC1
RC1:
'Comment on egusphere-2026-4196', Minghu Ding, 19 Aug 2026
This study focuses on the limitations that seasonal sea ice in Hudson Bay imposes on navigable access to the Port of Churchill. By correcting climate model biases and introducing a novel path-finding approach that requires continuous navigable routes, it provides a relatively detailed projection of future shipping navigable periods. The navigability of Type E vessels is evaluated under two criteria. While the study has certain value in terms of methodological improvements and scenario analysis, several issues still require clarification or improvement.
Major comments:
The goal of the paper is to evaluate different navigability assessment methods and simulation bias-correction methods. Therefore, the authors should more clearly and objectively quantify the degree of improvement that different methods bring to the results. Although Figs. 8 and 9 provide information, it is very difficult to precisely read the NP values for specific years or specific warming levels and the differences among methods from the figures. I would suggest the addition of a quantitative table to present the results more clearly and objectively. For example, in the historical period, which method or combination of methods yields NP closest to observations? By how much is the multi-model ensemble bias reduced? What are the projected results of each method?

2.3: “shipping routes were represented as a 200 km-wide zone”. A 200 km resolution is overly coarse, and the choice of this width lacks sufficient justification. Many CMIP6 models have resolutions around 100 km. The 200 km choice sacrifices specific precision, though it remains meaningful for regional accessibility. I would suggest that the models should first be interpolated to a uniform resolution (e.g., the same 25 km as the observations) before calculations are performed, and that the sensitivity of the results to resolution be examined.

When the authors apply sea-ice data from different models to the same 200 km-wide shipping zone for assessment, do they first interpolate all model data onto a uniform spatial grid? If no uniform interpolation is performed, then a considerable portion of the inter-model differences shown in Figs. 5–7 may arise from differences in model spatial resolution. For example, lower-resolution models may be unable to resolve the narrow islands and channels in Hudson Strait, thereby systematically overestimating navigability.

Type E (pleasure craft) vessels are selected as the assessment object. Type E vessels are defined as “non-ice-strengthened” in the POLARIS system. Are Type E vessels the main vessel type currently or in the future operating at the Port of Churchill? Using the Type E vessels as the baseline will systematically underestimate future navigable periods. It is recommended that the authors investigate the main vessel types operating at the port before conducting the assessment, and/or add discussion to clearly state the rationale and limitations of adopting Type E vessels.

Minor comments:
The authors state in the introduction that “the southern route is currently the preferred operational route”. But in the manuscript, they set up two routes. These two routes differ completely in destination distance, so a direct comparison between them lacks persuasiveness. And, what is the purpose of presenting this comparison? This should be clearly explained in the text. If there is no special objective, results for routes within the region are suggested to be merged for analysis as SI.

Line 190:

2.6: The definitions of OD, CD and NP are very complex and obscurely worded. How is the start day of the “sea ice year dynamically” determined? Specifically, which day is it? The text mentions “zones where the timing of maximum ice extent”, but provides no operable algorithm. The determination of OD and CD involves concepts such as “the median day for which a route is open/close”, “last day the route was closed”, and “the first closed day”, making it difficult for readers to intuitively understand their physical meaning. In particular, when NP exceeds 365 days, OD and CD become “undefined” and are specially treated, introducing an artificial discontinuity. Since the assessment is oriented toward the full year, it is recommended to define OD, CD and NP based on the simplest and most intuitive calendar year.

Line 275: “Model skill and projections are also evaluated using percentiles (median, 10th, 90th) against a reference dataset.” How exactly is this done?

Line 285: Are the weights independent of time? Then are the observations and simulations using climatological means? This should be clarified.

Line 350: “the opening day exhibits greater variability…2000s.” The authors attribute this to internal variability. In reality there is also the cause of fragmented drift ice, see Cook et al. (2024): Cook, A.J., Dawson, J., Howell, S.E.L. et al. Sea ice choke points reduce the length of the shipping season in the Northwest Passage. Commun Earth Environ 5, 362 (2024). https://doi.org/10.1038/s43247-024-01477-6
Citation: https://doi.org/10.5194/egusphere-2026-4196-RC1
RC2:
'Comment on egusphere-2026-4196', Anonymous Referee #2, 24 Aug 2026
Major comments
There is a mismatch between the vessel class analyzed and the operational/commercial shipping motivation of the paper. The paper is framed around the strategic economic importance of the Port of Churchill and integration into international markets. However, the analysis focuses on Type E vessels, which Figure 2 identifies as pleasure craft and which represent the least ice-capable vessel class considered. This vessel class is not representative of the cargo vessels currently operating in western Hudson Bay and Hudson Strait, nor of Canadian Coast Guard icebreaking vessels that are also in the area. The authors should either provide a stronger justification for why Type E is appropriate for the shipping questions posed in the introduction. Otherwise, the authors should narrow the framing of the manuscript to open-water/non-ice-strengthened vessel accessibility and reframe the need for such a study in the introduction/conclusion, or extend the analysis to one or more vessel classes representative of commercial shipping in the study area. The justification for the vessel class should also appear much earlier in the manuscript, such as in the Introduction or Methods section. (i.e. the current Introduction frames the study for an analysis of vessels in polar classes 2 to 7).

