Library Rights Statement

Transcription

1 Library Rights Statement In presenting the dissertation, Deep Variability in the Kuroshio Extension, in partial fulfillment of the requirements for an advanced degree at the University of Rhode Island, I agree that the Library shall make it freely available for inspection. I further agree that permission for copying, as provided for by the Copyright Law of the United States (Title 17, U.S. Code), of this thesis for scholarly purposes may be granted by the Librarian. It is understood that any copying or publication of this thesis for financial gain shall not be allowed without my written permission. I hereby grant permission to the University of Rhode Island Library to use my thesis for scholarly purposes. Andrew Dale Greene Date

2 DEEP VARIABILITY IN THE KUROSHIO EXTENSION BY ANDREW DALE GREENE A DISSERTATION SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY IN OCEANOGRAPHY UNIVERSITY OF RHODE ISLAND 21

3 DOCTOR OF PHILOSOPHY DISSERTATION OF ANDREW DALE GREENE APPROVED: Dissertation Committee: Major Professor DEAN OF THE GRADUATE SCHOOL UNIVERSITY OF RHODE ISLAND 21

4 ABSTRACT The Kuroshio Extension is the continuation of the Kuroshio western boundary current in the North Pacific Ocean. In 24 an array of current and pressure recording inverted echo sounders was deployed within the Kuroshio Extension as part the Kuroshio Extension System Study (KESS). One of the goals of KESS was to investigate the relationship between upper-ocean circulation patterns and the deep-barotropic currents. In order to understand the processes which couple the upper and deep ocean in the Kuroshio Extension it is helpful to first provide a comprehensive investigation of dynamical balances in the deep Kuroshio Extension. In the deep Kuroshio Extension, variability is generated locally and remotely. Like the Gulf Stream, meandering of the upper-baroclinic jet locally generates deep cyclones and anticyclones by stretching and squashing the lower water column. While this baroclinic stretching process dominates some events, there are other cases in which a seemingly spontaneous spin-up of an abyssal cyclone occurs without meander-induced baroclinic stretching. These strong deep eddies had cyclonic vorticities greater than.2f, where f is the Coriolis parameter, and are caused by topographic stretching when water columns are advected off isolated seamounts in the region. Strong incident currents drive water columns off seamounts to form cyclones with relative vorticity consistent with a layered potential vorticity conservation calculation. Besides locally-generated strong eddies, larger-scale and weaker deep cyclones and anticyclones were observed to propagate into the KESS region from the northeast. Daily snapshots of deep and upper ocean streamfunction demonstrated that as the deep eddies encountered the Kuroshio Extension, upper-ocean meanders steepened and deep eddies intensified. The joint intensification of the upper and deep circulations was also seen by strong energy and vertical coherence in the 3 6 day band near the Kuroshio

5 jet axis. Using complex empirical orthogonal function analysis, the phase speeds and wavelengths were calculated for the deep-pressure signals, identifying them as short barotropic topographic Rossby waves. Joint complex empirical orthogonal function analysis of upper and deep streamfunctions revealed that near the Kuroshio Extension axis, upper and deep signals were phase shifted laterally downstream, with the deep signal leading the upper. This arrangement is consistent with the joint baroclinic development patterns. Observations were then compared against an two-layer intermediate geostrophic model and the Ocean General Circulation Model for the Earth Simulator (OFES), both of which agreed with the KESS observations. Since OFES reproduced upper and deep pairs of cyclonic and anticyclonic features that propagated southwestward across the Kuroshio Extension, with phase speeds and intensities matching those observed in KESS, it was further utilized to help provide regional context to the KESS observations. Outside the KESS region, 3 6 day bandpass filtered bottom pressures were also consistent with Rossby-wave propagation on a combined planetary and topographic beta plane. This result also suggested that the likely origin of the waves was the Shatsky Rise. Furthermore, the meandering Kuroshio Extension was shown to be the likely generation source of the waves insomuch as the dominant Kuroshio Extension meander in the 3 6 day band has a zonal wavelength that can couple to the topographic Rossby wave field on the Shatsky Rise.

6 ACKNOWLEDGEMENTS First, I would like to thank my advisor Professor Randolph Watts for his neverending hopefulness and optimism, whether it be on research cruise in which there was one setback after another or on land analyzing a blender of data. He always kept me buoyed up at the most frustrating of times. He also taught me how to be a careful scientist and how asking the right questions is critical to being a successful scientist. Next, I can not thank Kathleen Donohue enough for her advice on an array of matters that concern a graduate student. Georgi Sutyrin is owed a large amount of gratitude for supplying me with a number of complex numerical simulations. I would also like to thank the Watts/Donohue/Wimbush/Park group for their always insightful commentary and advice regarding my research, it was probably single-handedly one of the best sources of constructive criticism available, anywhere. Nelson Hogg and Stephanie Waterman also provided invaluable commentary regarding the interpretation of my results. My committee members are also owed a large amount of gratitude for their suggestions and taking the time to read my dissertation. I also need to thank my office mates Magdalena Andres and Qiang Li for providing a great working environment over the years. Lastly, but certainly not least, my forever supportive and incredible wife Jessica Hiatt, who for the past 3 years has supported me and been a shoulder to complain on. It is because of her my aspiration to complete a Ph.D. was realized. iv

