Responses of Atlantic Water Inflow Through Fram Strait to Arctic Storms

Changes in the volume transport of Atlantic water into the Arctic Ocean can affect the heat and mass balance in the central Arctic Ocean. To understand the impacts of Arctic storms on the inflow through Fram Strait, we implemented the NEMO ocean model for the Arctic Ocean, to simulate the decadal variations of the water volume transport through Fram Strait. The simulations suggest that the water inflow tends to be weaker in the decades of the 1960 and 2010s but stronger in the 1980s. The decadal variation is associated with decadal variability of the storm density in the Greenland Sea. When there is an increased storm density near Fram Strait, the southerly wind anomalies dominate the Atlantic water pathway. As a response, there is an increased Atlantic inflow through Fram Strait.


Introduction
Fram Strait is one of the major pathways for Atlantic water into the central Arctic Ocean (Aagaard, 1989;Karcher & Oberhuber, 2002;Rudels et al., 1994).Atlantic water enters the Arctic Ocean along the west Spitsbergen Strait at depths about 200-900 m and spreads over the whole Arctic Ocean through a cyclonic boundary current and thermohaline intrusion, forming the warm Atlantic Water Layer.Moreover, the west Spitsbergen Current (WSC) carries about 40 × 10 12 W of heat and plays an important role in changes to ocean temperature in Canada Basin (Aggard et al., 1987;Hanzlick, 1983;Karcher et al., 2003;Long & Perrie, 2015;Mclaughlin et al., 2009).
The WSC carries warm, saline Atlantic water above the shelf slope of Svalbard.The inflow is topographically controlled and shows a strong barotropic component (Quadfasel et al., 1987).Of the two branches of WSC flowing into the central Arctic Ocean, the nearshore WSC branch follows the Spitsbergen slope, whereas the offshore WSC branch moves along the western boundary of the Yermak Plateau.The offshore branch is relatively deep, and there is no significant air-sea interaction.However, the heat flux associated with the nearshore WSC waters, on their way into the central Arctic Ocean, is about 200 Wm 2 during ice-free winter conditions (Aggard et al., 1987).
The total transport of the WSC is about 6.6 Sv, with significant seasonality and internannual variability (Beszczynska-Möller et al., 2012;Kawasaki & Hasumi, 2016).Its total volume transport tends to be stronger in winter, and weaker in summer (Schauer et al., 2004).On interannual time scales, observations show two warming events in the 1990and 2000s (Beszczynska-Möller et al., 2012).The warming events are abrupt and pulse-like (Bourgain & Gascard, 2012;Karcher et al., 2003;Polyakov et al., 2005), progressing from Fram Strait toward the central Arctic Ocean along the edge of the central Arctic Basin (Karcher et al., 2003;Dmitrenko et al., 2008).Previous studies also suggest that the large-scale atmospheric surface forcing, such as the spatial shifts of the leading atmospheric circulation patterns (Beszczynska-Möller et al., 2012;Karcher et al., 2003;Kawasaki & Hasumi, 2016;Zhang et al., 2008) and the ice export through Fram Strait (Wang et al., 2020), are the main two mechanisms for the interannal variability.
Due to limited observations, the decadal variations of the Atlantic water inflow are still an open scientific question.The objective of this paper is to investigate the decadal variations of the west Spitzbergen current.Section 2 describes model and data analyses.Section 3 validates the model climatology with available observations and reanalysis data.Section 4 shows the simulated decadal variations of water volume transport through Fram Strait.Section 5 presents the conclusions.

Ocean Model
NEMO is a coupled ice-ocean model developed by several research groups in Europe (Barnier et al., 2006).Its ocean component OPA is a threedimensional, primitive equation ocean model with complete thermohaline dynamics (Madec & the NEMO team, 2008).In addition, a two-level LIM2 ice model solves the ice dynamics and thermodynamics (Bouillon et al., 2009;Fichefet & Morales Maqueda, 1997).NEMO has been successfully applied in climate studies on both global and regional scales (Barnier et al., 2006).NEMO 3.6 is implemented for the whole Arctic Ocean (Figure 1).The model grid was selected from a NEMO tri-polar global spherical coordinate domain with a ¼ degree horizontal resolution (ORCA-R025; Barnier et al., 2006).The model has 50 z-vertical levels.Horizontal diffusion is handled with a Laplacian scheme, and a bi-Laplacian scheme is applied for horizontal viscosity.A Mellor-Yamada turbulence kinetic energy scheme is employed to represent the vertical mixing dynamics (Mellor & Yamada, 1982).In addition, the Neptune effect is implemented to represent the cyclonic rim currents associated with topographic stress (Holloway, 1992).
This study focuses on the role of Arctic storms in the decadal variability of Atlantic water inflow through Fram Strait.Lateral boundary conditions are prescribed as climatology.Initial and boundary conditions for ocean salinity and temperature are provided by the Polar Science Center Hydrographic Climatology (PHC3.0;Steele et al., 2001).Lateral boundary currents are prescribed by Glorys climatology (Ferry et al., 2010(Ferry et al., , 2012)).River runoff is prescribed as climatology from ORCA-R025 (Barnier et al., 2006).The atmospheric surface forcing is taken from the 3-hourly JRA55 reanalysis data set (Y. Harada, et al., 2016;S. Kobayashi et al., 2015).The model was given a spin-up of 40 years, forced by JRA55 data from 1961 to 2000, and integrated for the entire period from 1958 to 2021.