Several conclusions concerning higher vessel classes extend beyond the analysis that was performed. In the Discussion, the authors state that “the deployment of higher-class vessels, such as those built to polar standards, would considerably extend the NP since these vessels can navigate through thicker ice without significant difficulty.” This has not been quantified in the analysis. I suggest either conducting the analysis for representative higher vessel classes or revising this statement. Similarly, the statement that “even with extreme warming scenarios and the most permissive criteria, ice-strengthened vessels would still be necessary for 12-month shipping along these routes” follows logically from the finding that Type E vessels do not achieve year-round navigability. However, given the operational importance of higher vessel classes in this region, it would be considerably more useful to quantify how the opening date, closing date, and navigable period change across representative Polar Classes.

The choice of PIOMAS rather than Canadian Ice Service charts as the historical reference for the POLARIS analysis requires more justification. Around line 136, the manuscript states that Canadian Ice Service charts are preferred for operational purposes but that PIOMAS is preferred for comparison with climate models because PIOMAS and CMIP6 both use sea ice thickness rather than ice type. However, POLARIS itself is defined in terms of sea ice type, and the conversion from modeled sea ice thickness ranges to POLARIS risk values is therefore already an approximation. The manuscript should explain more clearly why similarity in the variable is preferable to using an operational ice product that more directly represents the ice types used by POLARIS. The statement that PIOMAS is a “more homogeneous” dataset should also be clarified.

Given that the bias correction is ultimately applied to derived navigability metrics (OD, CD, NP) rather than directly to sea ice thickness, could Canadian Ice Service charts instead provide the historical benchmark? The delta-shift correction is applied to quantities such as opening day, closing day, and navigable period. Therefore, it is not obvious that the historical reference dataset must contain the same underlying physical variable as CMIP6. Could POLARIS navigability calculated directly from CIS chart ice types be compared with historical CMIP6-derived navigability over their overlapping period, with that bias correction then carried forward into the CMIP6 future projections? Even if the authors ultimately retain PIOMAS, this alternative should be discussed and the reasons for not using it made explicit.

The sensitivity of the path-continuity analysis to spatial resolution should be considered and discussed. The manuscript emphasizes the importance of requiring connected navigable grid cells rather than evaluating a spatial average, but the datasets and CMIP6 models have different spatial resolutions. This could influence the probability of finding a continuous navigable path. It would be helpful to report the approximate number of grid cells spanning the 200-km corridor for the different datasets/model resolutions and to discuss resolution dependence as a limitation of the path-finding approach. The authors criticize the use of regional averaging in previous studies but using this approach with a coarse resolution approaches the same result as regional averaging, so it should be made clear how much higher the resolution of this study is than regional averaging to highlight the benefits of this approach. Also, is the blockiness of the grid cells in figure 1 representative of the grid cell size used in the study’s analysis?

Minor comments
As the paper stands now, I suggest specifying “low-ice or ice-free navigability” in the title. The paper specifically evaluates expansion of the ice-free/low-risk season and does not quantify the portion of the year that remains navigable for more ice-strengthened vessels. However, the framework could be used for the entire shipping season beyond low-ice or ice-free navigability if the authors choose to conduct that analysis.

There is a grammatical error in the abstract: “representing and increases of 137 days” should be “representing an increase of 137 days.”

Figure 2 includes POLARIS classes PC2 through PC7 even though only Type E is analyzed. The inclusion of additional classes in the figure emphasizes the need to explain clearly why Type E was selected.

Check the grammar around line 114: “…and therefore in the operational shipping season.”

“Charts” around line 136 should not be capitalized.

The shipping zone in Figure 1 is captioned as “dark grey color,” but it appears dark blue for me when printed. Canadian spelling (“colour”) may also be more appropriate.