7 PREFACE This thesis is written in manuscript format suitable for publication in a scientific journal and consists of three separate chapters, all of which are in various stages of publication. Manuscript 1 was published in the Journal of Marine Research in 29. Manuscript 2 was recently submitted and Manuscript 3 is in preparation. Additionally, two appendices provide supplementary details regarding the principal instrument used in the thesis, the current and pressure equipped inverted echo sounder, data processing and techniques, and quality control analysis. v

8 TABLE OF CONTENTS ABSTRACT ii ACKNOWLEDGEMENTS iv PREFACE v TABLE OF CONTENTS vi LIST OF FIGURES ix MANUSCRIPT 1 Deep Cyclogenesis by Synoptic Eddies Interacting with a Seamount Abstract Introduction Observational data PV structure from KESS A case study and a puzzle Modeling the interaction of synoptic eddies with a seamount The barotropic model equations Numerical setup Results of simulations Discussion KESS observations Advection of cyclones by the synoptic field KESS vs Gulf Stream Cyclones Evidence of submesoscale filaments Conclusions vi

9 1.8 References MANUSCRIPT 2 Evidence of Vertical Coupling between the Kuroshio Extension and Topographic Rossby Waves Abstract Introduction Observations and Numerical Models Methods Observations of Deep and Upper Streamfunction Interpretation and Discussion Topographic Rossby Waves Conceptual Model Numerical process model OFES Model Conclusions References MANUSCRIPT 3 The 3 6 Day Topographic Wave Field in the Kuroshio Extension: Observations and the Ocean General Circulation Model for the Earth Simulator Abstract Introduction Model Inter-comparison between KESS and OFES data sets Dynamical interpretation TRWs in the 3-6 day band Kuroshio/wave coupling Summary vii

10 3.7 References APPENDIX A CPIES Processing and Techniques A..1 Travel Time Interpretation of τ Leveling and calibrating τ A..2 Leveling p, u and v A..3 Optimal Interpolation A..4 GEM A.1 Comparing CPIES products vs in-situ data A.1.1 Mapped τ versus in-situ τ A.1.2 Temperature comparisons A.1.3 Velocity comparisons A.2 References APPENDIX B Eddy Detection B.1 Results B.1.1 Mapping p and ζ with pressure gauges B.1.2 Mapping p and ζ with current meters B.1.3 Mapping p and ζ with pressure gauges and current meters Biblogrpahy viii

11 LIST OF FIGURES 1.1 Topography in the KESS region Daily maps of lower layer potential vorticity (colorbar at middle range is approximately f /H ). The deep pressure anomaly field is superimposed in solid grey (positive anomaly) and black (negative anomaly) contours (CI =.25 dbar). Three bold blue contours denote the 3 m, 5 m, and 7 m depth of the 6 o C isotherm, with the northernmost contour representing the 3 m isoline. The upper jet path had a steep trough at E during this interval, because the first crest at E had shifted north of its mean position. The baroclinic structure remains separated to the north of the strong cyclogenesis event near 33 N 146 E Semi-daily currents incident upon seamount A, averaged from the center out to a radius of 8 km. The black vectors denote the time period in which the case study (Section 3) was performed. The vertical dotted line at yearday 32 is at the start of the strong deep synoptic currents. The strengthening of deep currents coincides approximately with a regime shift from stable to unstable meandering in the upper Kuroshio jet ix

12 1.4 Quasi-geostrophic potential vorticity field from the barotropic model. Top row (bottom row) shows the cyclogenesis process for the case when the anticyclonic (cyclonic) part of the Rossby wave field is incident upon the seamount. Left, middle, and right panels show respectively elapsed time,.5, 1 period of the synoptic Rossby wave field with color-shaded contours, with yellow and red colors indicating high PV. Grey and black contours denote the streamfunction, where grey (black) indicates anticyclonic (cyclonic) circulation and the zero streamline between them has a solid white contour. Seamount at (,) km. The cyclonic feature near (-28, ) km in the top row or (-36, ) km in the bottom row started as a large PV anomaly that separated from the seamount and formed a cyclone Zonal section of non-dimensional vorticity. The solid and dashed black lines represent the potential vorticity initially (t = ) and after one Rossby wave period (t = T R ). Solid grey line denotes the relative vorticity anomaly after one Rossby wave period. The positive PV feature at X = -24 to -3 km contains entirely cyclonic vorticity, acquired from stretching as the PV anomaly was swept off the seamount x

13 1.6 Upper three panels: Eddy kinetic energy at 3 sites from bottom current measurements bandpass filtered between the inertial period and 8 days. We think of this band as submesoscale and suggestive of the presence of small scale variability such as filaments. Site B5 is representative of all KESS CPIES sites that were well-separated from seamounts and had negligible EKE in this band. Sites F4 and F5 were respectively 54 km and 69 km away from a seamount, and their EKE in this band increased starting yearday 35, shortly after the increase in strength of the deep mesoscale currents. Lower panel: Mean EKE, at each site, in the inertial to 8 day time band as a function of distance from the nearest seamount KESS observational array. Positions of CPIES are denoted by solid black diamonds. Open circles identify the tall current meter moorings locations. Superimposed is the mean altimetric sea surface height (cm) between using AVISO Rio5 as the mean field. Topography between 4 m is shaded in light grey colors. Sea surface height standard deviation > 31 and 39 cm is color shaded in yellow and salmon, respectively Relationship between travel time τ and geopotential anomaly φ. Hydrographic data is denoted by the black dots with superimposed cubic spline in gray xi

