The Equatorial Undercurrent and TAO Sampling Bias from a Decade at SEA

SEPTEMBER 2014 L E S L I E A N D K A R N A U S K A S 2015

WILLIAM R. LESLIE Department of Geology, Oberlin College, Oberlin, Ohio, and Woods Hole Oceanographic Institution, Woods Hole, Massachusetts

KRISTOPHER B. KARNAUSKAS Woods Hole Oceanographic Institution, Woods Hole, Massachusetts

(Manuscript received 16 December 2013, in ﬁnal form 30 April 2014)

ABSTRACT

The NOAA Tropical Atmosphere Ocean (TAO) moored array has, for three decades, been a valuable resource for monitoring and forecasting El Niño–Southern Oscillation and understanding physical oceanographic as well as coupled processes in the tropical Pacific influencing global climate. Acoustic Doppler current profiler (ADCP) measurements by TAO moorings provide benchmarks for evaluating numerical simulations of subsurface circulation including the Equatorial Undercurrent (EUC). Meanwhile, the Sea Education Association (SEA) has been collecting data during repeat cruises to the central equatorial Pacific Ocean (1608–1268W) throughout the past decade that provide useful cross validation and quantitative insight into the potential for stationary observing platforms such as TAO to incur sampling biases related to the strength of the EUC. This paper describes some essential sampling characteristics of the SEA dataset, compares SEA and TAO velocity measurements in the vicinity of the EUC, shares new insight into EUC characteristics and behavior only observable in repeat cross-equatorial sections, and estimates the sampling bias incurred by equatorial TAO moorings in their estimates of the velocity and transport of the EUC. The SEA high-resolution ADCP dataset compares well with concurrent TAO measurements (RMSE 5 2 0.05 m s 1; R2 5 0.98), suggests that the EUC core meanders sinusoidally about the equator between 60.48 latitude, and reveals a mean sampling bias of equatorial measurements (e.g., TAO) of the EUC’s zonal 2 velocity of 20.14 6 0.03 m s 1 as well as a ;10% underestimation of EUC volume transport. A bias-corrected monthly record and climatology of EUC strength at 1408W for 1990–2010 is presented.

1. Introduction of SEA cruises and the TAO array, a natural focal point is the Equatorial Undercurrent (EUC), a prominent year- Under way measurements of conductivity–temperature– round feature that delivers cold, nutrient-rich water to depth (CTD), as well as horizontal subsurface velocity the eastern Pacific, especially in zones of enhanced up- via acoustic Doppler current profiler (ADCP) have been welling along the equator (Johnson et al. 2001) and the made routinely on educational cruises of the Sea Edu- western coasts of the Galápagos and South America cation Association (SEA) and archived locally over the (Lukas 1986). The EUC flows eastward along the equa- past decade. The spatial and temporal characteristics of 2 torial thermocline at a rate of approximately 1 m s 1 with the SEA cruises provide a unique lens in which to cross seasonal transport exceeding 40 Sverdrups (Sv; 1 Sv [ validate and better understand several important features 2 106 m3 s 1; Johnson et al. 2002). The EUC is both subject concurrently observed by the National Oceanic and At- to and a major player in ENSO dynamics as evident by the mospheric Administration (NOAA) Tropical Atmosphere sweeping changes in the tropical ocean circulation and Ocean (TAO) array in the equatorial Pacific Ocean thermohaline structure amidst the subsequent breakdown (McPhaden et al. 1998). Given the sampling distribution in the Walker circulation during El Niño (Firing et al. 1983). The EUC is also vitally important to smaller island ecosystems throughout the equatorial Pacific Corresponding author address: Kristopher B. Karnauskas, Woods Hole Oceanographic Institution, 360 Woods Hole Rd., MS such as Jarvis Island (Gove et al. 2006) and the Gilbert 23, Woods Hole, MA 02544. Islands (Karnauskas and Cohen 2012), where the E-mail: [email protected] strength of topographic upwelling (and thus delivery of

DOI: 10.1175/JTECH-D-13-00262.1

Ó 2014 American Meteorological Society Unauthenticated | Downloaded 10/01/21 09:37 PM UTC 2016 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 31