Cyclone Tracking
A cyclone identification and tracking scheme is applied to the 6-hourly JRA-55 sea level pressure fields from 1960 to 2021 to obtain storm estimates in the Arctic (Y.Harada, et al., 2016;Kobayashi, S. et al., 2015).The scheme identifies the storms, based on the structure of the MSLP fields, by comparing the value of the Laplacian of pressure (LP), ∇ 2 p, at each grid point to those at neighboring grid points (Murray & Simmonds 1991, 1995).The algorithm objectively identifies both open and closed low pressure systems, including small-scale storms.The scheme has been successfully applied in a number of studies (Neu et al., 2013;Raible et al., 2008;Uotila et al., 2009).Details of the identification and cyclone tracking scheme have been described by Simmonds and Keay (2000) and Simmonds et al. (2008).

Model Validation
The Atlantic water near West Spitzbergen is relatively warm, salty and well-mixed.The ocean temperature ranges from 1°C offshore to 4°C nearshore.The temperature of the Arctic outflow along the Greenland Shelf is mainly uniform with a temperature below 1°C (Figure 2a).The salinity of Atlantic water is above 34 psu, but the Arctic outflow is relatively fresh with a salinity minimum below 31 psu (Figure 2c).The NEMO simulation can reproduce the spatial patterns of ocean temperature and salinity but slightly overestimates the temperature of the Atlantic water inflow by about 1°C and the salinity of the outflow by about 1 psu (Figures 2b and 2d).
The simulated meridional currents are compared to Glorys reanalysis data (Ferry et al., 2010(Ferry et al., , 2012) ) in Figures 2e  and 2f.The Atlantic water inflow is mostly barotropic over the continental shelf.The maximum inflow is 6-12 cm/s.In addition, the outflow has a maximum of 15-18 cm/s above 200 m (Figure 2e).The NEMO simulations show a similar spatial pattern but overestimate the inflow by about 3 cm/s and underestimate the outflow by about 6 cm/s (Figure 2f).
Since 1997, year-round field measurements have been carried out in Fram Strait (Beszczynska-Möller et al., 2012), and the total transport has been computed along 5°E-9°E at 78.8333°N (Figure 3).The total transport of Atlantic water inflow shows a significant seasonal variation and tends to be stronger in winter but weaker in summer (Figure 3a).Following Beszczynska-Möller et al., 2012, the total transport is also estimated for NEMO simulations.Compared to the field measurements, NEMO can simulate the seasonal variability well, with an correlation coefficient of 0.61 (Figure 3a).On the interannual scale, both the model simulation and the observations show an increased inflow for the annual water volume, in 1999-2001 and 2012 (Figure 3b).

Decadal Variations of Water Volume Transport Into the Central Arctic Ocean
Changes in the storm density can play an important role in the poleward heat transport, in particular, when there is an increased frequency of intense Arctic cyclone occurrence, as seen in recent decades (Zhang et al., 2023).Consistent with previous studies (Hoskins & Hodges, 2019;Simmonds et al., 2008;Zhang et al., 2004), winter storms originating in the Northwest Atlantic move northeastward into the Greenland Sea (Figure 4a).While most of the storms continue eastward into the Barents Sea, some storms move northward into Fram Strait.The maximum storm density is about 30 storms per winter season (November-April) near Fram Strait and 24 in the

