Interannual variability of Pacific Winter Water inflow through Barrow Canyon from 2000 to 2006

Abstract

We examined the interannual variability of Pacific Winter Water (PWW), both upstream in the northeastern Chukchi Sea and Barrow Canyon using mooring observations from 2000 to 2006, and downstream in the Canada Basin using hydrographic data acquired in 2002–2006. The interannual variation of PWW salinity is governed by two factors: (1) variability in the salinity of Pacific Water that flows northward through Bering Strait in winter; and (2) the input of salt associated with sea ice formation during winter in an intermittent coastal polynya located along the Alaskan coast between Cape Lisburne and Point Barrow. During the winters of 2000/2001 and 2001/2002 an increased transport of cold and saline PWW (S > 33.5) to the basin via Barrow Canyon was observed. In 2000/2001 enhanced ice formation in the polynya contributed to the increased salinity of PWW, whereas in 2001/2002 the salinity of water entering through the Bering Strait was higher, and this resulted in more saline PWW being delivered to the basin. In the following four winters (2002/2003, 2003/2004, 2004/2005 and 2005/2006) the transport of cold and saline PWW in winter to the basin was less than that in the two preceding winters. In three of these four winters (2003/2004 being the exception) the coastal polynya was less active, thus reducing the input of salt due to brine enrichment. In the winter of 2003/2004, however, warmer water within the polynya region constrained ice formation and thus less cold and saline PWW was produced, despite the fact that the coastal polynya was active and frequently open.

Appendices

Appendix 1

1.1 Interpolation of mooring data and error estimation

Temperature, salinity, and velocity data measured in Barrow Canyon were linearly interpolated to every 2 m between the observational levels shown in Fig. 4. The shallowest data were extended from the shallowest observational level to the surface, and the deepest data were extended from the deepest observational level to the bottom. Most of the mooring data were recorded at hourly intervals. We calculated daily averaged data using a 25-h running mean. Some instruments failed (Fig. 4); thus, we had to use different interpolation methods to create time series of temperature, salinity, and velocity.

1.2 Interpolation of Stns. BCC-00 and BCC-01 data

At Stn. BCC-00, the Aanderaa recording current meters (RCMs) at 195 and 235 m failed to record data from December 2000 to September 2001 and from January 2001 to September 2001, respectively. We thus extended the velocity observed at 140 m to the bottom, decreasing by 4.7 % every 10 m. This rate was estimated from the acoustic Doppler current profiler (ADCP) data observed at Stn. BCC from October 2001 to June 2002 and from October 2005 to September 2006; these periods had good temporal records of temperature, salinity, and velocity. From July 2002 to August 2002 at Stn. BCC-01, the upward-looking ADCP at 150 m failed and we used velocity data recorded from the surface to 50 m at Stn. BCE-01 instead. Between 50 and 150 m, we linearly interpolated velocity data observed at 50 m at Stn. BCE-01, 75 m at Stn. BCC-01, and 150 m at Stn. BCC-01.

To evaluate the error arising from insufficient velocity data at Stns. BCC-01 and BCC-00, the PWW transport obtained from real velocity profiles (ADCP data) was compared to that calculated from estimated velocity profile data from October 2001 to June 2002 and from July to August 2006. There were good temporal records of temperature, salinity, and velocity during these periods (Fig. 4). The difference between the annual mean transport using ADCP data and that estimated using the interpolated data was only a small percentage of the total transport. Therefore, the error of PWW transport arising from these interpolation methods was negligible.