FIG. 1. Overview map of the tropical Pacific Ocean indicating the 21 SEA cruises utilized in this study (black lines), subsets of SEA cruises grouped by the longitude of equatorial crossing (colored ellipses, corresponding to Fig. 2b), and the 1408W TAO mooring also used. cold, nutrient-rich water) is dependent on EUC ve- 2. Description of the SEA dataset and comparison locity. Nonetheless, global climate models do not ad- with TAO equately simulate the EUC (Karnauskas et al. 2012) a. Temporal and spatial sampling and may, therefore, suffer in predictions of how the mean circulation and ENSO system may respond to The SEA dataset used herein draws from 21 ‘‘Semester natural and anthropogenic climate forcing, including at Sea’’ cruises of the Sailing School Vessel (SSV) Robert C. impacts on ecosystems. Seamans (http://www.sea.edu/ships_crew/seamans)from This paper characterizes the high-resolution SEA 2003 to 2012 with a total of 23 equatorial crossings (the dataset and applies it toward understanding the nature equator was crossed twice during two cruises; Fig. 1). The of the EUC as well as illuminating potential sam- EUC is known to have significant longitudinal and pling biases in sustained, equatorial observations of the seasonal variations (Johnson et al. 2002), thus sub- EUC by moored buoys. In the following section, sam- sampling the SEA dataset is necessary to control for pling aspects of the SEA data are described and ADCP those variables. The cruises are clearly separable into measurements are compared against TAO observations. two distinct seasons: boreal spring (February–May) In section 3, SEA data are analyzed and compared with and December (Fig. 2a). In figures throughout this previous observations to illuminate some characteristics paper, we denote spring results as red and December of the EUC including its spatial and temporal variability. results as blue unless otherwise noted. Furthermore, Focusing on SEA cross-equatorial sections, profiles at ADCP data were collected near the equator along three the equator are compared against those at the core of roughly similar cruise tracks (Figs. 1 and 2b). The mean the EUC in section 4 to address issues concerning the longitudes of equatorial crossings for the three similar use of TAO measurements to estimate EUC velocity tracks are western (1588W), central (1448W), and east- and transport. Finally, a summary and discussion of the ern (1298W). Figures throughout this paper use red, scientific implications of this work and future directions green, and blue to denote the western, central, and are provided in section 5. eastern tracks, respectively, unless otherwise noted. The

FIG. 2. Histograms of (a) calendar day and (b) longitude of equatorial crossing for each of the 23 SEA equatorial crossings. (c) Scatterplot of calendar day vs longitude of equatorial crossings with ellipses denoting subgroups with similar longitude and season (west/December, central/spring, and east/December).

Unauthenticated | Downloaded 10/01/21 09:37 PM UTC SEPTEMBER 2014 L E S L I E A N D K A R N A U S K A S 2017 entire study area falls within the central Pacific where the EUC is relatively swift; the climatological maximum EUC velocity and transport occurs at 1258W(Johnson et al. 2002). Overall, the 23 SEA equatorial crossings provided a convenient distribution: 13 spring crossings and 10 winter crossings, as well as 10 crossings in the western track, 8 in the central track, and 5 on the eastern track. Further subgroups of five cruises a piece are created basedonbothcomparableseasonandlongitude:west/ December (red), central/spring (green), and east/De- cember (blue), enabling us to further isolate the potential confounding variables of season and longitude and facil- itate consistency over time as well as providing a fair cross comparison with TAO measurements from a single longitude. More western tracks have been occupied in recent years (Fig. 3). All SEA ADCP data analyzed in this paper FIG. 3. Scatterplot of decimal year vs longitude of equatorial are available online at http://hdl.handle.net/1912/6746. crossings along with simultaneous daily ADCP measurements from the 08,1408W TAO mooring (gray vertical bars). b. Comparison with ADCP measurements from the 1408W TAO mooring TAO measurement (Fig. 4). The vertical profiles of zonal In this section, comparisons of SEA and TAO mea- velocity from SEA and TAO shown in Fig. 4b were taken surements at the equator are used to evaluate the suit- on 29 May 2009 within 1 h and within 20 km of one an- ability of SEA ADCP measurements made exactly on other. The comparison reveals excellent agreement be- the equator as a faithful proxy for TAO moored ADCP tween the two datasets with a small offset below the measurements. The TAO mooring used in this paper is EUC. The difference in maximum equatorial zonal ve- 2 positioned on the equator at 1408W(Figs. 1 and 3). locity between SEA and TAO (Fig. 4c)is20.02 m s 1 2 2 ADCP measurements from this TAO mooring were with an RMSE of 0.05 m s 1 to 250-m depth or 0.01 m s 1 missing during the periods indicated by breaks in the to the EUC core (89-m depth). gray bars (i.e., from 2007 to 2009 and again after 2010). Second, a comparison of maximum equatorial veloc- First, a single, nearly simultaneous and collocated ity for the east/December subset is compared against ADCP velocity profile is compared against an equivalent TAO measurements of zonal velocity, which allows