Around line 160, the discussion of northern and southern Hudson Strait routes is initially confusing because the manuscript has already introduced the Churchill–Labrador Sea route and the Churchill–community routes, neither of which is north/south. Since the difference between the northern and southern Hudson Strait passages is negligible at the scale of Figure 1, I suggest stating that the shipping zone was defined broadly enough to encompass both the northern and southern entrances to Hudson Strait.

The heading “Shipping Route Mask Setup” is confusing because this section is defining the possible shipping area rather than masking an area out. A heading such as “Shipping Zone Definition” would be clearer.

Around line 167, I suggest changing “two methods are used to define navigability” to “two definitions” or “two sets of criteria are used to define navigability.” I don’t quite think it’s a method.

Section 2.4 would be clearer if it were framed explicitly as the classification of individual grid cells rather than route detection. A title such as “Grid-Cell Navigability Classification” or “Navigability Assignment” would better distinguish it from Section 2.5, where the path-finding algorithm actually determines whether navigable cells form a continuous route. Related wording in Section 2.4 should refer to grid-cell navigability rather than saying that a “route” is low risk before route connectivity has been assessed.

The heading “Path continuity framework” should be capitalized consistently with the other section headings.

I suggest rewriting the first two sentences of Section 2.5 to more directly explain that route navigability is determined by whether connected navigable grid cells form a continuous path between origin and destination, i.e. "A path continuity framework ensures navigability is assessed as a route of connected navigable grid cells between the origin and destination." But this is at your discretion.

It would be helpful to report the approximate grid-cell width of the 200-km shipping zone for each dataset or model resolution rather than stating only the physical width of the corridor. (The authors state the resolution of OSISAF and PIOMAS, but not CMIP6, or how varying grid cells were handled despite stating in section 2.3 there were varying resolutions).

There appears to be a formatting problem around line 253, where text or mathematical notation is missing beside the Gamma function?

There also appears to be missing text around line 273.

N_eff is defined on line 290 but then no longer referenced or used.

I do not understand the term “projection time” around line 406. This should be explained.

Figure 3 contains a category labelled “too distant,” but this category is not explained in the text. If it means that a cell lies outside the permitted 200-km shipping zone or beyond an allowed connection distance, this should be explicitly stated or the category should be renamed.

The description of opening and closing dates in Section 2.6 is difficult to follow. The section begins by defining the sea ice year before explaining how the opening and closing dates themselves are identified, whereas readers may initially expect a direct definition of opening and closing dates. The repeated use of “median day” also makes the sequence difficult to reconstruct. I suggest revising this section so that the logic and order of the calculations are stated more explicitly and the terms “median open day,” “median closed day,” opening day, and closing day are clearly distinguished.

The authors should acknowledge in the Discussion that the ice-free or low-ice defined navigable period does not necessarily correspond directly to the operational shipping season. There are regulatory constraints that limit whether vessels can proceed in earlier or later navigability. Refer to Andrews et al., 2017 for this.
Citation: https://doi.org/10.5194/egusphere-2026-4196-RC2

Armina Soleymani, Alex Crawford, Lawrence Mudryk, and Feiyue Wang

Supplement

https://doi.org/10.5194/egusphere-2026-4196-supplement

Armina Soleymani, Alex Crawford, Lawrence Mudryk, and Feiyue Wang

Viewed

Total article views: 310 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
220	74	16	310	34	19	14

HTML: 220
PDF: 74
XML: 16
Total: 310
Supplement: 34
BibTeX: 19
EndNote: 14

Views and downloads (calculated since 23 Jul 2026)

Month	HTML	PDF	XML	Total
Jul 2026	58	19	6	83
Aug 2026	130	46	6	182
Sep 2026	32	9	4	45

Cumulative views and downloads (calculated since 23 Jul 2026)

Month	HTML	PDF	XML	Total
Jul 2026	58	19	6	83
Aug 2026	130	46	6	182
Sep 2026	32	9	4	45

Viewed (geographical distribution)

Total article views: 273 (including HTML, PDF, and XML) Thereof 273 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 04 Sep 2026

Short summary

Hudson Bay’s seasonal sea ice limits maritime access to the Port of Churchill, and although we know future warming will lengthen the navigable period, how quickly it will lengthen is unclear. Based on several methods for making future projections, the navigable period for the route from Churchill to the Labrador Sea has lengthened by about 4 weeks since the 1980s. In the future, it is projected to lengthen by about 2 weeks for each additional 0.5 °C of global warming.


Total:	0
HTML:	0
PDF:	0
XML:	0