1.3 Interpolation of Stn. BCC-02 data

At Stn. BCC-02, the Seabird Microcat (SBE) instrument at 50 m failed from September 2002 to September 2003, and the SBE at 158 m failed from November 2002 to September 2003. In the upper layer, temperature and salinity at Stn. BCC were in between values measured at Stns. BCE and BCW (Fig. 5). Therefore, mean temperature and salinity values (Stns. BCE-02 and BCW-02) were calculated and used for Stn. BCC-02 above 50 m. Temperature and salinity properties at Stn. BCW-02 at 152 m were used to determine the T–S curve at Stn. BCC-02 from 89 to 252 m. However, the vertical salinity profile at Stn. BCC-02, which is nearly the same as the density profile, could not be estimated from BCW-02 (Fig. 5). Thus, the monthly averaged profiles of the salinity gradient from 89 to 252 m were calculated for the following three cases using the differences of velocity at 150 m at Stn. BCC: (1) southwestward or weak northward current (current speed in northwestward direction, v < 20 cm/s), (2) northeastward current (v > 20 cm/s), and (3) strong northeastward current (v > 40 cm/s), using data from October 2001 to August 2002 and from October 2005 to September 2006. Calculated monthly averaged salinity profiles for these three cases were used to estimate the salinity profile at Stn. BCC-02 from November 2002 to September 2003. From September 2002 to September 2003, no velocity data from the surface to 120 m were available at Stn. BCC-02, and thus the velocity measured at 50 m at Stn. BCE-02 was used from the surface to 120 m at Stn. BCC-02.

To evaluate the error at Stn. BCC-02 due to the missing data, we applied the interpolation method used for Stn. BCC-02 to data from October 2001 to June 2002 and from July to August 2006. We then compared PWW transport values obtained from reliable temperature, salinity, and velocity data with those calculated from the interpolated data for these periods. PWW transport (32 < S < 33.5) decreased by 0.042 Sv for 2001–2002 and by 0.047 Sv for 2005–2006 when compared to values calculated using real temperature, salinity, and velocity profile data. In Fig. 10, the error bar indicates the magnitude of the potential underestimation of PWW transport at Stn. BCC-02 in 2002–2003.

1.4 Interpolation of Stn. BCC-03

The SBEs at 180 and 250 m at Stn. BCC-03 failed from October 2003 to September 2005; thus, no temperature and salinity data were collected below 120 m. Warm Atlantic Water is generally observed near the bottom at Stn. BCC (Fig. 5); thus, the mean properties of Atlantic Water from Stn. BCC at 250 m (T = 0.3, S = 34.64) from October 2001 to June 2002 and from October 2005 to September 2006 were used instead. However, the temperature and salinity recorded at 120 m at Stn. BCC-03 were extended to the bottom when cold and saline brine water (T < −1.5 °C) was observed at the bottom of Stn. BCW-03. Temperature and salinity properties at Stn. BCW-03 at 158 m were used to determine the T–S curve at Stn. BCC-03 from 120 to 250 m. The monthly averaged profile of the salinity gradient between 120 and 250 m at Stn. BCC-03 was calculated, as we interpolated salinity data for Stn. BCC-02 between 89 and 252 m from November 2002 to September 2003. Using the T–S curve observed at Stn. BCW-03 and calculated monthly averaged salinity profiles, temperature and salinity properties from 120 to 250 m at Stn. BCC-03 were estimated from September 2003 to September 2005. The upward-looking ADCP at 240 m failed from August 2004 to the end of September 2005, and only RCM velocity data at 57 m were available for Stn. BCC-03. Thus, the velocity measured at 57 m was applied to the velocity from the surface to the bottom at Stn. BCC-03.

We evaluated the error in PWW transport calculations arising from these data deficiencies at Stn. BCC-03. Based on coverage of velocity data, observational periods of Stn. BCC-03 can be divided into two periods: from October 2003 to September 2004 and from October 2004 to September 2005. First, we applied the interpolation method used for the first period of Stn. BCC-03 to data from October 2001 to August 2002 and from October 2005 to September 2006. PWW transport (32 < S < 33.5) decreased by 0.027 Sv for 2001–2002 and by 0.038 Sv for 2005–2006 when compared to values calculated from real data. Saline PWW transport (S > 33.5) decreased by 0.020 Sv for 2001–2002 and by 0.002 Sv in 2005–2006 when compared to values calculated from real data. Second, we applied the interpolation method used for the latter period of Stn. BCC-03 to data for October 2001 to August 2002 and October 2005 to September 2006. PWW transport (32 < S < 33.5) decreased by 0.040 Sv for 2001–2002 and by 0.062 Sv for 2005–2006 when compared to values calculated from real data. Saline PWW transport (S > 33.5) calculated using these assumptions decreased by 0.034 Sv for 2001–2002 and by 0.022 Sv in 2005–2006 when compared to values calculated from real data. The error bars on PWW transport in Fig. 10 indicate the possible magnitude of the underestimation for 2003–2004 and 2004–2005.

