Atmospherically forced sea-level variability in western Hudson Bay, Canada

. In recent years, signiﬁcant trends toward earlier breakup and later freeze-up of sea ice in Hudson Bay have led to a considerable increase in shipping activity through the Port of Churchill, which is located in western Hudson Bay and is the only deep-water ocean port in the province of Manitoba. Therefore, understanding sea-level variability at the port is an urgent issue crucial for safe navigation and coastal infrastructure. Using tidal gauge data from the port along with an atmospheric reanalysis and Churchill River discharge, we assess environmental factors impacting syn-optic to seasonal variability of sea level at Churchill. An atmospheric vorticity index used to describe the wind forcing was found to correlate with sea level at Churchill. Statistical analyses show that, in contrast to earlier studies, lo-cal discharge from the Churchill River can only explain up to 5 % of the sea-level variability. The cyclonic wind forcing contributes from 22 % during the ice-covered winter– spring season to 30 % during the ice-free summer–fall season due to cyclone-induced storm surges generated along the coast. Multiple regression analysis revealed that wind forcing and local river discharge combined can explain up to 32 % of the sea-level variability at Churchill. Our analysis further revealed that the seasonal cycle of sea level at Churchill appears to be impacted by the seasonal cycle in atmospheric circulation rather than by the seasonal cycle in local discharge from the Churchill River, particularly post-construction of the Churchill River diversion in 1977. Sea level at Churchill shows positive anomalies for September–


Introduction
through Hudson Bay to Hudson Strait, which may soon become a federally-designated 114 transportation corridor (e.g., Andrews et al., 2017;Pew Charitable Trusts, 2016). 115 In 2016, the University of Manitoba and Manitoba Hydro launched a project on "Variability and 116 change of freshwater-marine coupling in the Hudson Bay System", named BaySys, which aimed 117 to assess the relative contributions of climate change and river regulation to the Hudson Bay 118 system. Here, we are specifically focused on the impact of the Churchill River diversion on 119 variability of sea level at the Port of Churchill. Additionally, we put our findings in the context of 120 wind forcing over the entire Hudson Bay, elaborating on the suggestion by Dmitrenko et al. 121 (2020) that cyclonic wind forcing generates onshore Ekman transport and storm surges along the 122 coast. 123 We also revisit earlier results by Gough and Robinson (2000) and Gough et al. (2005). Using 124 tidal gauge and river discharge data from 1974 to 1994, Gough and Robinson (2000) suggested 125 that the Churchill River discharge dominates sea-level variability at Churchill. They explained 126 the seasonal elevation of sea level during late fall by a recirculating mechanism that links the 127 spring pulse of river discharge in the downstream James Bay (Figure 1) to sea level at Churchill 128 (Gough and Robinson, 2000;Gough et al., 2005). In this paper, we present an alternative 129 mechanism and show that (i) the Churchill River discharge plays a secondary role for generating     , 1984, 1987). When data gaps occurred, then the upstream hydrometric gauge below Fidler 244 Lake (station #06FB001) was used to infill data, with streamflow data adjusted to account for the 245 difference in contributing area between Fidler Lake and the Churchill outlet, following the 246 procedure of Déry et al. (2005). When the upstream hydrometric data were also unavailable, a 247 secondary step was taken to infill data gaps. Missing data on a given day were infilled using the to non-stationarity in the discharge record, which was implemented in our analysis.

267
The sea level record in Churchill is impacted by the post-glacial isostatic adjustment, with

286
In this section, we examine the impact of cyclonic wind forcing and local river discharge on sea

308
A closer look at the daily data reveals that the sea level seasonal maximum from October-     Table 1). For the ice-377 free period from June to November, correlations for whole period, and pre-and post-diversion 378 increase to 0.54, 0.52 and 0.55 (Table 2), respectively, compared to 0.47, 0.49 and 0.47 for the 379 ice-covered period from December to May (Table 3). We test the difference between correlations 380 estimated for the ice-covered and ice-free seasons using the Fisher z-transformation (Fisher,381 1921). Statistical assessment shows that the only differences between correlations estimated for 382 whole period and post-diversion are statistically significant at the 99% confidence level.

383
The relationship between vorticity and SLA changes significantly from one year to another. The     (Table 1).

422
Our results also reveal the important role of sea-ice cover and river diversion in modifying 423 controls on sea level variability. During the ice-free seasons from 1960-1976, the contribution of 424 wind forcing is 27%, and the role of river discharge is negligible ( pre-and post-diversion periods (Table 3). Summarizing these results, we point out that the sea- 5a) appears to be linked to the spring freshet of the Churchill River (Figures 5a and 5c).

472
However, post-diversion this positive SLA in June vanishes due to the abrupt decrease in the Churchill River discharge during the spring freshet from ~1,500 to 700 m 3 s -1 (Figure 5c).

474
Gradual decreases in Churchill River discharge from June/July to April for both pre-and post-  Figures 8b and 8c).

492
The zero SLA contour in Figure 8b and 8c is displaced relative to the zero vorticity and the long- level. This is consistent with a previous concern about significant impact of Churchill River 566 discharge on SLA in Churchill.

567
Third, our analysis shows that the seasonal cycle in sea level variability with positive SLA 568 during fall is observed not only in Churchill, but also along the eastern coast of Hudson Bay in 569 Innukjuak (Figures 1 and 9). While the sea level record at Innukjuak is short and not continuous, Innukjuak is consistent with seasonal amplification of atmospheric vorticity (Figures 5b and 9). coast of Hudson Bay with SSH differences between fall and summer ranging from ˃5 cm in 596 James Bay to ~1 cm along the northwest coast ( Figure 10). This confirms our results that a 597 positive SLA during fall is generated over the entire coast of Hudson Bay, and particularly in Churchill and Innukjuak, in response to enhanced cyclonic wind forcing (Figures 5a, 5b, and 9). Churchill is inconsistent.

602
One may suggest that seasonal SSH elevation in Figure 10 can be partly due to the thermosteric 603 and halosteric sea-level rise. During summer, the Hudson Bay coastal domain receives large 604 amount of fresh and warm water from river runoff. The seasonal tendency for river discharge, 605 however, is opposite to that for the SSH in Figure 10. (e.g., Figure 6) increasing risks to re-supply and fuel-transfer operations.

672
The origin of seasonality in cyclonic wind forcing, its climatic aspects and ocean response to