eddy separation is examined using a recently discovered relationship between .... inducing separation. Other studies have also noted that peripheral cyclonic.
WHAT DO WE KNOW AND WHAT CAN WE PREDICT ABOUT THE TIMING OF LOOP CURRENT EDDY SEPARATION? Robert R. Leben and Daniel J. Honaker Colorado Center for Astrodynamics Research and Department of Aerospace Engineering Sciences, University of Colorado, 431 UCB, Boulder CO 80309-0431 USA ABSTRACT/RESUME The feasibility of statistically forecasting Loop Current eddy separation is examined using a recently discovered relationship between Loop Current retreat after eddy separation and the subsequent eddy separation period [1]. The strawman hypothesis is that the timing of tLoop Current penetration and eddy separation is affected more by the initial condition of the Loop Current at the onset of intrusion rather than by the various dynamical mechanisms that interact during intrusion to cause eddy detachment. Thus, the latitude of Loop Current retreat after eddy separation can be used to forecast the period of the next Loop Current eddy separation cycle. This has been attempted for the three most recent eddy separation events. The strengths and weaknesses of the statistical forecasts are examined, as well as potential applications to operational planning and hurricane forecasting. 1. INTRODUCTION The ability to accurately predict the time that a Loop Current eddy will ultimately separate from the Loop Current has become a grail of sorts in Gulf of Mexico oceanography, and for good reason. Approximately 30% of the crude oil and 20% of the natural gas produced in the U.S. comes from the Gulf of Mexico, much of it from the continental slope and deepwater of the north-central Gulf. Strong Loop Current and Loop Current eddy currents affect day-to-day operations during offshore oil and gas exploration and production activities, frequently shutting down operations and making planning and scheduling difficult for this very expensive enterprise. The Loop Current and Loop Current eddies also play an active role in the rapid intensification of Gulf of Mexico hurricanes. This was the case in Hurricanes Katrina and Rita, both of which intensified over the intruded Loop Current and reached category 5 before making landfall along the northern Gulf coast [2]. Clearly if Loop Current intrusion and eddy separation could be predicted a year or even just months in advance there could be significant socioeconomic benefits. Efforts along these lines using operational data-assimilative nowcast/forecast ocean models have not been very successful. This is because the Loop
Current is an inherently unstable current system in which the dominant processes affecting intrusion and separation are not clearly understood, making accurate forecasting using numerical models very difficult. A review of the 13-year altimetric record of Loop Current metrics and comparisons with the Loop Current penetration and eddy separation cycles, however, suggests that there may be a more fundamental controlling influence on LC eddy separation. Thus, it may be feasible to statistically forecast Loop Current eddy separation. 2. LOOP CURRENT EDDY SEPARATION The Loop Current is the energetic section of the Gulf Stream System that “loops” in a clockwise direction within the Gulf of Mexico after entering as a northward flowing current through the Yucatan Channel and later exiting to the east through the Florida Straits. The northward penetration of the Loop Current into the eastern Gulf deep basin varies with time and at irregular intervals becomes great enough to produce large anticyclonic rings of recirculating current known as Loop Current eddies. Separation is defined as the final detachment of an eddy from the Loop Current with no later reattachment. Although eddies frequently detach from and reattach to the Loop Current during intrusion, the ultimate detachment or separation occurs most frequently at intervals of about 6 and 11 months [1,3]. The range of eddy separation periods that have been observed is from 2 weeks to 18 months [1]. Unfortunately, the dominant dynamical mechanisms controlling Loop Current eddy separation are poorly understood making accurate prediction of the highly variable Loop Current penetration and eddy separation cycle very difficult. A popular hypothesis is that the primary source of Loop Current eddy separation variability comes from upstream conditions influencing the circulation within the Gulf through the connectivity of the western boundary current system in the subtropical North Atlantic gyre. This conjecture is supported by modeling studies that show propagating anticyclonic eddies in the upstream current eventually passing through the Yucatan channel and affecting the Loop Current