14 2.3 Left Panel: Variance preserving spectra of deep-streamfunction anomaly. Middle Panel: Variance preserving spectra of upper-streamfunction anomaly. In the left and middle panels the line color represents instrument position as indicated by dot color on the right hand panel. Superimposed in dashed lines is the mean position of the Kuroshio Extension jet axis along with its northern and southern boundaries, as determined by the CPIES array. The scale for ψu 2 is ten times that for ψ2 D Squared coherence (right panel) and phase (left panel) between deep and upper ocean streamfunctions. Cross-spectra were averaged in cpd frequency bands ( 3 6 day periods). Mean axis of the Kuroshio Extension is denoted by the dashed black line. Coherence and phase is masked for grid points greater 7 km away from an operating instrument An example of a deep anticyclone passing southwestward under a Kuroshio Extension meander crest. Deep streamfunction is color-shaded where anticyclones (cyclones) are represented by the orange (blue) colors. Superimposed in black lines is the upper-ocean streamfunction, with the axis of Kuroshio Extension is denoted by a white contour line. Four-day snapshots are shown from left to right. As the anticyclone encounters and interacts with the Kuroshio Extension the meander crest steepens.. 63 xii

15 2.6 Hovmöller diagrams of 3 6 day band-pass filtered bottom pressures (middle row), with superimposed 3 6 day band-pass filtered upperocean streamfunction (bottom row). Cyclonic (anticyclonic) upper-ocean streamfunction anomalies are denoted by the thin grey (black) contour lines. The Hovmöller section is along a NE-SW line connecting the northern and southernmost CPIES (top row). The white line in the bottom row represents the daily intersection-latitude of the Kuroshio Extension and Hovmöller line section. Gaps or large jumps signify times when the Kuroshio Extension s path is S shaped and has multiple intersection points. (Motivated by personal communication with Hogg and Waterman) First mode CEOF for 3-6 day band-pass filtered deep-ocean streamfunction. Mean Kuroshio Extension axis is represented by dashed black line. Amplitude and phase are masked for grid points greater 7 km away from an operating instrument. Upper left panel: Spatial amplitude. Upper right panel: Phase. Lower panel: Amplitude time series corresponding to 1st mode spatial amplitude. Note that the 1st, 2nd, and 3rd modes recovered 64%, 16% and 11% of the variance, respectively Same as Figure 2.7 except for 3 6 day band-pass filtered upper-ocean streamfunction. Note that the 1st, 2nd, and 3rd modes recovered 51%, 21% and 12% of the variance, respectively xiii

16 2.9 First-mode joint CEOF of 3 6 day band-pass filtered deep and upper ocean streamfunction. Amplitude and phase are masked for grid points greater 7 km away from an operating instrument. Upper left panel: Spatial amplitude of the deep streamfunction portion of the 1st mode joint CEOF is shown by color shaded contours. The upper-ocean spatial amplitude is represented by black contour lines. Upper right panel: Phase difference between of deep and upper ocean. The mean Kuroshio Extension pathway is denoted by dashed black line. Lower panel: Amplitude time series of the first joint mode. Note that the 1st, 2nd, and 3rd modes recovered 54%, 2% and 13.4% of the variance, respectively Top row: Barotropic topographic Rossby wave dispersion curves for 3 and 6 days. The grey-shaded region represents the wavenumber space consistent with 3 6 day topographic Rossby waves in the KESS region. Superimposed are the wavenumbers and standard deviation bars associated with the horizontal gradient of phase of the 1st mode CEOF of 3 6 day band-pass filtered deep streamfunction. Wavenumbers were averaged in 1.5 boxes centered on the position indicated by the colored dots in the bottom left panel. Bottom right panel shows the number of points included in each 1.5 box average. The mean Kuroshio Extension axis is represented with a dashed line in the bottom panels. (Motivated by personal communication with Hogg and Waterman) xiv

17 2.11 Top row: Tall current meter mooring locations. Middle row: First mode EOF eigenvector of 3 6 day band-pass filtered u, v currents as a function of depth. Bottom row: Amplitude time series corresponding to the first mode. The black and grey lines in the middle and bottom rows denote the u and v EOF components, respectively. First mode variance for each velocity component is shown in the upper-left hand corner of each panel in the middle row Idealized interaction between an upper-ocean jet and a deep eddy in a two-layer system. The top row is a plane view of a deep anticyclone represented by dashed lines and the upper jet indicated by solid black lines. The gray dashed line represents the cross-section depicted in the bottom rows. The bottom row shows cross-sections of the jet/anticyclone configuration with jet and anticyclone are represented by the solid lines and tall grey shaded cylinder, respectively. Left panel: Initial configuration of an eastward-flowing (into the page) jet with an anticyclone downstream an upper-ocean meander crest. Right panel: Configuration after the crest has steepened and locally the jet has shifted northward (positive y) as indicated by the grey line and the anticyclone has intensified due to the local squashing of the lower water column xv