21 FIG. 4. (a) Vertical–meridional section of zonal velocity (m s ) from SEA cruise on 29 May 2009 clearly indicating the EUC. The black 2 contour line is set at 1.5 m s 1. (b) Vertical proﬁles of zonal velocity from the same cruise (thick gray) and a nearly collocated and simultaneous (within 1 h and 20 km) measurement from the 1408W TAO mooring (black). (c) The difference in equatorial zonal velocity 2 2 (SEA 2 TAO). The RMSE is 0.05 m s 1 to 250-m depth or 0.01 m s 1 to the core of the EUC (89-m depth), and the difference in maximum 2 zonal velocity is 20.02 m s 1.

Unauthenticated | Downloaded 10/01/21 09:37 PM UTC 2018 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 31

2 FIG. 5. Scatterplots of (a) SEA vs TAO equatorial maximum zonal velocities (R 5 0.98) and (b) SEA at the EUC core vs TAO equatorial maximum zonal velocities (R2 5 0.64) in east/ December subgroup. Note that the nonunity slope and nonzero intercept is due to the slight offset in longitude between the TAO mooring and the eastern group of SEA equatorial crossings. for the largest possible comparison of SEA and TAO 3. Spatiotemporal variability of the EUC in the (N 5 5 cruises) while controlling for longitude and season SEA dataset (Fig. 5). In this comparison, values of maximum zonal The strength of the EUC is calculated in terms of velocity from both the equator and from the latitude of maximum zonal velocity at the latitude of the EUC core the EUC core in SEA are compared to the equatorially and 2D transport at the same latitude (the vertical in- confined TAO measurement. ADCP velocity profiles tegral of a zonal velocity profile as in the one shown from 60.058 latitude (;5 km) are averaged to form each in Fig. 4b).1 Figure 7a relates maximum zonal velocity equatorial measurement. A very strong linear relation- and 2D transport, which shows only a modest correla- ship is observed between SEA and TAO equatorial tion (R2 5 0.41). However, relating 2D transport with measurements, which indicates that SEA sampling at the volume (3D) transport suggests a stronger correlation equator offers a faithful proxy for an equatorial TAO 2 2 (R 5 0.71). This is noteworthy because it suggests that mooring (RMSE 5 0.05 m s 1; R2 5 0.98). The high different indicators of EUC strength are not necessarily correlation deteriorates in the comparison between TAO interchangeable, and especially that maximum zonal (equatorial) maximum velocity and SEA core maximum velocities is a poor predictor of 2D (and therefore vol- velocity (R2 5 0.64), implying that even with the effects of ume) transport. The weak correlation between maxi- longitudinal and seasonal variations removed, TAO mum zonal velocity and transport is not surprising given equatorial measurements do not closely capture the the varying spatial extent of the EUC as shown by variability of the maximum velocity at the EUC core. Johnson et al. (2002). Also, given the strong correlation We note that while one correlation may be significant in Fig. 7b, if 2D transport is used as a proxy for volume and the other not, the two correlations may not be transport, then purely equatorial measurements might significantly different from one another with such small underrepresent both if the equatorial measurement sample sizes. We can, however, conclude that SEA misses the position of the core of the EUC. equatorial measurements are a good proxy for TAO The EUC is known to shoal from ;200-m depth in the equatorial measurements. west to ,100 m in the east across the Pacific basin, Finally, a time series with the 15 cruises grouped by closely following the thermocline (Johnson et al. 2002). season and longitude is shown to further broaden the Furthermore, the EUC appears weaker and spatially scope of comparison to include interannual variability (Fig. 6). With an expected offset due to unavoidable differences in longitudinal sampling, TAO measurements 1 2D transport is computed as the vertical integral of zonal ve- tend to closely follow the SEA equatorial measurements. locity (eastward only) from the shallowest depth at which the zonal This confirms that SEA equatorial measurements are in velocity becomes eastward down to the last depth at which the good agreement with TAO moored measurements on an zonal velocity is eastward or 350 m, whichever is shallower [to interannual time scale. Further, the consistent offset be- avoid including, e.g., the Equatorial Intermediate Current (EIC) in the calculation]. Volume (3D) transport is calculated by integrating tween the SEA equatorial (dashed) and SEA core (solid) 2 all zonal velocities .5cms 1 between 28N and 28S, and shallower lines show that rarely do equatorial measurements cap- than 300 m. These schemes were deemed appropriate based on ture the EUC’s maximum zonal velocity. This point is careful analysis of each of the SEA transects in this region and the further investigated in section 4. sensitivity to minor changes in these parameters is small.