to isolate a dominant mechanism in any observed eddy shedding cycle. In [1], a review of the altimetric record of Loop Current metrics and comparisons with the Loop Current penetration and eddy separation cycles suggests that there may be a more fundamental controlling influence on separation that is related to the retreat of the Loop Current after an eddy separates. This relationship was first found with an objective technique to track the Loop Current using altimeter data. 3. ALTIMETRIC TRACKING OF THE LOOP CURRENT
Figure 1. Contoured sea surface height (5 cm interval; white contours are negative, black positive) from 13 Mar 2002 overlaid on the nighttime composite sea surface temperature image from GOES 8 for the same day (courtesy of Nan Walker, LSU). The bold line is the 17-cm contour used to track the Loop Current boundary and eddies. behavior within the Gulf [4,5]. Further support of this hypothesis comes from the vorticity flux into the Gulf calculated from in situ current measurements in the Yucatan Channel that show correlation with the changes in the Loop Current penetration and eddy separation as observed by satellite altimetry [6]. The source of the vorticity flux is thought to be upstream eddies [7]. Transport variations in the Loop Current [8], including variations in the deep outflow [9], have also been proposed as potential dynamical influences on Loop Current intrusion and eddy separation. Another proposed mechanism for the exhibited variability is the influence of peripheral and low-layer eddies. In model simulations, a lower-layer eddy dipole pair forms beneath Loop Current eddies during eddy separation [10] and contributes to the ultimate detachment of the eddy from the Loop Current. Simulations in [11] showed that the growth of cyclones in the deep lower layer below the Loop Current recirculation are caused by an instability of the forming ring. Early in the separation process, one of the instability-generated deep cyclones pairs with a deep anticyclone to form the dipole pair inducing separation. Other studies have also noted that peripheral cyclonic eddies, including Loop Current frontal eddies [12], Tortugas eddies [13; 14] and Campeche Bank eddies [15], are often associated with eddy separation events and are thought to play a role in the ultimate formation and detachment of a Loop Current eddy. The large variety of mechanisms discussed above that are implicated in eddy separation makes it very difficult
The primary goal of altimetric Loop Current tracking is to objectively monitor the time-dependent behavior of the Loop Current and accurately estimate Loop Current metrics from the continuous time series of altimeter observations. An accurate description of the Loop Current from observations is important for the study of dynamical processes influencing Loop Current penetration and eddy separation. The Loop Current can be identified in sea surface height by using a contour that closely tracks the edge of the high-velocity core of the Loop Current (see Fig. 1). The 17-cm contour was found to work well for the Colorado Center for Astrodynamics Research (CCAR) Gulf of Mexico sea surface height maps, however, this contour will typically be different for other altimeter products or model output. CCAR maps are based on sea surface height anomaly fields added to a model mean based on a data assimilation hindcast performed by Drs. Lakshmi Kantha and Jei Choi [16]. The 17-cm tracking contour is autonomously tracked in the presence of other contours (such as those associated with detached Loop Current eddies) using a MATLAB program, which identifies the Loop Current contour and calculates a variety of Loop Current metrics such as the length, area, volume and circulation associated with the current, and its maximum northward and westwards extent. We identify the timing of Loop Current eddy separation events using the Loop Current length time series since the breaking of the 17-cm contour between the Loop and a detaching eddy into separate contours causes a discrete change in the Loop Current length equal to the circumference of the separating eddy (see for example Quick Eddy, which is just about to separate in Fig. 1). This objective method for tracking the Loop Current and detecting Loop Current eddy separation events gives separation periods comparable [1] to the periods determined by subjective tracking methods [3]. We quote an exact day of separation determined by the breaking of the tracking contour; however, estimated uncertainties in separation period may be as great as one month [3]. The day that the breaking of the contour occurs is identified as the “time” of eddy separation.
Figure 2. The 20 Loop Current eddy separation events identified in the altimeter record are shown above. Sea surface height maps on the separation dates are shown in the panels to the right (note that the values above 40 cm and below 30 cm have been clipped). Eddy separation dates were objectively determined by breaking of the 17-cm tracking contour, which causes a discrete change in the Loop Current length (left panel). The length time series is overlaid with red lines corresponding to the 20 events identified.