18 2.13 Upper and deep streamfunction fields from numerical model. Black bold lines denote the upper-ocean jet. Solid (dashed) gray contour lines represent the anticyclonic (cyclonic) portion of the topographic wave field. The contour interval for the deep field is.25 dbar. The top and bottom rows show the interaction at elapsed time t = and 3 days, respectively. Initially the upper ocean jet was straight and during the interaction with barotropic TRWs the jet developed steep meander crests and troughs and the TRW amplitude intensified Same as Figure 2.6 except 3 6 day band-pass filtered bottom pressures and geopotential from OFES are included (left panels), with KESS observations on the right panels. All Hovmöller plots have a yearbase of 24 and the aspect ratio of each has been adjusted such that an equally sloped line on each represents the same phase speed Bottom-pressure variance, in units of squared centimeters, from one year of bottom-pressure measurements from the KESS array (top row), model year 24 from OFES (middle row), and 9 years of OFES model data (bottom row). The KESS CPIES locations are denoted by white dots. Grid points are masked where no data exist Same as Figure 3.1, except 3 6 day band-pass filtered bottom-pressure variance is shown xvi

19 3.3 Hovmöller diagrams of 3 6 day band-pass filtered bottom pressures from OFES (top row) and KESS (bottom row). The Hovmöller section is along NE SW line connecting the northern and southernmost CPIES (bottom left panel). The black dots represent CPIES locations. Hovmöller plots have a yearbase of 24 and the aspect ratio of each has been adjusted such that an equally sloped line on each represents the same phase speed First and second mode CEOF of 3 6 day band-pass filtered bottom pressures, top and middle rows respectively. The domain is similar to the KESS observational array. Left panels: Spatial amplitude. Right panels: Phase. Bottom row: Amplitude time series corresponding to 1st and 2nd mode spatial amplitude. Note that the 1st, 2nd, and 3rd modes recovered 35.3%, 25.8% and 1.7% of the variance, respectively First mode CEOF of 3 6 day band-pass filtered deep streamfunction from KESS array measurements. Amplitude and phase is masked for grid points greater 7 km away from an operating instrument. Upper left panel: Spatial amplitude. Upper right panel: Phase. Lower panel: Amplitude time series corresponding to 1st mode spatial amplitude. Note that the 1st, 2nd, and 3rd modes recovered 64%, 16% and 11% of the variance, respectively xvii

20 3.6 Squared coherence (left panel) and phase (right panel) between bottom pressure and upper-ocean geopotential from 1 year of KESS measurments (top row), OFES model year 24 (middle row), and 9 years of model data (bottom row). All spectra were smoothed using 5% overlapped windows spanning 186 days and then band averaged in 3 6 day period band, with a central period of 41.3 days. The black contour line outlines regions where the squared coherence exceeds the 95% confidence limit. Coherence and phase is masked for grid points that are shallower than 513-dbar. Positive phase would imply the deep ocean is leading the upper ocean. In the regions of large coherence the deep ocean is leading the upper ocean Same as Figure 3.3 except 3 6 day band-pass filtered upper-ocean geopotential from OFES (left panel) and KESS observations (right panel) are included. Cyclonic (anticyclonic) upper-ocean streamfunction anomalies are denoted by the thin grey (black) contour lines. The white line in the bottom row represents the daily intersection latitude of the Kuroshio Extension and Hovmöller line section. Gaps or large jumps signify times when the Kuroshio Extension s path is S shaped and has multiple intersection points Same as Figure 3.4 except the domain has been extended to (3 45 N, E. The 1st, 2nd, and 3rd modes recovered 31%, 18% and 12% of the variance, respectively xviii

21 3.9 Local topographic Rossby wave dispersion curves for 3 and 6 day periods. The grey shaded region represents the wavenumber space supported by the local 3 6 day Rossby wave field. Superimposed are the wavenumbers and standard deviation bars calculated from the horizontal gradient of the 2nd mode CEOF phase of 3 6 band-pass filtered bottom pressure. Wavenumbers were averaged in 1.5 boxes centered on the position indicated by the colored dots in the bottom right panel. Wavenumbers were only estimated at grid points where the 2nd mode CEOF amplitude exceeded.4. At the top of each subplot α and γ, the meridional and zonal components to topographic β, are listed. Lower right panel: Group velocity vectors associated with each (k, l) with superimposed smoothed topography with 1 m contour interval Same as Figure 3.9 except the TRW dispersion relation was Doppler shifted in regions where the average latitude of the grids points used to calculate (k, l) was between N An example of a typical eastward-propagating meander as it is about to encounter the Shatsky Rise. The black and white contour lines represent the geopotential field at t = t and t = t + 9 days, respectively. Superimposed is 1/3 gridded bathymetry used in the OFES simulations day snapshots of 3-day averaged 3 6 day band-pass filtered EKE. Superimposed in thin white lines is the Kuroshio Extension on the midday of the 3 day EKE average, the midday is shown at the top of each subplot in yearbase 24. Top row is representative of times in which the deep EKE is high near Shatsky Rise and the Kuroshio is meandering over the Shatsky Rise. The bottom row represents a straighter Kuroshio path near the Shatsky Rise and low EKE xix