Unauthenticated | Downloaded 10/01/21 09:37 PM UTC SEPTEMBER 2014 L E S L I E A N D K A R N A U S K A S 2019

Figure 9a shows a positively skewed distribution of the depth indicating an average depth of 120 m. Fur- thermore, Fig. 9b showsthattheEUC’scoreisover twice as likely to be found outside of 60.1258 latitude of the equator than within those bounds. The distribution shown in Fig. 9b also implies that TAO moorings are unlikely to capture the actual core velocity, which further highlights the necessity for a quantitative basis for determining the sampling bias associated with equatorial observations of EUC strength (section 4). What explains the observed distribution of the latitude of the EUC core? We propose a simple model to explain why the EUC core is more often observed at some position y away from the equator. Consider the 21 FIG. 6. Time series comparing (a) maximum zonal velocity (m s ) simple model: 2 and (b) 2D transport (m2 s 1) for the EUC core as measured by SEA (solid lines), the equator as measured by SEA (dashed lines), and y 5 A sinx, the equator as measured by TAO (filled circles). The lines are grouped into west/December (red), central/spring (green), and east/ where the amplitude A represents the maximum lat- December (blue). itudinal meander of the EUC core. Solutions of the model for varying parameter A are shown in Fig. 10a. Figure 10b broader in the western Pacific and becomes progressively depicts the corresponding histograms of y for each solu- narrower and faster as it moves eastward with a maxi- tion using identical bin widths as in Fig. 9. The sinusoidal mum zonal velocity observed near 1258W(Johnson et al. function y 5 0:48 sinx,wherex may represent time or 2002). To confirm that the SEA dataset captures such zonal distance, appears to provide the best fit to the ob- spatial and temporal structure of the EUC, Fig. 8 depicts servations (Fig. 9b), thus suggesting that the EUC me- EUC core values of zonal velocity, 2D transport, and anders in a sinusoidal fashion with an amplitude of 0.48 depth against both longitude (Figs. 8a,c,e) and the day of latitude (or ;44 km) over this longitudinal range. the year (Figs. 8b,d,f). The seasonal cycle and longitudi- On the equator, where the Coriolis parameter is zero, nal variation are consistent with previous observational we may actually expect a broad spectrum of wave energy and modeling studies stating the EUC strengthens/shoals and other influences on the strength and/or latitudinal during springtime and weakens/deepens during Decem- position of the EUC. A period of 5–10 days can be vi- ber, and the general strengthening and shoaling is from sually discerned from relatively short TAO records in west to east (Johnson et al. 2002). the eastern equatorial Pacific based on Karnauskas et al. Histograms of EUC core depth and latitude (Fig. 9) (2010, their Fig. 1a). Spectral analysis of a daily time further highlight the spatial variability of the EUC. series of maximum eastward velocity measured by ADCP

21 2 21 FIG. 7. (a) Scatterplots of (a) maximum zonal velocity (m s ) vs 2D transport (m s ) and (b) 2 2D transport (m2 s 1) vs volume transport (Sv).