Appendix 2

2.1 Calculations of heat flux and salt flux

The algorithms used to estimate heat and salt fluxes generally follow those of Drucker et al. (2003) and Martin et al. (2004). We applied the estimation method used by Drucker et al. (2003) to estimate F _T and F _L. Shortwave radiation F _R was calculated based on the methods used by Cavalieri and Martin (1994). Turbulent heat flux (F _T) was calculated using the formula as follows:

$$ F_{\text{T}} = \rho_{\text{air}} C{\text{cp}}W(T_{\text{air}} - T_{\text{surf}} ) $$

(4)

where ρ _air is the density of air (1.3 kg/m³), cp the specific heat of air at constant pressure (1004 J/deg/kg), W the air speed at 10 m, T _air the air temperature at 10 m, and T _surf the surface temperature of thin ice. The heat transport coefficient (C) was assigned a value of 2.0 × 10⁻³ to take both sensible and latent heat flux into account, following the methods of Andreas and Murphy (1986). Longwave radiation (F _L) was estimated as:

$$ F_{\text{L}} = \varepsilon_{\text{ice}} \sigma T_{\text{surf}}^{4} - \varepsilon_{\text{air}} \sigma T_{\text{air}}^{4} $$

(5)

where ε _ice is the thin ice emissivity (0.98), and σ is the Stefan-Boltzman constant. The effective atmospheric emissivity ε _air is given by ε _air = 0.7829(1 + 0.2232CL^2.75), following the methods of Maykut and Church (1973), and CL is the fraction of cloud cover. Shortwave radiation was estimated using the parameterizations found by the empirical formula by Laevastu (1960) as

$$ F_{\text{R}} = (1 - \alpha )(1 - 0.6{\text{CL}}^{3} )F_{{{\text{R}}0}} $$

(6)

where α is the surface albedo of thin ice (0.3) (Maykut 1978), and F _R0 is the mean daily incoming solar radiation, estimated using Zillman’s (1972) empirical formula. The fluxes were calculated using NCEP 6-h gridded reanalysis meteorological data (wind speed, air temperature, atmospheric pressure, and cloud cover).

As long as the ice is not melting, the surface heat loss F should be equal to the conductive heat flux F _ice. We assumed that the conductive heat flux through the ice, F _ice, was a linear function of the temperature difference across the ice, as follows:

$$ F_{\text{ice}} = ki(T_{\text{surf}} - T_{\text{water}} )/h_{\text{t}} $$

(7)

where ki = 2.03 W/m/K is the conductivity of sea ice, T _water the seawater temperature, and h _t the ice thickness calculated from SSM/I (Drucker et al. 2003). We assumed that there was no snow cover on the thin ice.

The ice production (V _ice) and salt flux (SF_P) in the polynyas, in cubic meters per unit area per day, are given using the following equations (Cavalieri and Martin 1994; Martin et al. 2004):

$$ V_{\text{ice}} = F_{\text{ice}} /(\rho_{\text{ice}} L) $$

(8)

$$ {\text{SF}}_{\text{P}} = \rho_{\text{ice}} V_{\text{ice}} (S_{\text{water}} - S_{\text{ice}} ) \times 10^{ - 3} $$

(9)

where F _ice is the net surface heat loss, ρ _ice = 920 kg/m³ the ice density, and L = 3.34 × 10⁵ J/kg the latent heat of fusion of sea ice. The salinity of seawater is represented by S _water and the salinity of frazil ice by S _ice, with both given in practical salinity units. The salinity of frazil ice S _ice is assumed to be:

$$ S_{\text{ice}} = 0.31S_{\text{water}} $$

(10)

according to the laboratory experiments of Martin and Kaufman (1981).