22 day period Rossby wave dispersion circles for a flat (left panel) and sloped ocean bottom (right panel). For the sloped case, zonal topographic β is 5 times stronger than planetary β, i.e., γ = 5β. Superimposed on each dispersion diagram, as a red line, is the zonal wavenumber of a 5 km wavelength eastward-propagating meander First mode CEOF of 3 6 day band-pass filtered upper-ocean geopotential. Upper left panel: Spatial amplitude. Upper right panel: Phase. Lower panel: Amplitude time series corresponding to 1st mode spatial amplitude. Note that the 1st, and 2nd modes recover 25% and 15% of the variance, respectively Rossby wave dispersion circle for a 45 day period wave near E and N, i.e, northern portion of Shatsky Rise. In this region zonal and meridional topographic β are γ = 12.6β and α = 2.6β, respectively. The zonal wavenumber of the most frequently occurring meander in the 3 6 day band is denoted by the vertical red line. The two intersection points with dispersion circle represent 4 and 45 km waves km smoothed model bathymetry. Locations shaded grey represent regions in which the local Rossby wave field can couple to the dominant wavelength of 3 6 day eastward-propagating meanders in the Kuroshio Extension. Only the Shatsky Rise has topography suitable for Kuroshio/wave coupling. Only the regions between 34 and 38 N, i.e., where there are eastward-propagating meanders (Fig. 3.15), were used for solving equation xx

23 3.17 Top row: Correlation coefficient function between EKE at Shatsky Rise and the KESS region, where a positive time lag indicates that EKE at Shatsky Rise is leading KESS EKE. EKE at the Shatsky Rise (middle row) and the KESS region (bottom row). Each region is defined respectively as E, N and E, N. The two time series have.41 correlation coefficient at a 25 day time lag that is significant to the 99% confidence level A.1 Relationship between travel time τ and the depth of the 6 C isotherm Z 6. Hydrographic data is denoted by the black dots with superimposed cubic spline in gray. The rms difference between the spline fit and the data is shown in the top right corner A.2 Regional hydrocasts available in the KESS region. The grey (black) dots and bars denote float (CTD) profiles. The top panel shows the geographical locations of all the hydrocasts. On the bottom from left to right are histograms of the number of casts as a function of maximum pressure, year, and month A.3 Seasonal correction to travel time. Top row: Relationship between τ 25 and τ 25 14, where the black line is a cubic spline fit. Red (blue) colors represent summer (winter) months. Bottom row: Residual of the spline fit plotted as a function of yearday. The black curve represents the seasonal cycle in τ, calculated from a 1-day low-pass filter of 15-day averages of the residual with a 5% overlap A.4 Relationship between τ 14 and τ 4. Hydrographic data is denoted by the black dots with superimposed cubic spline in gray xxi

24 A.5 Correlation coefficient between groups of instruments as a function of separation distance for low-pass filtered τ (top row), high-pass filtered τ (middle row), and p (bottom row). Correlations were grouped into 1 km bins and averaged. The red line is a Gaussian function with the e-folding length scale denoted on the right hand side of each subplot. Error bars are the standard deviation within each 1 km bin A.6 Left panel: Temperature GEM contoured in thin black lines as function of pressure and τ 14. Superimposed is the uncertainty in the GEM temperature. Right panel: Temperature versus τ 14 at various pressure levels (top right of each subplot) with superimposed spline fit. A set of these curves from p = to p = 63 dbar make up the GEM (left panel). The rms difference between the curve and data is shown at the bottom left of each subplot A.7 Argo float and CPIES measured τ agree nicely. The rms difference between the two and predicted mapped τ error also agree well A.8 Functional dependence of the rms difference between τ CP IES and τ float on mapped tau error and closest distance from an operating instrument, top and middle rows respectively. In the top (middle) row rms differences were averaged in.25 ms (1 km) bins. In the top row, the grey line has a slope of one. In the bottom row mapped tau error is plotted against distance from an instrument. The errorbar for each bin is the standard deviation divided by the square root of the estimates within each bin A.9 Temperatures at 5 dbar derived from the CPIES/GEM technique agree well with Argo float temperatures. The grey line has a slope of one. Error estimates are also in good agree with rms differences xxii

25 A.1 Comparison between CPIES mapped temperatures (blue) and those directly measured by the moored profiler (red) at sites K2, K5, and K7. Temperatures were smoothed with a 4-day low-pass filter prior to comparison. The latitude and pressure level of each comparison is shown in the upper left corner, where the chosen pressure level had the minimum amount of data gaps for that site. Predicted temperature error and actual rms difference agree nicely A.11 Predicted temperature rms error profiles (grey line) at sites K2, K5, and K7 agree well the observed rms difference profiles (black). On average the predicted error is the same or slightly larger than the actual error A.12 Comparison of velocities at 2 m between CPIES derived (blue) and the upward-looking ADCP (red) velocities. The left (right) column shows the zonal (meridional) component of velocity. At the top of each subplot is rms difference and predicted error in velocity for each site B.1 The colored contours represent the normalized pressure field of the analytic eddies. Going across left to right the eddy e-folding radii are L = 3, 5, and 8 km, respectively. From top to bottom the center of the analytic eddies are offset, 3, and 45 km from a pseudo CPIES instrument (black dots). Instruments were spaced 88 km zonally along lines separated 88 km meridionally, diagonally offset xxiii

