Velocity and Temperature Variability
on the Chukchi Slope
Abstract: The Canada Basin Acoustic Propagation Experiment (CANAPE) surveyed the Chukchi Slope in 2016 and 2017 with shipboard and moored measurements. I used their hydrography and velocity measurements, as well as wind and sea ice measurements, to characterize the Chukchi Slope variability. In general, the velocity on the Chukchi Slope shows two cores. The shallow portion flows westward along the shelfbreak as the Slope Current and the deeper portion flows eastward as the Shelfbreak Jet. The temperature profile has three major layers: meltwater, Pacific water, and Atlantic water. These features are not constant, and sea ice is the greatest predictor of the Chukchi Slope Current variability. The first mode of the empirical orthogonal function (EOF) analysis explains 65% of the variance in the temperature for 2016-2017. This mode resembles the mean temperature in summer and winter as defined by sea ice cover. When the sea ice forms and melts, the Slope Current reverses direction. I investigated the effects of wind stress and Ekman veering, but do not find significant correlation. Note, however, that the presence of sea ice limits the influence of wind stress on the water column, and that the shallowest velocity measurement is 20 m.
The following is a summary based on my Physical Ocean Science and Engineering Master’s defense
given at the University of Delaware on Tuesday, December 17, 2019.
Figures were created by me for the full-text PDF (with some modifications for slideshow readability) unless noted.
Introduction
The Arctic and the Chukchi Sea
Sea Ice Changes
Arctic summer sea ice extent is drastically different than previous years. Shown here is the difference between the September sea ice extent in 1996, the year I was born, and 2019, the year of this thesis defense. The large orange dot indicates the area of the Chukchi Sea I studied to characterize the “new Arctic”.
The Chukchi Sea
The Chukchi Sea is bordered by Siberia and Alaska. To the northeast is the Beaufort Gyre, which is a major freshwater storage area for the Arctic Ocean. Warm Atlantic water enters the Chukchi Sea from the northwest.
This schematic by Corlett and Pickart (2017) depicts the currents in the Chukchi Sea. Three main branches that enter through Bering Strait eventually lead to Barrow Canyon. From there, all the water was thought to flow east, but we are now seeing a westward flow as well.
I am looking closer at the Chukchi Slope, marked with an orange dot, to characterize this westward flow in 2016 and 2017.
Study Area
My study area is a portion of the Chukchi Slope, as you can see marked by the red bounding box in the map inset above.
I used
- satellite-measured sea ice concentration averaged over the area in the light gray shaded box
- reanalysis winds over the area in the map inset
- CTD casts taken in 2016 (green x) and 2017 (red +)
- moorings (circles) with thermistors and one ADCP
Basic Structure
What did the Chukchi Slope look like in 2016 and 2017?
Hydrographic Profile and Water Masses
A good place to start looking at the Chukchi Slope is temperature, salinity, density with depth. Above is the data from the deepest 2017 CTD station along the mooring line. As you can see, the density looks nearly identical to the salinity at these temperatures, so I will focus my discussion on other parameters.
There are four main layers of water masses. I follow the naming convention of Corlett and Pickart (2017) and define
- Top: Meltwater (MW) – cold and fresh river runoff and melted sea ice
- Upper middle: Bering Summer Water (BSW) – warm and moderately salty summer Pacific water
- Lower middle: Remnant Winter Water (RWW) – cold and moderately salty winter Pacific water
- Bottom: Atlantic Water (AW) – warm and very salty Atlantic water that traveled the Arctic
Although this is a good representation of the Chukchi Slope, it is only one cast out of 125. So I will show them all using a temperature-salinity (TS) plot.
Temperature-Salinity
Each measurement has a “signature” based on its temperature and salinity. The resultant TS plot shows which water masses were measured in a dataset. Here, the normalized count shows how much of each water mass was present in the 2016 (top) and 2017 (bottom) CTD surveys.
