sign in sign up
<<
Home , Argo (oceanography)

Proceedings of 21th Conference on Climate Variability and Change, Phoenix, 11-15 January 2009

7A.6 Detection of Thermohaline Structure and Meridional Overturning Circulation Above and Below the Surface

Peter C. Chu1), Charles Sun2), and Chenwu Fan1)

1) Naval Ocean Analysis and Prediction Laboratory Department of Naval Postgraduate School, Monterey, California 2) NOAA/NODC, Silver Spring, Maryland

Since rapid changes in MOC could have 1. Introduction implications for regional changes of climate, correlation analysis between our reanalyzed Ocean , , and currents datasets (3D ocean fields) and the surface available at the present time do not have atmospheric data (such as NCEP reanalyzed sufficient resolution to describe the variability in wind stress, air-ocean heat and moisture fluxes) Meridional Overturning Circulation (MOC). We will improve understanding of the physical use our unique data access, our strong experience mechanisms behind fluctuations in the in analysis of sparse and noisy ocean data, and thermohaline structure and MOC. our data system already in place at NOAA/NODC and the Naval Postgraduate 2. Atlantic MOC School (NPS), to produce and distribute three dimensional temperature, salinity, and velocity A major feature of the basin-scale circulation in fields using the Global Temperature-Salinity the North Atlantic is the existence of the robust Profile Program (GTSPP) and Argo profile and AMOC. Driven largely by the deep-water track data together with the Navy’s Master formation in the Labrador and Nordic , the Observational Oceanographic Data Set AMOC is the main mechanism for northward (MOODS), at spatial resolution equal and higher heat transport in the Atlantic Ocean. Cross- than the standard product (1o × 1o). The equatorial heat flux associated with the AMOC temporal and spatial resolution was improved by occurs through northward transport of warm merging with data from the Ocean Surface surface waters from the southern Atlantic Analyses – Real Time (OSCAR) derived and southward transport of cold deep waters by from satellite altimeter and scatterometer. Close the deep western (DWBC). partnership between NPS and NOAA/NODC Changes in the Labrador Cold water formation shows the impact of the temporally varying affect the MOC and subsequently the surface temperature, salinity, and velocity data on temperature (SST) in the tropics (Curry and operational monitoring of MOC, thermohaline McCartney, 1996; Broeker, 1997; Curry et al., structure, and associated rapid climatic change. 1998). More flexible and user-driven data processing and distributing system will be implemented, to During the World Ocean Circulation optimize data use by both the scientific and Experiment (WOCE) 1990-1998, nine operational communities. With the reanalyzed realizations conducted along “48oN” section three dimensional ocean fields for two decades (from the English Channel to the Grand Banks of (1990-2008), we indentified temporal and spatial Newfoundland) showed significant inter-annual variability of MOC and thermohaline structure. variability in climate-relevant key parameters such as heat and fresh water transports. With a phase lag of one to two years, the transport is almost linearly correlated to the changes of the dominant mode of low-frequent atmospheric variability in the North Atlantic, the North 1) Corresponding author address: Peter C. Atlantic Oscillation (NAO) [Bersh, (2002)]. Chu, Naval Postgraduate School, Monterey, Frankignoul et al. (2001), Taylor and Stephens CA 93943, email: [email protected] (1998), Rossby and Benway (2000), Volkov (2005) and others have documented an seems to be a reason of the spatial variability approximately two year lag between the shifts of along 48oN and a 1-2 years phase lag between the core and NAO events. NAO and variability of oceanic heat and freshwater transports in mid-latitudes (Brand and Hypothetically, such linkages can be explained Carsten, 2005). by traveling of oceanic perturbations generated by the atmosphere. From the theoretical point of To understand the AMOC variability due to view, Rossby waves significantly affect the anomalies in surface wind and/or large-scale thermohaline ocean circulation and forcing, it is necessary to use the planetary wave transport across the North Atlantic basin from dynamics with detecting Rossby waves their east to west at speeds of a few cm/s (depending vertical scales and structure from observations on latitude) (Gill, 1982; Rhines, 2004). It takes (Liu, 1999; Yang, 2000). The months or years to cross the North Atlantic signatures are clearly detected from satellite Basin. Several scenarios for ocean wave altimetry [for example, Chelton and Schlax teleconnections between mid-latitude and (1996); Osychny and Cornillon (2004)], SST tropical Atlantic were proposed in scientific [Cipollini et al. (1997); Hill et al (2000)] and literature (Doscher et al., 1994; Huang et al., color [Cipolloni e t al. (2001); Killworth et al. 2000; Hakkinen and Mo, 2002; Yang and Joyce, (2004)] in the North Atlantic that allows 2003 and others). However, the question of what estimating some propagating features of long scenario is of the most importance in transferring baroclinic Rossby waves but not their vertical decadal signals between the mid-latitude and structure. Tropical Atlantic is still unanswered. At present, observational support for detecting Numerous studies (mostly using results of the AMOC variability is quite poor. This is numerical modeling) demonstrate strong wave mainly due to lack of three dimensional ocean variability in the tropics and connections data (temperature, salinity, velocity) for the between this variability and large-scale entire Atlantic with sufficient temporal and atmospheric perturbations. After analyzing the spatial resolution. The lack of data causes less flow of Antarctic Intermediate Water along the capability in identifying long baroclinic waves equator in an idealized regional model of the in the Atlantic other than satellite observations tropical Atlantic Ocean, Jochum and Malanotte- (Brand and Carsten, 2005) although Rossby Rizzoli (2003) found that the flow is dominated wave signal should appear in various ocean by the Rossby wave activity related to the annual fields, such as velocity, temperature, salinity and cycle. Thierry et al. (2004) studied the deep biological substances at different depths. seasonal variability in realistic and simplified Following Killworth et al. (2004), baroclinic GCMs of the equatorial Atlantic Ocean and Rossby waves in the North Atlantic may induce found that the annual velocity fluctuations are vertical displacements for ocean surface up to 10 dominated by the lowest odd meridional mode cm, for bottom boundary of up to 50 Rossby wave packets. These Rossby waves are –100 m and for nitrocline up to 10 m. Below generated by the reflection of the directly wind the thermohaline, observed large perturbations of driven, shallow packets at the temperature and salinity also may be associated eastern boundary. These facts show the with the vertical structure of baroclinic Rossby importance of the Rossby and Kelvin wave waves. Thus, it is urgent to produce a quality dynamics in the inter-annual variability of mid- three dimensional dataset for temperature, latitude Atlantic circulation and in teleconnection salinity, and velocity dataset with sufficient (linkage) between tropical and mid-latitude inter- spatial resolution (1o × 1o) and temporal annual variability. Boning and Schott (1993) resolution (1-3 months) for the entire Atlantic for found deep current fluctuations with magnitude determining the AMOC variability. of 5 cm s-1 induced by the seasonal cycle of the wind stress and consistent with the long 3. Detection of MOC below the Ocean equatorial Rossby waves. Surface