26 B.2 Top row: Fraction of the true pressure (left panel) and vorticity (right panel) fields recovered using a pseudo array of pressure gauges as a function of eddy e-folding size L and closest distance to a pseudo instrument D. Instruments were spaced in a pattern similar to the KESS array. The contour interval is.5. Bottom row: Average fractional pressure (left panel) and vorticity (right panel) field recovered as function of L B.3 Top row: Fraction of the true pressure (left panel) and vorticity (right panel) fields recovered using a pseudo array of current meters as a function of eddy e-folding size L and closest distance to a pseudo instrument D. Instruments were spaced in a pattern similar to the KESS array. Bottom row: Average fractional pressure (left panel) and vorticity (right panel) field recovered as function of L B.4 Top row: Fraction of the true pressure (left panel) and vorticity (right panel) fields recovered using a pseudo array of pressure gauges and current meters as a function of eddy e-folding size L and closest distance to a pseudo instrument D. Instruments were spaced in a pattern similar to the KESS array. Bottom row: Average fractional pressure (left panel) and vorticity (right panel) field recovered as function of L B.5 Mapping with combined pressure gauges and current meters is better than just mapping with pressure gauges (red line) or current meters (blue line) alone for all sized eddies. Differences between mapping with p + uv and only p and only uv is shown in red and blue respectively. The difference between mapping with only uv and p is shown in green. The top (bottom) row shows the differences for mapped pressure (vorticity). 142 xxiv

27 MANUSCRIPT 1 Deep Cyclogenesis by Synoptic Eddies Interacting with a Seamount by Andrew D. Greene 1 ; Georgi G. Sutyrin 2 ; D. Randolph Watts 3 Published in the Journal of Marine Research, May PhD Candidate, Graduate School of Oceanography, University of Rhode Island, Narragansett, Rhode Island. agreene@gso.uri.edu 2 Senior Marine Research Scientist, Graduate School of Oceanography, University of Rhode Island, Narragansett, Rhode Island. gsutyrin@gso.uri.edu 3 Professor, Graduate School of Oceanography, University of Rhode Island, Narragansett, Rhode Island. rwats@gso.uri.edu 1

28 MANUSCRIPT 1 Deep Cyclogenesis by Synoptic Eddies Interacting with a Seamount Andrew D. Greene, Georgi G. Sutyrin, and D. Randolph Watts Graduate School of Oceanography University of Rhode Island Narragansett, Rhode Island 1.1 Abstract Strong deep eddies with cyclonic vorticity greater than.2f were detected using an array of bottom current and pressure measurements in the Kuroshio Extension System Study (KESS) in Daily maps showed these deep eddies developed locally. As in the Gulf Stream, meandering of the upper baroclinic jet generates deep cyclones and anticyclones by stretching and squashing the lower water column. However, unlike the Gulf Stream, the smaller vertical stretching and greater water depth in the Kuroshio Extension limits the relative vorticity generated by this vertical coupling process to about.1f. In the deep Kuroshio Extension the strong cases of vorticity generation and cyclone development are related to stretching driven when water columns are advected off isolated seamounts in the region. The large observed values of relative vorticity are consistent with a straightforward calculation of deep layer potential vorticity conservation. A barotropic model is used to illustrate the topographic generation of cyclones by ambient currents in synoptic eddies. Positive potential vorticity filaments also develop during the cyclogenetic process with width L R = O(2 km), where L R is the topographic Rhines scale, and travel anticyclonically around the seamount. Observational evidence lends support to the existence of submesoscale filaments, insomuch as current 2

29 meter records near the flanks of seamounts exhibited bursts of eddy kinetic energy when bandpass-filtered between the inertial period and 8 days. 1.2 Introduction Synoptic eddies in the abyss are common in the ocean, especially beneath meandering western boundary currents (Shay et al., 1995; Watts et al., 1995). Here we present data from the Kuroshio Extension System Study (KESS) array. Recent analyses indicate highly variable deep velocities (> 25 cm s 1 ) and relative vorticities O(.2f ) which, unlike deep cyclones in the Gulf Stream, can not be explained alone by stretching effects caused by the main thermocline meandering nor by currents crossing large scale topography. Moveover, the deep barotropic portion of the jet below the Kuroshio Extension is only 5 cm s 1, which is too weak to generate such strong deep eddy variability (> 25 cm s 1 ) by simple meandering or barotropic instability processes. Therefore we search for other sources of deep flow variability. Besides the local generation of strong deep eddies, remotely generated eddies passed through the Kuroshio Extension region. High and low pressure centers repeatedly propagated into the region from the east with diameters typically 25 km and relative vorticities <.1f, i.e., larger and weaker than the strong cases of locally generated cyclones. As they drifted through the array they encountered the upper Kuroshio Extension jet, rings, and topography. Of these interactions the one that can consistently generate O(2%) vertical stretching is the interaction of the deep flow with seamounts. The sea floor in the Kuroshio Extension region has a gentle trend downward toward the southeast from m. Several seamounts in this region rise above this broad slope (Fig. 1.1). The height of the seamounts above the local bathymetry is 8 2 m with radii of 5 75 km. The strongest deep cyclonic eddies were detected offset from a seamount by 1 km or less. Therefore, we investigate the interaction of the deep flow 3