Geophysical Research Letters
10.1029/2023GL107777 central Barents Sea.However, there are significantly fewer storms in summers (Figure 4b), and the analyses in this study only focus on the fall-winter storm activity, during November-April.
The average storm density near Fram Strait shows a decadal variation from 1960 to 2021 (Figures 5a-5c).In the 1960s, more storms move eastward into the Barents Sea, whereas the Fram Strait branch of these Arctic storm tracks is relatively weak; the maximum storm density near Fram Strait is about 18.However, the Fram Strait branch tends to be relatively stronger in the 1980s, with a maximum storm density that is about 33.By comparison, the spatial pattern in the 2010s is similar to that of the 1960s, and the Fram Strait branch is significantly weaker with a maximum storm density of 24.
Due to the strong surface winds associated with Arctic storms, the decadal variations in storm density can have significant impacts on the water volume transport through Fram Strait (Figure 5d).For example, when the average storm density near Fram Strait is 14 in the 1960s, the average water volume transports is 5.9 Sv.However, when there is increased storm activity in the 1980s, with an average storm density of 20, the simulated water volume transport tends to be stronger, and the average water volume transport is 8.5 Sv.Moreover, the storm density near Fram Strait and the water volume transport show significant values for their correlation coefficient, reaching 5.9.This suggests the importance of the impacts of storm variability on the water volume transport through Fram Strait.
Climatologically, the atmospheric circulation along Norwegian coast is cyclonic.Fram Strait is dominated by northerly winds in winter with southerly winds prevailing along the Norwegian coast (Figure 5e).When there are fewer storms near Fram Strait in the decades of the 1960 and 2010s, there is an anticyclonic wind anomaly in the central Greenland Sea, and Fram Strait that is dominated by northerly wind anomalies which tend to reduce the water volume transport through Fram Strait (Figure 6).However, when there are more storms near Fram Strait, there is a cyclonic wind anomaly in Fram Strait, and southerly wind anomalies dominate the Atlantic water inflow pathway.Thus, the Atlantic water inflow tends to intensify (Figure 6).
Ocean heat flux along 5°E-9°E at 78.83°N is dominated by a linear trend in ocean temperature (Figure 7).There is an increasing trend in ocean temperature from 1960 to 2021, whereby the ocean temperature increases from about 3°C in 1960s to about 4.5°C in 2010s.Due to the increased ocean temperature, the heat flux associated with the Atlantic water inflow tends to increase.The average increase is from 25 TW in the 1960s to about 40 TW in 2010s.However, the increasing trend in the heat flux is not as significant as that of the ocean temperature.For example, the heat flux tends to be weak in the 1960 and 2000s, suggesting the impacts of water volume transport.Consistent with observations (Polyakov et al., 2005), the heat flux shows two peak warming events, one in the early 1990s and the other in the 2010s, which are mostly associated with interannual variations of the water volume transport through Fram Strait (Figure 5d).

Conclusions
The simulated water inflow through Fram strait tends to be weaker in the 1960 and 2010s but stronger in the 1980s.The decadal variation in water volume transport is associated with the decadal variability in storm density in the Greenland Sea.When there is an increased storm density near Fram Strait, there is a southerly wind anomaly near Fram Strait, and the Atlantic water inflow tends to intensify.However, when there are fewer storms near Fram Strait, there is an anticyclonic wind anomaly in the Greenland Sea, and the northerly winds are intensified, which tends to reduce the Atlantic inflow.Moreover, the decadal variations of storm density near Fram Strait can be related to NAO.For example, in the 1980s, the storm density tends to increase during the positive phase of NAO.
The heat flux associated with the Atlantic water inflow reflects the impacts of the linear trend in ocean temperature and the decadal variations in water volume transport.Thus, there is an increasing trend in the ocean flux  : 1960-1970, 1980-1990, 2010-2020.into the central Arctic Ocean from 1960 to 2021 and the heat flux tends to be weak in the 1960 and 2010s, showing a decadal variation.

Figure 1 .
Figure 1.Model domain and bathymetry for NEMO.Units are in m, and the thick red line across Fram Strait shows the location of hydrographic sections mentioned in the text.

Figure 2 .
Figure 2. Cross sections for annual ocean temperature (°C, upper panel), salinity (middle panel) and currents (lower panel) as a function of depth (m) along the line indicated in Figure 1, showing: PHC data in panels(a, c), Glorys data in panel (e), and NEMO results averaged for 1970-1999 in panels (b, d, f).

Figure 3 .
Figure 3. Water volume transport averaged along 5°E-9°E at 78.83°N, showing observations (black) and NEMO results (red): (a) monthly water volume transport, where correlation coefficient is 0.61, and (b) annual water volume transport.Unit: Sv.

Figure 4 .
Figure 4. Storm density in panel (a) November-April and (b) May-October averaged for 1970-1999.The black box shows the area where the average storm density is computed.Unit: storms per grid in November-April.

Figure 5 .
Figure 5. November-April storm density for decadal periods (a) 1960-1970, (b) 1980-1990, and (c) 2010-2020.(d) Shows November-April time series of storm density averaged over the area indicated by the black box in Figure 4a (black line) and annual water volume transport averaged along 5°E-9°E at 78.83°N (red line in Figure 1) with correlation coefficient of 0.59, and (e) shows the 10 m wind (m/s) for November-April averaged for 1970-1999.

Figure 6 .
Figure 6.Upper panel: the 10 wind anomaly (m/s) for November-April relative to the average for 1970-1999.Lower panel: cross sections for annual meridional current (m/s) as a function of depth (m) along the line indicated in Figure 1, showing NEMO results, for decadal periods: 1960-1970, 1980-1990, 2010-2020.