Therefore, the propagation of long baroclinic Argo is an internationally coordinated, broad- Rossby waves generated in the eastern North scale global array of temperature and salinity Atlantic in conjunction with the meridional profiling floats (Fig. 1), and a major component movement of the zero line of wind stress curl

2 of the global ocean observing system. There are both temperature and salinity profiles, (c) sound- 3,000 temperature and salinity profiling floats speed profiles, and (d) surface temperature deployed worldwide. (drifting ). The measurements in the MOODS are, in general, irregular in time and space. Due to the shear size and constant influx of data to the Naval Oceanographic Office from various sources, quality control is very important. The primary editing procedure included removal of profiles with obviously erroneous location, profiles with large spikes (temperature higher than 35oC and lower than – 2oC), and profiles displaying features that do not match the characteristics of surrounding profiles such as profiles showing increase of temperature with depth. The MOODS contains more than 6 Fig. 1. Argo float (after the website: million profiles worldwide (Chu, 2006). http://scrippsnews.ucsd.edu/Releases/?releaseID=6 96). 4. Detection of MOC above the Ocean The Global Temperature and Salinity Profile Surface Program (GTSPP) is a cooperative international project to develop and maintain a global ocean Data collected from satellite altimeter, Temperature-Salinity resource with data that are scatterometer, and SST sensors can be used to both up-to-date and of the highest quality. It is a detect the MOC and Thermohaline structure. For joint World Meteorological Organization example, the satellite altimeter (JASON-1, GFO, (WMO) and Intergovernmental Oceanographic ENVISAT) and scatterometer data are analyzed Commission (IOC) project. Functionally, GTSPP and processed into near-real time ocean surface reports to the Joint Commission on current dataset on 1o × 1o resolution for world Oceanography and Marine (60o S to 60o N), which is posted online (JCOMM), a body sponsored by WMO and IOC as “Ocean Surface Current Analyses – Real and to the IOC’s International Oceanographic Time (OSCAR)”. It provides invaluable Data and Information Exchange committee resources online for various uses include large (IODE). GTSPP played a key role in the WOCE scale climate diagnostics and prediction, Upper Ocean Thermal Data Assembly Centre fisheries management and recruitment, and contributed to the final WOCE Data monitoring debris drift, larvae drift, oil spills, Resource DVD Version 3. The WOCE was fronts and eddies, plus opportunities for search ended in 2002, with some of its activities and rescue, naval and maritime operations. The continuing through a new program, CLImate methodology for OSCAR combines geostrophic, VARiability (CLIVAR). The GTSPP is also Ekman and Stommel shear dynamics, and a recognized by GOOS as an end-to-end data complementary term from the surface buoyancy management system for the oceanographic gradient. Interested readers are referred to the community. Many nations contribute data to the web site: (www.oscar.noaa.gov/index.html). GTSPP and without their contributions the project could not exist. Contributions to the data 5. Optimal Spectral Decomposition management portion of GTSPP are provided by (OSD) Australia, Canada, France, Germany, Japan and the USA. The quality control procedures used in The ocean observational data, no matter detected GTSPP were developed by Integrated Science below or above the ocean surface, is usually Data Management (ISDM). Readers can find noisy and sparse. Fig. 2 demonstrates typical useful information from the following website: observation coverage for two month observation http://www.nodc.noaa.gov/GTSPP/overview/abo period (October-November, 04). Beside, in utGTSPP.html. some cases major physical characteristics are not

explicitly identified by the data. For example, the The Navy’s Master Oceanographic evident current systems such as Gulf Streams, Observational Data Set (MOODS) is a Kuroshio, …, are not represented in the OSCAR compilation of ocean data observed worldwide data. consisting of (a) temperature-only profiles; (b)