Unauthenticated | Downloaded 10/01/21 09:37 PM UTC 2020 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 31

FIG. 8. Scatterplots of (a) maximum zonal velocity, (c) 2D transport, and (e) depth of the EUC as a function of the longitude where color denotes season (red for spring and blue for December). (b),(d),(f) As in (a),(c),(e), but as a function of date, where color denotes longitude (red for west, green for central, and blue for east). at the 1408W equatorial TAO mooring (see Fig. 15a, due to the small sample size and the relatively narrow described in greater detail below) indicates variability longitudinal range sampled by SEA. A more compre- with statistically significant power (99% confidence level) hensive dataset is clearly needed to establish such second- occurring within two distinct subannual frequency bands: order dependencies. 3–17 and 50–60 days. Such observed temporal variability could have been interpreted solely as the EUC varying in 4. Estimating the TAO sampling bias using the strength over time, but appears equally likely to be the SEA dataset EUC core meandering about the equator allowing for TAO to momentarily observe values closer to the true The SEA data are strongly correlated with corre- maximum core velocity. Given the time scales noted sponding TAO measurements on the equator and are above, several phenomena are implicated such as internal qualitatively consistent with previous observational de- waves (3–15 days; Farrar and Durland 2012), tropical scriptions of the tropical Pacific Ocean circulation and instability waves (15–40 days; Lyman et al. 2007), intra- structure. Yet, they also provide the essential off-equatorial seasonal Kelvin waves (60–75 days; Kessler et al. 1995), perspective. In particular, the SEA dataset enables us to and the Madden–Julian oscillation (30–90 days; Madden quantify the actual EUC core velocity (and transport) and Julian 1971). when the EUC is not on the equator, which is the ma- Finally, we test for dependence of the EUC core lat- jority of the time as shown above (Fig. 9b). Figure 12 itude on longitude and season. Results suggest that lat- compares SEA measurements from the equator with itudinal variations of the EUC core are more prominent those from the latitude of the EUC core. There is a strong in the west and during boreal spring (Fig. 11). Error bars positive linear correlation, but more interesting is the are set to two standard errors (;95% confidence); these offset between the best-fit and one-to-one lines in both results are not statistically significant, which is likely zonal velocity (Fig. 12a) and 2D transport (Fig. 12b). This

Unauthenticated | Downloaded 10/01/21 09:37 PM UTC SEPTEMBER 2014 L E S L I E A N D K A R N A U S K A S 2021

FIG. 9. Histograms of (a) depth (m) and (b) latitude (8) of the EUC core for the 23 SEA equatorial crossings. The mean values are indicated by solid black lines. suggests that the equatorial profiles are systematically Drenkard and Karnauskas 2014), this knowledge should undersampling the maximum EUC velocity and trans- lead to improved confidence in EUC estimates and, port. For example, the mean maximum (core) EUC ve- hence, understanding of its role in basin-scale ocean locity estimated by SEA profiles at the latitude of the core circulation and climate. 2 is 1.42 m s 1, whereas the mean estimated strictly from The distribution of biases in maximum zonal velocity 2 SEA equatorial profiles is 1.28 m s 1. This difference, and 2D transport further underscores the TAO sam- which is both statistically and physically significant, has pling bias as estimated by SEA measurements. Biases likely been suspected but not established quantitatively. are also calculated and expressed as percentages of their Since equatorial observations are frequently used to quan- mean values, so as to mitigate the effect of seasonal and tify EUC strength as well as validate models and reanalysis longitudinal variations (Fig. 13). The percent bias b of products (e.g., Izumo 2005; Karnauskas et al. 2010, 2012; quantity X is calculated as

FIG. 10. (a) A simple model y 5 A sinx for 0.1 # A # 1.0, where A represents the maximum latitudinal meander of the EUC core and x may represent time or zonal distance. (b) Corresponding histograms of y for the range of values of A.

Unauthenticated | Downloaded 10/01/21 09:37 PM UTC 2022 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 31

FIG. 11. Scatterplots of the absolute latitude of the EUC core vs (a) longitude and (b) calendar day. Color scheme for (a) is west is red, central is green, and east is blue. Color scheme for (b) is red is spring and blue is December.