Occasionally a detached eddy will reattach to the Loop Current, so in those cases the time associated with the final detachment of the eddy is referred to as the eddy separation time. The exact timing of a separation event, therefore, is dependent on the criteria selected to define separation and is complicated by the ambiguity of associating an exact time with what is clearly a continuous and complicated process. The tracking contour also impacts Loop Current and eddy statistics, such as length or areal extent that are estimated using the tracking contour. Nevertheless, an objective definition of separation provides a useful benchmark for comparing Loop Current eddy separation events. Table 1. Loop Current eddy separation events from the 1 Jan 1993 through 15 Mar 2006 altimeter record. The separation dates for Eddies Walker and Xtreme are estimates based on near real-time altimetry. Separation Eddy Industry Date Period Number Eddy Name (months) 1 11 Jul 1993 11.5 Whopper 2 10 Sep 1993 2.0 Xtra 3 27 Aug 1994 11.5 Yucatan 4 18 Apr 1995 7.5 Zapp 5 8 Sep 1995 4.5 Aggie 6 14 Mar 1996 6 Biloxi 7 13 Oct 1996 7 Creole 8 30 Sep 1997 11.5 El Dorado 9 22 Mar 1998 5.5 Fourchon 10 2 Oct 1999 18.5 Juggernaut 11 10 Apr 2001 18.5 Millennium 12 22 Sep 2001 5.5 Odessa/Nansen 13 28 Feb 2002 5.5 Pelagic 14 13 Mar 2002 0.5 Quick 15 5 Aug 2003 17 Sargassum 16 31 Dec 2003 5 Titanic 17 28 Sep 2004 9 Ulysses 18 14 Sep 2005 11.5 Vortex 19 *19 Feb 2006 5 Walker 20 *2 Mar 2006 0.5 Xtreme The Loop Current length time series and maps at the time of separation for each Loop Current separation event in the altimeter record are shown in Fig. 2. Horizon Marine, Inc. names the Loop Current eddies in alphabetical order as anticyclones shed from the Loop Current and/or impact offshore operations in the northern Gulf of Mexico. A complete list is published on the web at http://horizonmarine.com/namedlces.html. The names appear in the EddyWatch™ reports provided to the Gulf of Mexico offshore oil and gas industry by subscription. The eddy number, separation date, separation period, and industry eddy name are tabulated for each of the observed events in Table 1.
A total of 20 Loop Current eddy separation events have been identified in altimeter record from 1993 through the drafting of this paper (15 March 2006). The most recent eddy separation date is tentative for Eddy Xtreme, since there is a remote possibility that the eddy may reattach to the Loop Current. All separation events identified to date in the altimeter record using the SSH 17-cm tracking contour have been monitored by the EddyWatch™ service, although a number of smaller anticyclonic eddies (7 total) were also named that causes the breaks in the alphabetical sequence. Only one marginal eddy separation event was identified by the objective tracking procedure (Eddy Odessa/Nansen, Eddy 12), which dissipated so quickly that an estimate of the eddy area could not be made using the tracking contour. These smaller eddies are of Loop Current origin, but form on the outer edge of the LC through the interaction of frontal cyclones with the current. Ideally they should not be counted since they are better categorized as minor, peripheral eddies. 3. RELATIONSHIP BETWEEN LOOP CURRENT RETREAT AND EDDY SEPARATION PERIOD The analysis of the Loop Current metrics derived from the automated tracking shows a strong relationship between Loop Current retreat and subsequent eddy separation periods with more southerly retreats exhibiting longer separation periods. As discussed in [1] the 1993 through 2003 record exhibited an average separation period following retreat of 16.2 months when the entire Loop Current retreated below 25ºN, which much longer than the 5.5 month average period for instances where part of the Loop Current remained north of 25ºN. The updated results are shown in the left panel of Fig. 3. Eddy separation periods for the Loop Current intrusion/eddy separation events #2 through #20 are plotted versus the values of the Loop Current northern extent (i.e. the Loop Current maximum latitude) immediately following the previous eddy separation event. These northern Loop Current extent values were calculated by finding the minimum value of the northernmost Loop Current 17-cm contour latitude coordinate in the five-day window following separation of a Loop Current eddy. One retreat value (#3 at 26.34°N shown in bold in the figure) did not reflect the initial state of the Loop Current because a warm filament affected the 17-cm tracking contour just after eddy separation. Instead, a value of 25.6°N is chosen for event #3 that corresponds to the latitude where the Loop Current remained for nearly three months just after separation. Otherwise, the Loop Current retreat values are relatively stable and can be calculated in a variety of ways by using mean or extrema values within a variety of windows.