3 trajectory data are used for the middle month such as March-May 04 for April 04, April-June 60°N 04 for May 04, …, and March-May 05 for April 05. Temperature data are much better because 50°N spatial gaps in float coverage are smaller, and the level of noise in the data is lower than for 40°N velocity observations. Therefore monthly T- fields are reconstructed in our study. Therefore, 30°N hereafter, ˆ ′ Uxnoise (,)00ttt=+ UxUx (,) (,) 0, ° 20 N ˆ ′ TtTTtnoise (,)xx=+ (,t) (,) x, (3)

10°N are treated as “noise”, which should be removed through the reconstruction process. 0°N Following Eremeev et al. (1993) and Chu et al. (a) 10°S (2003a, 2005, 2007) a velocity nowcast 70°W 50°W 30°W 10°W Fig. 2. ARGO subsurface tracks of floats parked at Uxnow(,) o t for quasi-geostrophic currents at 1000m and 1500m in October-November, 2004 any point of the area of interest Ω is written by (after Chu et al., 2007). the form of parameter-weighted sums of the

harmonic Z ()x and basis Ψ ()x functions An optimal spectral decomposition (OSD) io k 0 method was developed to overcome such as weaknesses in ocean observational data. Let S Ux(,)tatZ=×∇ ()[ k ()] x velocity U and temperature T fields be now00∑ s s s=1 decomposed within the observation period tobs as (4) follows, K +×∇Ψbt()[kx ( )]. ˆ ′ ∑ kk0 Ux(,)tt=+ Ux () Ux (,) + Ux (,t) + U (,) x t, k =1 TtT(,)xxxxx=+ () TtTtT (,) +ˆ (,) +′ (,) t. Similarly, a temperature field is represented by a sum of parameter-weighted basis functions (1) Ξ ()x (Chu et al., 2004), Here, [ Ux(), T () x ] are time averaged velocity m 0 M and temperature fields; [ Ux(,)t , Tt(,)x ] T,tT(xx )=+ () ct ()() Ξ x. (5) now00 cl∑ m m 0 represent seasonal variability and dynamical m=1 processes with temporal scale, τ ≤ tobs; The spectral coefficients as(t), bk(t), cm(t) in Eqs. Uxˆ (,)t , Ttˆ(,)x ] represent faster, mesoscale (4) and (5) are functions of time, k is the vertical unit vector positive upward, ∇=(,)∇∇ is variability with characteristic temporal ϕ λ scaleτ << t ; and [Ux′ ( , t) ,(,)Tt′ x ] are the obs the horizontal gradient operator. Tcl ()x is the measurement noises, x = (,,)ϕ λ z is a right- climatic temperature field from World Ocean handed coordinate system with ϕ the latitude, Atlas (Locarnini et al. 2005). λ the longitude, and z positive upward. Where (S, K, M) are the truncated mode numbers. Our goal is, using irregularly spaced and noisy The harmonic functions { Zio()x } are the Argo float observations collected within τ obs =14 solutions of the horizontal Laplace equation; and months (between March 2004 and April 2005), to reconstruct the basis functions { Ψ k ()x 0 } { Ξm ()x 0 } are the solutions of the horizontal Poisson equation with Uxnow(,) ott=+ UxUx () o (,) o , (2a) appropriate boundary conditions. Here, we list TtTTt(,)xxx=+ () (,), (2b) now 000 the equations for determining { Ψ ()x }, at 1000 m depth, i.e., k 0

x0 ==−(ϕλ , ,zm 1000 ) . To reduce measurement errors and noises for velocity calculation, original three-month float

4 ∇Ψ2 =−λ Ψ The subtropical gyre has two western boundary hm mm, currents - the Gulf Stream (off the east coast of

ni∇Ψhm|0,Γ = (6) the U.S.) and the North Atlantic Current (off the east coast of Newfoundland/Flemish Cap). The mM= 1, 2,..., , North Atlantic Current makes a major bend where Γ (z) is the lateral boundary, offshore at the southern entrance to the Labrador 22222 ∇≡∂∂h //x +∂∂y , and n is the unit vector Sea (called the “northwest corner”) and then extends eastward into the mid-Atlantic and then normal to Γ()z . The basis functions { Ψ } are m turns northward into the subpolar region as the independent of the data and therefore available Subarctic Front. The subtropical gyre also has an prior to the data analysis. The harmonic eastern boundary current called the Canary Current. The wind-driven subtropical gyre is functions { Zio()x } are the solutions of the largely gone by about 2000 m depth. horizontal Laplace equation. After the harmonic functions { Zio()x } and basis functions The deep-water formation in the Labrador and { Ψ ()x } { Ξ ()x } are given, the Vapnik Nordic seas drives the Meridional Overturning k 0 m 0 Circulation (MOC). It transports cold deep water (1982) variational principle is used to determine southward mainly by the Deep Western the optical spectral truncation (Kopt, Mopt) in (4) Boundary Current (DWBC), and transports p and (5) from a series of velocity ( Uobs ) and warm surface water northward from the southern Atlantic Ocean by the cross-equatorial currents. p p temperature ( Tobs ) observations at xo . For the Changes in the Labrador Cold water formation North Atlantic Ocean circulation and affect the MOC and subsequently the sea surface thermohaline structure using the Argo/GTSPP temperature (SST) in the tropics, and in turn data, we find that affect the . SK==2, 20, M = 38 . (8) opt opt There is much evidence that the dominant mode The spectral coefficients, of the low-frequency atmospheric variability in t t the North Atlantic, the North Atlantic Oscillation a,= [aa1 ,...,S ] b = [bb1 ,...,K ] , opt (NAO), affects the Labrador Cold Water (LCW) t c = [cc ,..., ] , (8) formation and ocean heat transport variability. 1 M opt are estimated using an appropriate variation For example, during the World Ocean Circulation Experiment (WOCE) from 1990 to method. Here, the superscript ‘t’ shows the o transpose operator. The OSD method has two 1998 nine realizations of the so-called “48 N” important procedures: optimal mode truncation section along the line between the English and determination of spectral coefficients. After Channel and the Grand Banks of Newfoundland the two procedures, the generalized Fourier showed significant inter-annual changes in the spectrum (1) is used to provide data at regular climate for relevant key parameters of the large- grids in space and time. Readers may find scale circulation in the North Atlantic, such as detailed description on accuracy of the OSD heat and fresh water transports. With a phase lag method from Chu et al. (2007). of 1-2 years, the transports are near-linearly correlated to the change of the NAO index [Bersh, 2002]. Frankignoul et al. [2001], Taylor 6. Mid-Depth North Atlantic and Stephens [1998], Rossby and Benway Circulation and Thermal Field [2000], and Volkov [2005] have documented an approximately 2 year lag between the shifts of 6.1. Background the Gulf Stream core and the NAO events.