X 2 X results reported above that transport is not accurately b 5 100 3 eq core , X predicted by the maximum zonal velocity alone. Fur- core thermore, no dependence of the sampling bias on core where subscript eq denotes the value of X as measured latitude is found, suggesting that the inferred bias does directly on the equator and subscript core denotes the not appear to be simply a function of the distance of the value of X as measured at the latitude of the core of the EUC core from the equator. If the EUC in terms of its EUC. The X may be either maximum zonal velocity or 2D zonal velocity distribution in the latitude–depth plane transport. Equatorial measurements are consistently un- was constant, and it simply translated northward and derestimating the EUC strength in terms of zonal velocity southward relative to the stationary equatorial mooring, 2 with a mean bias of 20.14 6 0.03 m s 1 (210.38% 6 then the sampling bias would be dependent only on the 2.14%), and in terms of 2D transport with a mean bias of latitude of the center of the EUC. Since this is not the 2 219.04 6 8.92 m2 s 1 (29.87% 6 4.37%). Error bars are case, we hypothesize that there is substantial variability set to two standard errors of the mean, indicating 95% in the structure of the EUC (i.e., the shape and distri- conﬁdence limits. A nontrivial sampling bias in both zonal bution of zonal velocity about the centroid). Sampling velocity and 2D transport is evident, which is germane to biases in maximum zonal velocity and transport are also the interpretation of equatorial measurements as estimates tested for dependence on longitude and season (Fig. 14). of the EUC’s actual strength. There is no statistically signiﬁcant dependence of bias on Comparing the estimated biases in 2D transport with either longitude or season, which again may be limited by those of maximum zonal velocity (not shown) yields the relatively small sample size and longitudinal domain a weak positive correlation, which is consistent with sampled by the SEA dataset. A wider range in longitudes

21 FIG. 12. Scatterplots of (a) maximum zonal velocity (m s ) on the equator vs in the EUC core. 2 (b) As in (a), but for 2D transport (m2 s 1).

Unauthenticated | Downloaded 10/01/21 09:37 PM UTC SEPTEMBER 2014 L E S L I E A N D K A R N A U S K A S 2023

FIG. 13. Histograms of (a) bias of maximum zonal velocity and (b) 2D transport (%). and/or more cruises may further illuminate differences in (1990–2010 with two approximately year-long gaps). biases from the western, central, and eastern sections of Bias-corrected monthly values X0 are computed as the broader Paciﬁc basin.

Given the estimated bias in EUC velocity and 2D 0 X X 5 eq , transport noted above with little or no evidence of lon- 1 1 b gitudinal or seasonal variation within the study area, we can create a bias-corrected monthly record and where Xeq and b are as deﬁned above. Shown in Figs. 15a climatology of EUC strength from the ;19 years of and 15c are daily mean, monthly mean, and bias-corrected observations at the 1408W equatorial TAO mooring monthly mean time series of EUC velocity and 2D

FIG. 14. Distribution of biases in (a),(c) maximum EUC zonal velocity and (b),(d) transport as a function of (a),(b) longitude and (c),(d) calendar day. Error bars indicate 62 standard errors. Color scheme for (a) and (b) is west is red, central is green, and east is blue. Color scheme for (c) and (d) is red is spring and blue is December.

Unauthenticated | Downloaded 10/01/21 09:37 PM UTC 2024 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 31