In both years, the characteristic parabolic bottom line is clearly visible. However, while the plot for 2017 looks “normal” for the area, signatures in 2016 are condensed along the 26kg/m^3 isopycnal. This spike is outside the scope of my study, but is worth noting.
Year-long Temperature
In addition to temperature measurements in many places at one time, I also have temperature measurements in one place over many times. Above is the temperature at mooring UD-7, which is the deepest and matches the above sample CTD cast. On the left is the mean temperature with depth and on the right the time series with depth. Black arrow markers on the left-hand side indicate thermistor locations and the gray bar sea ice concentration greater than 40%.
The mean profile (left) resembles the single cast profile (above) at depth in the AW and RWW layers. The surface BSW and MW layers, however, change throughout the year. The surface waters are cold in November and December but warm in August through October. These changes appear to have something to do with sea ice cover, so I added that to the list of possible forcing functions.
This is the temperature measured from one mooring in one spot over time, but with additional moorings, we can see temperature in many spots.
Thermal Wind
To understand my findings over the mooring line, we must first understand thermal wind.
The concept of thermal wind is typically used by meteorologists to describe velocity shear based on horizontal density gradients. In a stratified system where there is a horizontal density gradient as pictured above, velocity shear develops to maintain equilibrium. The resultant winds are perpendicular and to the right (in the Northern Hemisphere) to the pressure gradient force that tilts opposite the density gradient.
Layers of water masses function in the same way as layers of air masses. A tilted isopycnal (horizontal density gradient) forces currents perpendicular to the slope of the isobars (constant pressure lines) that tilt in the opposite direction.
Hydrographic Surface
Above is the temperature measured on the 26kg/m^3 isopycnal according to the CTD casts of 2017. The axes are rotated so x is along and y is across the slope. Dots indicate CTD casts and the scales are such that point (0,26) is the deep water mooring line cast used as an example. Black contour lines are the depth of the isopycnal surface while color contours are temperatures on that surface.
The isopycnal slope from south to north would have a matching isobaric slope from north to south. Such a tilt forces a westward current, indicated by the overlay arrow. This is the Chukchi Slope Current.
Next, I will see if this current indicated by CTD survey and thermal wind exists throughout the year in the ADCP record.
Velocity Time Series
Above is the along slope velocity with time and depth such that red indicates into the page, or westward, and blue is out of the page, or eastward. Side markers indicate ADCP bin depth.
The overall structure indicates a strong westward flow near the surface and an eastward flow at depth. I identify these as the Slope Current and Shelfbreak Jet.
The lower gray bar indicates sea ice concentration in the area greater than 40%. The flow direction near the surface appears to reverse when the sea ice melts.
Above the velocity plot are wind velocity vectors, which are also rotated such that upward pointing vectors are along slope. Winds in December and September to the southwest and northeast force strong upper currents while weakening lower currents.
Based on these results, I added sea ice cover and wind to the list of possible forcing functions.
Velocity Variance Ellipse
Variance ellipses find the major and minor axes of a dataset. Above are the variance ellipses for the ADCP velocities in this study, with dark colors on surface and light colors at depth.
On the surface, the ellipses are wide and face northwest/southeast. This indicates surface currents are highly variable but mostly travel along the shelfbreak.
At depth, narrow ellipses facing partly onshelf indicate little changes in the west/east currents there over time.
The uppermost and lowermost few major axes look a little like Ekman spirals (explained below), so I added that to the list of possible forcing functions as well.
Forcing Functions
What causes the variability on the Chukchi Slope?
Sea Ice
First, I investigated sea ice as a forcing function.
Sea ice in my study area forms in December and melts in July, as seen via satellite imagery. The formation/melt cycle affects the water column temperature, as indicated in the single-mooring time series previously depicted. I looked closer at the effect with empirical orthogonal functions (EOF).
Empirical Orthogonal Function
An EOF deconstructs a time series into a “pattern” frozen in time and an amplitude time series that modifies the pattern. For each input time series, in this case each thermistor record, there is a mode that explains a certain percent of the variance in the original time series.