Figure 3. The left panel shows separation period versus the retreat latitude of the Loop Current immediately following the previous eddy separation. Points are numbered according to the events shown in Fig. 2. The right panel shows the same points with selected sequential events combined as discussed in the text. The regression line (red dash) is used to predict the eddy separation period from the observed retreat latitude of the Loop Current. Investigation of the incongruent period associated with separation event #8 leads to an interesting relationship. While events #10, 11 and 15 are clustered along the line extrapolating the nearly linear relationship of retreat to period above 26ºN, the period associated with event #8 is much shorter. Note, however, that if the periods of events #8 and 9 are combined and plotted with the initial retreat of event #8, then the point falls within the cluster of the other three southernmost retreat points. In fact, if all of the events exhibiting the incongruent periods (#6, 8 and 12) are combined with the period of the next eddy separation event (#7, 9, and 13, respectively), then a nearly perfect linear relationship is found between retreat latitude and separation period (Fig. 3, right panel). 4. PREDICTING EDDY SEPARATION PERIODS The statistical relationship described in the previous section shows a clear relationship between the Loop Current retreat after eddy separation and the time that a large Loop Current eddy will ultimately separate during the next Loop Current intrusion cycle. Thus, we can predict the next Loop Current eddy separation period by simply using a line fit to the data in the right panel of Fig. 3. Since this relationship was found three additional eddy shedding events have occurred: #18 Eddy Vortex, #19 Eddy Walker, and #20 Eddy Xtreme. These “hindcast” separation predictions will be discussed now. Eddy Vortex (#18): The Loop Current’s retreat latitude after separation of Eddy Ulysses (#17) was 25.36°N, which predicted that the separation period of Eddy Vortex would be 394 days. The predicted value was within 13% of the 351-day observed period. A total of
four detachment/reattachment events occurred during the Eddy Vortex intrusion cycle. The first detachment was on 21 Feb 2005 about 7 months before the ultimate separation of Eddy Vortex. The eddy remained detached from the Loop Current for 52 days until reattaching on 14 April 2005. The other three detachment events occurred on 14 May 2005, 18 June 2005 and 7 August 2005 with the eddy reattaching 7, 8 and 12 days later, respectively. The statistical relationship between retreat and period gives information only on the time of final separation of an eddy not on any intermediate behavior such as deep northerly intrusions or eddy detachments that the Loop might exhibit. Nevertheless, when an eddy detachment occurs within the predicted period then the eddy is more likely to reattach to the Loop Current. Operationally this information is very valuable since separation usually results in an eddy leaving an area where strong eddy currents are impacting off shore activities. In this case the Loop Current remained intruded for over 7 months, significantly impacting offshore activities. The far northerly intrusion of the Loop Current and Eddy Vortex at the time of separation was an important source of thermal energy that contributed to the severe intensity of Hurricane Katrina [2]. Because of the primary role that the Loop Current plays in the formation of strong Gulf hurricanes it may be possible in the future to incorporate statistical predictions of the Loop Current into the early season forecasts of the intensity of Gulf coast land-falling hurricanes. Eddy Walker (#19) and Eddy Xtreme (#20): The retreat latitude of the Loop Current after the separation of Eddy Vortex (#18) was 26.8°N, predicting that the
eddy separation period for Eddy Walker (#19) would be 96 days. The observed value was 154 days. This prediction does not agree as well with the observed period as the prediction for Eddy Vortex. The predicted period was only within 40% of the actual. The latitude of retreat of the Loop Current after the separation of Eddy Walker was the most northerly retreat observed to date. This predicted that the next Loop Current eddy, Eddy Xtreme, would separate soon, which it did on 2 March 2006. The timing of these separations and the far western intrusion of the Loop Current is reminiscent of the events observed in 2002 when the westernmost eddy observed in the altimetric record (#13 Pelagic Eddy) separated from the Loop Current on February 28th, followed closely by a second eddy (#14 Quick Eddy) on March 15th (see Fig. 1). Eddy “Y” (#21): Sea surface temperature imagery and near real-time altimetry tentatively show that the Loop Current retreated to about 25.8°N after the separation of Eddy Xtreme. This gives a forecast for the Eddy “Y” separation period of about 300 days, which places separation near 1 Jan 2007. Although separation is predicted to be after the 2006 hurricane season, the exact position of the Loop Current cannot be predicted during the intrusion cycle. It is likely, however, that the Loop Current will be intruded more than the long-term average, which is 26.2°N, increasing the chances that hurricanes passing over the Gulf will encounter warm subsurface waters associated with the Loop Current and intensify. The Loop Current position will be monitored in the coming months to see if conditions similar to 2005 are likely to occur again in 2006. We continue to look for other Loop Current metrics that are well predicted by earlier observations in order to predict the size and location of the Loop Current at separation and at the times in between. 5 REFERENCES 1.