30 with the local seamounts. Topographically induced anticyclonic circulation is commonly observed over the seamounts (Genin et al., 1989). Previous studies suggest that background currents and ambient eddy variability govern the topographic interactions. One explanation of seamount-trapped flows relies on the Taylor-Proudman theorem: an isolated obstacle diverts flow in a rotating fluid around an isolated seamount and generates a column of trapped circulation (Greenspan, 1968; Hopfinger and Van Heijst, 1993). More formal explanations of anticyclones over topography are provided by equilibrium statistical mechanics (Merryfield et al., 21; Nycander and LaCasce, 24). The initial value problem for uniform flow incident on a small amplitude seamount has been investigated earlier numerically in a number of barotropic and stratified models. Using a three-dimensional model with nine levels and periodic boundary conditions, Huppert and Bryan (1976) found that starting from rest, the evolution of the flow redistributes vorticity in such a way that anticyclonic vorticity remains over the topographic feature, while water shed from the seamount induces a cyclonic vorticity anomaly. For sufficiently strong incident flow, the shed fluid is advected downstream forming a cyclonic eddy. If the incident flow is relatively weak, interaction between anticyclonic and cyclonic vorticity patterns traps the cyclonic eddy and the eddy remains in the vicinity of the topographic feature. They interpreted their results by using simple analytical calculations based on an approximate representation of their streamfunction in terms of line vortices, and suggested that observations by Vastano and Warren (1976) of an eddy in the vicinity of the Atlantis II Seamount and existence of the large amount of high frequency energy near the bottom of the ocean measured by the MODE experiment may be partly explained in terms of this mechanism of topographic cyclogenesis. The different typical responses of a uniform flow impinging on an isolated seamount were synthesized using barotropic 4

31 models with downstream boundary conditions which allowed vorticity to be advected smoothly out of the domain (James, 198; Verron and Le Provost, 1985). Similar results were obtained for a two-layer contour dynamical model for a tall seamount penetrating the upper layer (Thompson, 1993). Generally, unsteady flow over a sloping bottom is known to produce a bottomintensified mean flow with a correlation between streamfunction and topography and anticyclonic motion over bumps (Dewar, 1998). In contrast, upper layer variability can produce surface-intensified topographic circulation over a tall seamount (Sutyrin, 26). Such rectified flows are related to potential vorticity stirring which becomes strongly anisotropic over steep slopes where topographic Rossby waves prevent the flow from achieving the scale of the topography. This may lead to the formation of alternating jets with spatial scale about 2 km or less, estimated from the topographic Rhines scale L R = (UH/sf) 1/2, with typical eddy velocity U.1 m s 1, ocean depth H 4 km, topographic slope s 1 2, and Coriolis parameter f = 1 4 s 1 (Vallis and Maltrud, 1993). For a uniform background potential vorticity gradient, lateral mixing of potential vorticity across the background gradient tends to be inhomogeneous and leads to staircases, which Dritschel and McIntyre (28) argue is why persistent jets are commonplace in planetary atmospheres and oceans and why such jets tend to sharpen themselves when disturbed. Our observations in the Kuroshio Extension and these previous studies motivated us to consider what kind of variability can be generated by time-varying currents passing an isolated seamount. The Synoptic Ocean Prediction Experiment (SYNOP) was the first field experiment that clearly demonstrated the deep-flow fields beneath the Gulf Stream were organized into coherent vortices. SYNOP revealed that deep cyclones developed and intensified jointly with large amplitude Gulf Stream meander troughs. Case studies of steep meander events revealed that the production of positive relative vorticity in the deep ocean 5

32 could be accounted for by the local meander-induced stretching of the lower water column (Savidge and Bane, 1999; Howden, 2). In 24 the multi-institutional collaborative project KESS was launched. One of the goals of KESS has been to understand the processes coupling the baroclinic and barotropic circulation. One working hypothesis was that meandering of the Kuroshio Extension couples the baroclinic front to deep eddies, via baroclinic instability, which in turn steer the baroclinic front. While this holds true for a number of events in the near neighborhood of the baroclinic front, there are additional cases away from the baroclinic front in which a seemingly spontaneous spin-up of a deep cyclone occurs without the baroclinic stretching produced by jet meandering. Since the array was designed to resolve mesoscale processes we conducted potential vorticity analyses on the deep layer. All deep eddies generated within the KESS array with vorticity ζ >.15f in KESS were cyclones. This paper is focused on deep cyclogenesis due to the influence of isolated seamounts. We will show that the interaction between deep synoptic currents and seamounts can produce detached cyclones. Since mesoscale observational arrays can smooth over submesoscale features we utilize a semi-spectral barotropic model to study the topographically generated cyclones. The rest of the paper is organized as follows. Section 2 describes the observational data. Section 3 shows how we calculate lower layer potential vorticity. Section 4 presents numerical simulations of the barotropic Rossby waves interacting with an isolated seamount. Section 5 provides a discussion and Section 6 offers some conclusions and suggests future work. 1.3 Observational data Forty-six Current and Pressure recording Inverted Echo Sounders (CPIES) were deployed in the Kuroshio Extension region, 3 4 N and E, during May 24 6