Basin-scale circulations in the North Atlantic are Although the North Atlantic is a subject of driven by surface winds and heat/freshwater intensive theoretical and observational studies fluxes. The trades (easterlies) and the mid- for many years, quantitative description on the latitude westerlies are the prevailing winds north large-scale mid-depth circulation and thermal of the tropics. They drive the subtropical gyre in field especially the temporal variability is quite the mid-latitude North Atlantic through Ekman limited. There are several attempts to construct convergence and subpolar gyre in the northern the large-scale mid-depth circulation pattern for side of the westerlies through Ekman divergence. regional seas in the North Atlantic with higher

5 resolution using subsurface float (SOFAR, 1000 m and 1500 m and grouped together to RAFOS, ALACE, PALACE, SOLO) trajectories represent the mid-depth (1000 m). This neglects directly or combining with hydrological the vertical shear. Second, the vertical shear observations [Lavender et al., 2000; 2005; Zhang causes increase or decrease of the distance et al., 2001; Bower et al., 2002; Kwon and Riser between the points of ascending from and diving 2005] but for the whole North Atlantic using to the parking depth. Third, the sequence of float climate data [Schmitz and McCartney, 1993; trajectory segments only approximates the real Reid, 1994; Chu, 1995; Lozier et al., 1995]. Lagrangian paths. Fourth, preliminary computations (not included here to be published The Argo float data are used to determine the in a separate paper) show that high resolution spatial-temporal structure of the basin-scale elements of circulation in the western North circulation and thermal field in the North Atlantic, such as the northern re-circulation gyre Atlantic basin (4oN-65oN) at mid-depth (1000 and the Deep Western Boundary Current m), to estimate the kinematical (DWBC) are also revealed by the Argo floats. characteristics of the circulations, and to For example Fig 3 clearly shows the existence of calculate the horizontal heat flux. To do so, the DWBC. However, such a resolution is not recently developed OSD method is used. Our available for the whole North Atlantic. The results are not inconsistent with the bulk of the high-energetic mesoscale eddies and narrow earlier studies, but the results presented here boundary currents are classified as “noise” and have better spatial and temporal coverage than removed from the analysis. Fifth, there are large most previous studies. The mid-depth (1000 m spatial gaps in Argo float trajectories. The depth) currents and temperature were boundary consists of three segments: Γ and reconstructed with the same spatial resolution of 1 o o 1 ×1 in the whole North Atlantic excluding the Γ2 approximate the 1000 m isobath and the near-equatorial area south of 4oN. Azores Plateau; Γ3 is an open boundary along o o 6.2. Circulations 4 N latitude. This domain is extended to 10 S ' ( Γ3 ) with zero tangential velocity and diffusion The mid-depth mean velocity (April 2004 to flux. March 2005) can be calculated from the Argo float track data using the bin technique. The 60°N arrows are the mean velocities averaged over x appropriate (4o ×4o) bines. The circulation 2 ° patterns, robust to variations of bin sizes, show 50 N x 1 well-known circulation features identified from 40°N Γ early analyses on the RAFOS float trajectories in 2 separate regions of the North Atlantic. 30°N The measurement cycle of an Argo profiling float includes four stages: ascending, surface 20°N Γ drifting, diving and deep drifting. The Argo float 1 can only get its position fixings while it ascends 10°N Γ (a) to the sea surface. The vector between two 40 cm/s 3 consecutive surface positions during the deep 70°W 60°W 50°W 40°W 30°W 20°W 10°W drifting divided by the time interval is taken as the mid-depth velocity vector. When the Argo Fig.3. Circulation velocities (tiny arrows) estimated float is diving, ascending and drifting below the from the original ARGO float tracks at 1000 m for sea surface, no data can be transmitted to the Dec 2003–Mar 2004. The figure scale is given for ground stations in real time. Velocity field after tiny arrows. Red arrows are circulation velocities the first step analysis shows noisy circulation obtained by averaged over appropriate bins (red patterns (Figs. 2 a, b) with large spatial gaps lines). The magnitude of these arrows are (from 230 km to 800 km). multiplied by scale equaled to 20 for better visualization of schematic circulation patterns. Γ , Γ and Γ are boundaries of the Uncertainty in the Argo float data causes errors 1 2 3 in the velocity field. First, the data extracted computation area. from the floats parking at two different levels:

6 The reconstructed annual mean circulation (April again to southeastward in October 2004 (Fig. 2004 – March 2005) is characterized by four 5d). Evidences for this result also can be found major gyres (Fig. 4): tropical double gyres (south in Elmoussaoui et al [2005], Fratantoni and of 20oN), anticyclonic subtropical gyre (20oN to Richardson [1999] and others. Clearly, the about 50oN), and cyclonic subpolar gyre (North reconstructed fields have too lower resolution to of 50oN). be used for the analysis of narrow boundary currents, such as IWBC. Therefore, fall-winter reverses of IWBC from northwestward to southeastward which may be observed south of 15oN [Stramma and Schott, 1999], are not explicitly extracted from our results. However, Chu et al. (2007) found the Rossby wave propagation in the tropical North Atlantic using the same ARGO track data.

Fig. 4. Mean circulation at 1000 m depth from April 2004 to March 2005 computed using the OSD method.

6.2.1. Tropical Gyre

The tropical gyre is an elongated cyclonic gyre with velocities in a core of less than 2 cm/s and Fig. 5. Evolution of circulation at 1000 m depth in does not correspond to the schematic flow 2004 for (a) January, (b) April, (c) July, and (d)) diagram for 500 to 1200 m proposed by Stramma October 2004. and Schott [1999] (see, also, Schmid et al. 2005). However, a spatial pattern corresponding to a 6.2.2. Subtropical Gyre cyclonic flow approximately around the center of Guinea dome (12oN, 22oW) is clearly seen in The North Atlantic subtropical gyre includes the Fig. 4. That agrees, for example, with Gulf Stream, which becomes the North Atlantic Elmoussaoui et al. (2005) who observed the Current at about 40o N and 45oW and flows quasi-permanent cyclonic flow of the Guinea northeasterly across the North Atlantic. It re- dome at the same depth and deeper. circulates at the west coast of England and splits into southwestward flowing current along the Seasonal evolution of the circulation in 2004 mid-ocean range and the southward flowing (Fig. 5) indicates a superposition of eddy-wave- Canary Current along the west coast of Spain, like perturbations (mainly cyclonic rotation) Portugal, and North Africa. The mean Canary propagating northwestward, and westward from Current is weak at 1000 m depth and merges the eastern Tropical North Atlantic. These with the North Atlantic Equatorial Current, thus perturbations affect the mid-depth flow along the completing the North Atlantic subtropical gyre African coast and probably the Intermediate (Fig. 5). The subtropical gyre contains three Western Boundary Current (IWBC). That results evident cyclonic eddies with the west one. The in seasonal reversals for boundary currents. The subtropical gyre has evident temporal variation currents are northwest near the African coast up at 1000 m depth. to 16oN in January 2004 (Fig. 5a) with a maximum speed of about 3 cm/s. As the 6.2.3. Subpolar Gyre cyclonic eddy propagated northwestward, the currents reversed to southward and The subpolar gyre of the North Atlantic is a southwestward in April, changed southward in region of complex dynamics, playing a key role April 2004 (Fig. 5b), and reversed the direction in the variability of climate. Its currents are

7 forced by buoyancy contrasts and overflows Temporal evolution of the temperature field at from neighboring seas, probably as much as by mid-depth shows a warm zone appears in the the wind. Because of the low stratification, mid-latitudes with around 8oC in the western topography steers the currents, even in the upper North Atlantic (23oN - 35oN) and around 10oC in layers. The different water masses, although well the eastern North Atlantic (30o N-50oN). This defined by the precision of hydrographic warm zone is sandwiched by two cold zones measurements, have salinity contrasts small with 5oC – 6oC in the tropical one, and colder enough that numerical models have difficulty to than 4oC in the polar one. The most evident maintain them. feature is the formation of a warm eddy in the western mid-latitudes. This eddy was formed in Fig. 5 shows that the fastest flows are along the December 2004 and sustained until February northern and western boundaries of the basin and 2005. near Flemish Cap, including the subsurface signature of the northward-flowing North 6.4. Error Statistics Atlantic Current. Flow around the Reykjanes Ridge and the local acceleration through the The reconstructed monthly temperature field Charlie Gibbs illustrate apparent T is evaluated at regular grid points. We topographic control, while reversing zonal flows now interpolate Tnow to the observational points for exist just north of the Azores Plateau. that month and compute the residuals,

TT' =− T. (7) 6.3. Temperature Field obs now To show the statistical characteristics of T’, the The mid-depth annual mean temperature field North Atlantic is divided into 4 presumably (April 2004 to March 2005) can be calculated dynamically different sub-basins: tropical region, from the Argo profiling data using the OSD western mid-latitudes, eastern mid-latitudes, and technique (Fig. 6). The thermal field shows well- polar region. The error histograms of T’ (Fig. 7) known features identified from early analyses on show non-Gaussian distribution since the the for the North Atlantic. kurtosis is much higher than 3. The error The annual mean temperature field from April standard deviation is around 0.6oC. The error T’ 2004 to March 2005 identified from the Argo is positively skewed with skewness ranging from profiling data is quite similar to the 0.86 to 1.00. The bias (mean T’) is quite small climatological mean temperature field except a from 0.05oC to 0.07oC. cold eddy occurs in the western tropical North Atlantic (10o–20oN) with the temperature cooler o 6.5. Mid-Depth Long Baroclinic Rossby than 5 C. Waves 60°N Identification of long Rossby wave of tropical North Atlantic at mid-depth (~1000 m) from the