Not all modes are significant. In the table above, which represents the entire temperature dataset, only the first few modes are significant.
Sea Ice - EOF mode 1
Above is the result of my EOF analysis. On the top is the comparison of the EOF patterns (left) to the seasonal means (right). For each panel, dots represent temperature sensors and the dark gray contour is the seafloor along the mooring line. The bottom panels represent the amplitude time series and the open water percentage (reverse of the sea ice concentration for easier comparison to the amplitude time series). I define summer with open water greater than 60% and winter when there is less.
EOFs work in such a way that when the amplitude in e) is -1, the temperature on the mooring line is c) while an amplitude of +2 indicates temperatures like a) multiplied by 2.
The positive first mode pattern closely resembles the summer mean temperature while the negative first mode pattern resembles the winter mean.
The amplitude time series matches closely with the open water percentage, though the times when the regime switches from negative to positive versus open water to ice-covered are slightly different.
Overall, I conclude that 65% of my dataset’s temperature variance can be explained by the presence/absence of sea ice cover.
Wind Stress
After sea ice, I looked at the effect of winds on the velocity and temperature on the Chukchi Slope. Above are two sample times during the hydrographic cruises in 2016 and 2017.
In 2016, the wind stress is facing nearly the same way over the entire area. This is not the case in 2017, when winds are conflicted.
To examine the effect of winds, I correlated the time series at each location to the velocity time series.
Wind Stress - Correlation with Surface Current
Above is the complex correlation of the wind stress with the surface currents, with the 2000m isobath and ADCP location for reference.
Overall, the correlation is very low, with a maximum of 0.27 near the Bering Strait. Winds in the Bering Strait force an inflow of water, so this correlation is expected.
There is another correlation “hot spot” in the Beaufort Gyre. The relationship here is unclear and beyond the scope of my study.
Regardless, with such low correlations, I am hesitant to draw any real conclusions. It is possible that the wind stress curl, rather than the wind stress, would produce better correlations.
Ekman Spirals
Next, I looked at Ekman spirals.
“Ekman spiral” refers to the effect of stressors on currents in the Northern Hemisphere. Above is a schematic of the predicted velocities with depth under a surface wind forcing, with darker colors on the surface lightening at depth.
In an Ekman spiral, surface currents are forced by wind and the Coriolis force such that they are 45 degrees to the right of the wind. Each successive layer of water below the surface is deflected further to the right and decreases in velocity.
With this theory, researchers can predict what velocities in the water column should be.
Ekman Spirals - Progressive Vector
One such study comparing theory to real-world results was Chereskin (1995). Above is a modified version of their progressive vector diagram.
Progressive vector diagrams estimate where a particle would travel over time based on the velocities at a single point. I added overlay vectors in direct comparison to the theoretical Ekman spiral depicted above.
In the case of the California Current in the Chereskin (1995) study, Ekman dynamics appear to be upheld. This is not the case in my study.
Above is a progressive vector diagram derived from a subset of my own data on the Chukchi Slope. I chose a time when currents were constant to the west for an extended time to compare to Chereskin (1995). Darker colors indicate surface currents with lighter currents at depth. The surface progressive vector also includes the wind direction at that time.
Under an Ekman spiral, the above westward flow should be governed by wind to the south, but that is not always the case here. In addition, the overlay arrows indicating overall direction of the layers rotates both very little and in the “wrong direction” to be considered an Ekman spiral.
In my area of the Chukchi Slope, the Ekman spiral doesn’t appear to work.
Conclusion
Summary of Results
What did the Chukchi Slope look like in 2016 and 2017?
- Shallow, westward Chukchi Slope Current which transports mainly Pacific water
- Deep, eastward Chukchi Shelfbreak Jet which transports mainly Atlantic water
What causes the variability on the Chukchi Slope?
- Sea ice, as evidenced by the EOF mode 1
- Not wind stress, though it would be beneficial to look at the wind stress curl
- Not Ekman dynamics