2.
3.
4.
R. R. Leben, Altimeter-derived Loop Current metrics, in Circulation in the Gulf of Mexico: Observations and Models, W. Sturges and A. LugoFernandez, Editors, AGU Geophysical Monograph 161, p. 181-202, 2005. R. Scharroo, W. Smith, and J. L. Lillibridge, Satellite altimetry and the intensification of hurricane Katrina, Eos, Vol. 86, No. 40, 4 October 2005. W. Sturges and R. Leben, Frequency of ring separations from the Loop Current in the Gulf of Mexico: A revised estimate, J. Phys. Oceanogr., 30, 1814–1819, 2000. S. J. Murphy, H. E. Hurlburt, J. J. O’Brien, The connectivity of eddy variability in the Caribbean
5.
6.
7.
8.
9.
10. 11.
12. 13.
14.
15.
16.
Sea, the Gulf of Mexico and the Atlantic Ocean, J. Geophys. Res., 104, 1431−1453, 1999. L.-Y. Oey, H.-C. Lee and W. J. Schmitz, Effects of winds and Caribbean eddies on the frequency of Loop Current eddy shedding: A numerical modeling study, J. Geophys. Res., 108, 3324, doi:10.1029/20002JC001698, 2003. J. Candela, J. Sheinbaum, J. Ochoa, A. Badan and R. Leben, The potential vorticity flux through the Yucatan Channel and the Loop Current in the Gulf of Mexico, Geophys. Res. Lett., 29, No. 22, 2059, doi:10.1029/2002GL015587, 2002. A. J., Abascal, Sheinbaum, J., J. Candela, J. Ochoa, and A. Badan, Analysis of flow variability in the Yucatan Channel, J. Geophys. Res., 108, No. 12, 3381, doi:10.1029/2003JC001922, 2003. G. A. Maul and F. M. Vukovich, The relationship between variations in the Gulf of Mexico Loop Current and Straits of Florida volume transport, J. Phys. Oceanogr., 23, 785−796, 1993. L. Bunge, J. Ochoa, A. Badan, J. Candela, and J. Sheinbaum, Deep flows in the Yucatan Channel and their relation to changes in the Loop Current extension, J. Geophys. Res., 107, 3223, doi: 10.1029/2001JC001256, 2002. S. U. Welsh and M. Inoue, Loop Current rings and the deep circulation in the Gulf of Mexico. J. Geophys. Res., 105, 16,951−16,959, 2000. L. M. Chérubim, W. Sturges and E. P. Chassignet, Deep flow variability in the vicinity of the Yucatan Straits from a high-resolution numerical simulation, J. Geophys. Res., 110, C04009, doi:10.1029 /2004JC002280, 2004. F. M. Vukovich and G. A. Maul, Cyclonic eddies in the eastern Gulf of Mexico, J. Phys. Oceanogr., 15, 105−117, 1985. T. N. Lee, K. Leaman, E. Williams, T. Berger and L. Atkinson, Florida Current meanders and gyre formation in the southern Straits of Florida, J. Geophys. Res., 100, 8607−8620, 1995. P. S. Fratantoni, T. N. Lee, G. Podesta, and F. Muller-Karger, The influence of Loop Current perturbations on the formation and evolution of Tortugas eddies in the southern Stratis of Florida, J. Geophys. Res., 103, 24,759−24,779, 1998. J. Zavala-Hidalgo, S. L. Morey and J. J. O’Brien, Cyclonic eddies northeast of the Campeche Bank from altimetry data, J. Phys. Oceanog., 33, 623629, 2003. W. D. Nowlin Jr., A. E. Jochens, S. F. DiMarco, R. O. Reid and M. K. Howard, Deepwater physical oceanography reanalysis and synthesis of historical data: Synthesis report, OCS Study MMS 2001-064, Synthesis Report, U. S. Dept. of the Interior, Minerals Management Service, Gulf of Mexico OCS Region, New Orleans, LA, 530 pp, 2001.