50°N 4 Argo track data is taken as an example to demonstrate the usefulness of the OSD method 5 to reanalyze sparse and noisy ocean data (Chu et ° 10 40 N 6 9 al., 2007). Argo float data (subsurface tracks and 7 temperature profiles collected from March, 04 8 30°N 8 through May, 05) are used to detect signatures of long Rossby waves in velocity of the currents at 7 1000 m depth and temperature, between the 20°N 6 ocean surface and 950 m, in the zonal band of 5 4oN -24oN in the Tropical North Atlantic. 10°N Different types of long Rossby waves (with the 5 characteristic scales between 1000 km and 2500

0°N km) are identified in the western [west of the 70°W 60°W 50°W 40°W 30°W 20°W 10°W Mid-Atlantic Ridge (MAR)] and eastern [east of the MAR] sub-basins. Fig. 6. Annual mean temperature fields at 1000 m depth computed from the ARGO profile data Current velocities computed along the original (April 2004 to March 2005). (non-smoothed) Argo tracks in November-

8 December, 2004 are shown in Fig. 7 as red A strong contribution from intensive eddies, arrows. such as that shown in inset B, narrow jets and 30°N 20 cm/s 5 cm/s measurement errors is clearly identified here. Visually, the velocity pattern corresponding to the original data looks quite chaotic, and there 20°N B are 500-600 km spatial gaps in observation coverage. To understand the reconstruction skill 60 A for such data we applied three criteria: (1) the 10°N 40 formal mean square error (the reconstruction 20 error) computed by the “laminar ensemble” 0 (a) 0 45 90 135 180 0°N technique (Turchin et al., 1971), (2) statistics of 70°W 60°W 50°W 40°W 30°W 20°W 10°W angle (α ) between the reconstructed and 30°N 5 cm/s 5 cm/s observed velocity at float locations, and (3) stability degree of the reconstructed snapshot on

20°N observation sampling (Perry and Chong, 1987).

60 7. Converting Idealized Currents into ° A 10 N 40 Realistic Currents 20 0 (b) 0 45 90 135 180 Near-real time ocean surface currents derived 0°N 70°W 60°W 50°W 40°W 30°W 20°W 10°W from satellite altimeter (JASON-1, GFO, 30°N o o 5 cm/s 5 cm/s ENVISAT) and scatterometer data on 1 × 1 resolution for world oceans (59.5o S to 59.5o N) posted online as “Ocean Surface Current ° 20 N Analyses – Real Time (OSCAR)”, provide invaluable resources online for various uses 60

° A include large scale climate diagnostics and 10 N 40 prediction, fisheries management and 20 recruitment, monitoring debris drift, larvae drift, 0 0 45 90 135 180 (c) 0°N oil spills, fronts and eddies, plus opportunities ° ° ° ° ° ° 70°W 60 W 50 W 40 W 30 W 20 W 10 W for search and rescue, naval and maritime 30°N 5 cm/s 5 cm/s operations. The methodology for OSCAR combines geostrophic, Ekman and Stommel

20°N shear dynamics, and a complementary term from the surface buoyancy gradient.

60

° A A major weakness of the OSCAR dataset is its 10 N 40 inability to represent the currents near the lateral 20 boundary. The most evident western boundary 0 (d) 0 45 90 135 180 0°N currents such as the Gulf Stream and Kuroshio 70°W 60°W 50°W 40°W 30°W 20°W 10°W are missing (Fig. 7b). The OSD method is used to reconstruct the OSCAR data. After the OSD Fig. 7. Sensitivity of the reconstructed circulation analysis, the reconstructed OSCAR data show patterns to filtration of the original data: OSD is realistic surface circulations including western applied to (a) the original data (November- boundary currents such as Gulf stream, December, 04); (b) the original data (November- Kuroshio, Brazilian Currents, Somali Currents, December, 04) filtered with a 2-month window; (c) the original data (October-December, 04) filtered and eastern boundary currents such as California with a 3-month window; (d) the original data Currents, Peru Currents, etc. (Fig. 7a). (October-December, 04) filtered by 4o × 4o bin averaging; Blue and red arrows correspond to the With the establishment of 4D (time and space) reconstructed circulation and original/filtered data. synoptic temperature, salinity, and velocity For α -histograms (inserts A) the x-axes is the (STSV) data, the MOC/ thermohaline structure angle α and the y-axes is the number of as well as their variability can be effectively comparisons (%), as described in subsection 4.3 identified. (after Chu et al., 2007).