33 July 26. Instruments were spaced on average 88 km zonally along lines separated 88 km meridionally, diagonally offset (Fig. 1.1). The Inverted Echo Sounder (IES) component measures the vertical acoustic travel time (τ) round trip from the sea floor to the surface and back. Variations in τ are mainly due to changes in depth of the thermocline. In this study τ is used with historical hydrography to estimate the 6 C isotherm depth Z 6, which in turn is used as a proxy for the base of thermocline. This method has been utilized with success in the North Atlantic Current (Meinen and Watts, 2), Subantarctic Front south of Tasmania (Watts et al., 21), and the Gulf of Mexico (Donohue et al., 26). Each IES was equipped with a pressure sensor (PIES) to measure deep pressure variations (p), the barotropic signal. Thirty-seven PIESs also had an Aanderaa TM Doppler current sensor tethered 5 m above the bottom (CPIES). All the measured variables, τ, p, u, and v, were processed to an hourly output interval, where u and v are the zonal and meridional velocities measured by the current sensor. In addition, there were 7 deep moorings with RCM-11 current meters, sampling at 2 m, 35 m, and 5 m, deployed across the Kuroshio Extension. All the data from the KESS array were low-pass filtered using a fourth order Butterworth filter with a three day cutoff period. This effectively removes the effects of the major tidal constituents, inertial oscillations, and surface waves while retaining the low-frequency mesoscale variability. Transients in the time series were reduced by chopping off 1 day (1/3 of the filtering period) at the beginning and end of each time series. Data gaps were filled by linear interpolation, so filtering could be carried out. Direct comparisons of CPIES estimated temperatures and velocities to measurements by the current meter moorings in the KESS array show good agreement (Donohue et al., 29 and Appendix A). Deep pressure and current measurements were used to generate optimally interpolated 1 (OI) daily maps of velocity, streamfunction, and relative vorticity ζ at 53 m. 1 Optimal interpolation is an objective analysis scheme used for data interpolation and mapping 7

34 Optimal interpolation was also utilized for mapping Z 6. CPIES that were deployed close to the flanks of seamounts recorded bursts of strong rectilinear currents with periods shorter than the ambient mesoscale variability. We believe these currents are associated with the topography and are highly localized, therefore they were not used in our OI mapping which maps horizontally nondivergent flow fields. But the individual current records will be used to provide evidence of filaments and submesoscale activity. 1.4 PV structure from KESS The mapped ζ fields revealed deep eddies with relative vorticity ζ values exceeding.2f, where f is calculated at 35 N, near the center of our array. Under the assumption that layered potential vorticity (PV) is conserved, and if initially the deep ζ, then the elevated relative vorticity in an eddy would have required a lower layer stretching of 2%. However, the ocean is so deep in the KESS region, m, that vertical excursions of the thermocline 5 m can only cause a 1-12% change in the lower layer thickness, with matching size (ζ/f o ). Moreover, observations demonstrated that these strong deep eddies developed with little to no baroclinic stretching. This quantifies the statement in the Introduction that the observed values of ζ cannot be explained by meandering of the baroclinic jet. We define the lower layer as the portion of the water column confined between the base of the thermocline and the ocean bottom. Variations in Z 6 represent vertical excursions of the base of the thermocline. Z 6 deepens 5-6 m across the Kuroshio (Bretherton et al., 1976). The optimally interpolated map minimizes the error between the estimate and the measurement and through the correlation functions enforces horizontal nondivergence. A Gaussian spatial correlation function was fitted to the data for τ and for p. The empirically derived correlation length scale for τ is 13 km (75 km) for temporal scales greater (less) than 6 days. The correlation length scale for deep pressures is 1 km. 8

35 Extension jet from north to south, and does not outcrop to the north. Lower layer thickness is H = Z Z 6 and on average H is 5 m, where Z is the bathymetry, version 8.2 (Smith and Sandwell, 1997), shown in Figure 1.1. Direct velocity measurements beneath the thermocline confirmed that deep vertical shear is weak. Between 2 m and 5 m the mean speeds differ by less than.3 cm s 1, therefore we treat the lower layers as barotropic in our PV calculations. The lower layer potential vorticity was calculated as where f is the planetary vorticity. PV = ζ + f H (1.1) A case study and a puzzle Daily maps of PV and pressure were generated to illustrate how deep barotropic eddies are generated by seamounts. Figure 1.2 illustrates the generation of a deep cyclonic vortex, where the colors are lower layer potential vorticity, the deep pressure field is represented by the black (low pressure) and grey (high pressure) contour lines, and the bold blue lines denote Z 6 at 3 m, 5 m, and 7 m, respectively from north to south side of the Kuroshio Extension. Between yearday an intense low pressure center develops near 33 N, 146 E. The cyclone had a pressure anomaly -.24 dbar ( -24 cm) at the center and maximum swirl speeds exceeding 31 cm s 1. At its most southern position, 33 N, the cyclone had a value of ζ equal to (.18 to.21)f o, where.3f o range in vorticity is the uncertainty estimate from OI mapped relative vorticity in the KESS array for a circular eddy with e-folding radius R L = 8 km. Uncertainties in mapped pressure and vorticity are discussed in Section 5a. Noteworthy is that during the days leading up to the deep eddy spin-up the observed baroclinic stretching was far too small to have produced the observed deep relative vorticity. This apparent (but unlikely) spontaneous spinup of a deep cyclone motivated us to investigate the effects of 9