9 OSD smooth data 2007 Jun 15

o 0.8(m/s) 60 N

30oN

0 Latitude

30oS

60oS

60oE 120oE 180 120oW 60oW 0

Original Data 2007 Jun 15

60oN 0.8(m/s)

30oN

0 Latitude

30oS

60oS

60oE 120oE 180 120oW 60oW 0 Longitude Fig. 7. Ocean surface currents on 15 June 2007: (a) after the OSD analysis, and (b) from original OSCAR data. which will be primarily based on the GTSPP- 8. Conclusions Argo, OSCAR, and Navy’s MOODS data. A more flexible and user-driven data processing Our main objective is to provide the scientific and distributing system will be implemented, to and operational communities with three- optimize data use by both the scientific and dimensional ocean fields (temperature, salinity, operational communities. and velocity) that have higher resolutions and better coverage than any products available. With 3D STSV data, temporal (seasonal and With our unique access to the baseline GTSPP- inter-annual) and spatial (various scales) Argo, OSCAR, and Navy’s MOODS data, our variability of AMOC can be identified through strong experience with the data processing, and quantifying the derived variables such as the our established near-real-time and online data meridional overturning (MO) streamfunction, distribution system in NODC, we plan to heat storage, meridional heat transport, etc.. The produce and distribute this 3D synoptic physical mechanisms causing the AMOC temperature, salinity, and velocity (STSV), variability can be determined through correlation analysis between the STSV data (including Chu, P.C., L.M. Ivanov, T.M. Margolina derived variables) and surface atmospheric data and O.V. Melnichenko, 2002: On probabilistic or investigation of the role of long baroclinic stability of an atmospheric model to differently Rossby waves (longer than 500 km) generated in scaled perturbations, J. Atm. Sci., 59, 19, 2860– the eastern Atlantic on the AMOC variability. 2873. Chu, P.C., L.M. Ivanov, T.M. Margolina, The impact of the AMOC variability on rapid T.P. Korzhova, and O.V. Melnichenko, 2003a: climate change can be evaluated through the Analysis of sparse and noisy ocean current data correlation analysis between STSV data the using flow decomposition. I. Theory, J. Atmos large-scale atmospheric variability (characterized Oceanic Tech., 20, 478 - 491. by NAO). Chu, P.C, Ivanov, L.M., Margolina, T.M., Korzhova, T.P., and O.V. Melnichenko, 2003b: With the gridded STSV data, we are analyzing Analysis of sparse and noisy ocean current data the variability and structure of the AMOC and using flow decomposition. Part 2. Applications explore the mechanisms contributing to AMOC to Eulerian and Lagrangian data. J. Atmos. variability using the new 3D STSV data together Oceanic Tech., 20, 492–512. with the NCEP surface atmospheric data. We Chu, P.C., L.M. Ivanov, and T.M. plan to work closely with the Naval Margolina, 2004: Rotation method for Oceanographic Office and NOAA/NCEP to reconstructing process and fields from imperfect apply the STSV data for quantitative assessment data. Int. J. Bifur. Chaos, 14 (8), 2991-2997. on the AMOC variability and operational Chu, P.C., L.M. Ivanov, O.V. monitoring of the AMOC variability and in turn Melnichenko, 2005 a: Fall-Winter Current the rapid climate change. Reversals on the Texas-Louisiana . J. Phys. Oceanog., 35, 902-910. Acknowledgments Chu, P.C., L.M.Ivanov, and T.M. Margolina, 2005 b: Seasonal Variability of the NOAA, NODC, ONR, and the Naval Black sea chlorophyll-a. J. Mar. Sys., 56, 243- Oceanographic Office sponsored this research. 261. Chu, P.C., L.M. Ivanov, O.V. Melnichenko, and N.C. Wells, 2007: On long baroclinic References Rossby waves in the tropical North Atlantic Bersh, M., 2002: North Atlantic observed from profiling floats. Journal of oscillation-induced changes of the upper layer Geophysical Research, 112, C05032, circulation in the northern North Atlantic Ocean. doi:10.1029/2006JC003698. J. Geophys. Res., 107, C10, 20-1 –20-11. Cipollini P., D. Cromwell, M. S. Jones, G.D. Boning, C.W., and F.A. Schott, 1993: Quartly and P.G. Challenor, 1997: concurrent Deep currents and the eastward salinity tongue in altimeter and infrared observations of Rossby the equatorial Atlantic: results from an eddy- wave propagation near 34oN in the Northeast resolving, primitive equation model. J. Geophys. Atlantic, Geophys. Res. Let., 24, 8, 8889-892. Res., 98 (C4), 6991-6999. Cipollini, P., D. Cromwell, P.G. Boning C.W. and J. Kroger, 2005: Chanselor and S. Raffago, 2001: Rossby waves seasonal variability of deep currents in the detected in global ocean color data, Geophys. equatorial Atlantic: a model study. Deep-Sea Res. Lett, 28, 2, 323-326. Research, I, 52, 99-121. Curry, R.G. and McCartney, 1996: Brandt, P., and C. Eden, 2005: Annual Labrador Sea Water carries northern climate cycle and inter-annual variability of the mid- signal south. Oceanus, 39, 24-28. depth tropical Atlantic Ocean. Deep-Sea Curry R.G., M.S. McCartney and T.M. Research I, 52, 199-219. Joyce, 1998: Oceanic transport of subpolar Broecker, W.S. ,1997: Thermohaline climate signals to mid-depth subtropical waters. circulation , the Achilles’s heel of our climate Nature, v. 391, 5, 575-577. system: Will man-made CO2 upset the current Doscher, R., C.W. Boning and P. balance? Science, 278, 1582-1588. Herrmann, 1994: Response of Circulation and Chelton, D.B., and M.G. Schlax, 1996: heat transport in the North Atlantic to changes in Global observations of oceanic Rossby waves. thermohaline forcing in northern latitudes: a Science, 272, 234-238. model study. J. Phys. Ocean., 24, 11, 2306-2320.

11 Emery, W.J., and R.E. Thompson, 1998: model of wind-driven, mid-latitude ocean Data analysis methods in . circulation. J. Phys. Ocean., 20, 9, 1349-1378. Pergamon, New York, 634 pp. Mundt, M.D., G.K. Vallis, and J. Wang, Enomoto, T., and Y. Matsuda, 1999: 1997: Balanced Models and dynamics for the Rossby wave packet propagation in a zonally- large- and mesoscale circulation. J. Phys. varying basic flow. Tellus, 51A, 588-602. Oceanogr., 27, 6,1133-1152. Gill, A.E., 1982: Atmosphere-ocean Osychny V., and P. Cornillon, 2004: Dynamics, Academic Press, San Diego. Properties of Rossby Waves in the North Frankignoul, C., G. de Coetlogon, T.M. Atlantic Estimated from Satellite Data. J. Phys. Joyce, and S. Dong, 2001: Gulf Stream Oceanogr., 34,1, 61-76. variability and ocean-atmosphere interactions. J. Polito P.S., and P. Cornillon, 1997: Long Geophys. Res., 31, 3516-2359. baroclinic Rossby waves detected by Hakkinen S., and K.C. Mo, 2002: The TOPEX/POSEIDON. J. Geophys. Res., 102, C2, low-frequency variability of the Tropical 3215-3235. Atlantic Ocean. J. Climate, 15, 3, 84-93. Rhines P., 2004: Oceanic and atmospheric Hill, K.L., I.S. Robinson, and P. Cipollini, Rossby waves. A lecture for the Norman Phillips 2000: Propagation characteristics of extra- symposium at the annual Amer. Meteorological tropical planetary waves observed in the ATSR Society meeting (available from the website: global record. J. http://www.ocean.washington.edu/research/gfd/p Geophys. Res.,105, C9, 21,927-21.945. apers-rhines.html). Huang, R.X., M.A. Cane, N. Naik, and P. Roebber, P.J. 1995: Climate variability in Goodman, 2000: Global adjustment of the a low-order coupled atmosphere-ocean model. thermocline in response to deep-water formation. Tellus, 47A, 473-494. Geophys. Res. Lett., 27, 6, 759-762. Rossby, T., and R.L. Benway, 2000: Slow Hurrell, J.W., and K.E. Trenberth, 1999: variations in mean path of the Gulf Stream east Global sea surface temperature analyses: of Cape Hatteras. Geophys. Res. Lett., 27, 117- multiple problems and their implications for 120. climate analysis, modeling and reanalysis, Bull. Susanto, R. D., Q. Zheng, and X.-H. Amer. Meteor. Soc., 80, 2661-2678. Yan, 1998: Complex singular decomposition Ivanov, L.M., A.D. Kirwan, Jr., and T.M. analysis of equatorial waves in the Pacific Margolina, 2001: Filtering noise from observed by TOPEX/POSEIDON altimeter. J. oceanographic data with some applications for Atm.Ocean. Tech., 15, 3, 764-774. the Kara and Black Seas, J. Mar. Sys., 28, 1–2, Taylor, A.H., and J.A. Stephens, 1998: 113–139. The North Atlantic Oscillation and the latitude of Killworth P.D., P. Cipollini, B. Mete Uz, the Gulf Stream. Tellus, 50A, 134-142. and J. R. Blundell, 2004: Physical and biological Thiebaux H.J., and M.A. Pedder, 1987: mechanisms for planetary waves observed in Spatial Objective Analysis: with Applications in satellite-derived chlorophyll. J. Geophys. Res., Atmospheric Science. Academic Press, 299 pp., C07002, 1-18. London. Large, W.C., and A.J. Nurser, 2001: Thierry, V., A.M. Treguer, and H. Ocean surface water mass transformation. Ocean Mercier, 2004: Numerical study of the annual Circulation and Climate, G. Siedler et al., Eds., and semi-annual fluctuations in the deep International Geophysics series, v. 77, Academic equatorial Atlantic Ocean. Ocean Modelling, 6, Press, 317-336. 1-30. Liu Z., 1999: Forced planetary wave Vapnik, V.N., 1982: Estimation of response in a thermohaline gyre. J. Phys. Ocean., Dependencies Based on Empirical Data. 29, 1036-1055. Springer-Verlag, 379 pp., New York. Lorenz, E.N. 1984: Irregularity. A Volkov D.L., 2005: Interannual variability fundamental property of the atmosphere. Tellus, of the altimetry-derived eddy field and surface 42,98-110. circulation in the extra-tropical North Atlantic Lumpkin, R., and K. Speer, 2003: Large- Ocean in 1993-2001. J. Phys. Oceanogr., 35, 4, scale vertical and horizontal circulation in the 405-426. North atlantic. J. Phys. Ocean., 33, 9, 1902- Walin, G., 1982: On the relation 1920. between sea-surface heat flow and thermal McWilliams, J.C., N.J. Norton, P. Gent circulation in the ocean. Tellus, 34, 187-195. and D. B. Haidvogel, 1990: A linear balance

12 Yang, J., 1999: A linkage between decadal climate variations in the Labrador sea and the Tropical Atlantic Ocean. Geophys. Res. Lett., v. 26, 8, p. 1023-1026. Yang H., 2000: Evolution of long planetary wave packets in a continuously stratified ocean. J. Phys. Oceanogr., 30, 2111- 2123. Yang, J., and T.M. Joyce, 2003: How do high-latitude North Atlantic climate signals the crossover between the deep Western Boundary Current and the Gulf Stream? Geophys. Res. Lett., 30, 2, 42-1 - 42-4.

13