21 FIG. 15. (a) Time series of daily mean (thin gray) and monthly mean (thick black) maximum zonal velocity (m s ) measured by ADCP from the TAO mooring at 08, 1408W over the period 1990–2010. The bias-corrected monthly mean record (described in section 4) is shown by a thin black line. (b) Monthly climatology (thick line) and bias-corrected monthly climatology (thin lines; dashed lines denote 95% 2 2 confidence limits on the bias correction) of maximum zonal velocity (m s 1). (c),(d) As in (a),(b), but for 2D EUC transport (m2 s 1; eastward zonal velocity integrated from 35- to 235-m depth). transport. There are clearly significant daily fluctuations in d The distribution of the latitude of the EUC core EUC strength about the monthly means (as discussed in throughout the SEA dataset suggests that the EUC section 3), which are likely a combination of temporal vari- core meanders in a sinusoidal fashion about the equator ations in EUC strength and latitudinal position relative to the with a meridional scale of ;0.48 latitude (;44 km). equator. Correcting for the 10.38% (9.87%) biases at each d On average, an estimate of peak EUC velocity or monthly mean velocity (2D transport) value results in bias- transport based on an equatorial profile of zonal corrected monthly records that correspond closely to (but do velocity is biased by 210%. not typically exceed) the positive spikes in the daily time Based on a high-resolution numerical ocean model 8 series. The 140 W EUC climatology can also be adjusted experiment capable of resolving the sensitivity of island- similarly (Figs. 15b,d), which increases the climatological 2 scale sea surface temperature (SST) to the strength of the EUC velocity by 0.09–0.14 m s 1 at the seasonal minimum 21 EUC (Karnauskas and Cohen 2012), a 10.4% increase in (February) and by 0.13–0.20 m s at the seasonal maximum the time-mean EUC velocity would lead to a 0.58C (May). The annual mean EUC velocity increases from 1.16 21 cooling of SST on the west side of the Gilbert Islands. to 1.27–1.33 m s in the bias-corrected climatology. Thus, an underestimate of the speed of the EUC of this magnitude also has practical consequences for ecological 5. Conclusions and discussion systems such as coral reefs across the equatorial Pacific. In theory, the EUC should have a mean tendency This paper directly compares SEA cruise measurements to be displaced to the north (south) of the equator to TAO moorings, when and where available, as well as to in the presence of mean southward (northward) merid- previous climatological observations of the EUC (Johnson ional wind stress (Charney and Spiegel 1971). It is et al. 2002); proposes a simple model of the EUC’s me- important to point out that the longitude where the cross- ander about the equator; and quantitatively estimates equatorial (meridional) component of the climatological a sampling bias in the maximum zonal velocity and trans- wind stress changes sign (i.e., from southward in the port of the EUC. The major conclusions are as follows. western Pacific to northward in the eastern Pacific) is d The SEA dataset compares well with TAO measure- ;1588W(Atlas et al. 2011), which is within the domain ments and is potentially a very useful resource; the sampled by the SEA dataset. Thus, it should be expected SEA ADCP data focusing on the EUC reproduce well that the TAO sampling bias estimated here for the central the mean longitudinal and seasonal structure of the Pacific be low compared to sites in the western or eastern EUC relative to prior observational work. Pacific where the large-scale mean meridional wind stress d Knowledge of the maximum zonal velocity (i.e., at the field is significantly nonzero and thus expected to set up core of the EUC) does not provide a close estimate of a larger mean tendency for the EUC to be displaced away 2D transport at the core of the EUC (R2 5 0.41), but from the equator. A preliminary survey of hundreds of 2D transport at the core of the EUC does provide cross-equatorial sections of zonal velocity measurements a fairly reasonable estimate of the full, 3D volume by shipboard ADCP not only reveals that the predicted transport of the EUC (R2 5 0.71). inverse dependence of the mean latitude of the EUC on

Unauthenticated | Downloaded 10/01/21 09:37 PM UTC SEPTEMBER 2014 L E S L I E A N D K A R N A U S K A S 2025 the mean cross-equatorial wind stress holds true, but that and oceanographic applications. Bull. Amer. Meteor. Soc., 92, the domain sampled by SEA in fact yields the smallest 157–174, doi:10.1175/2010BAMS2946.1. bias or difference between velocities on the equator and Charney, J. G., and S. L. Spiegel, 1971: The structure of wind- driven equatorial currents in homogeneous oceans. J. Phys. at the actual EUC core. Further analysis with this more Oceanogr., 1, 149–160, doi:10.1175/1520-0485(1971)001,0149: spatially comprehensive dataset, the Joint Archive for SOWDEC.2.0.CO;2. Shipboard ADCP (Firing 2013), along with numerical Drenkard, E. J., and K. B. Karnauskas, 2014: Strengthening of the models is necessary to confirm the robustness of the bias Pacific Equatorial Undercurrent in the SODA reanalysis: results presented herein and its dependence on other Mechanisms, ocean dynamics, and implications. J. Climate, 27, 2405–2416, doi:10.1175/JCLI-D-13-00359.1. variables. Farrar, J. T., and T. S. Durland, 2012: Wavenumber–frequency These results may lend further insight into the in- spectra of inertia–gravity and mixed Rossby–gravity waves in tegration of observational data with ocean and climate the equatorial Pacific Ocean. J. Phys. Oceanogr., 42, 1859– models to understand the strength of the mean ocean 1881, doi:10.1175/JPO-D-11-0235.1. circulation, ENSO events, and how they will respond to Firing, E., cited 2013: Notes from Acoustic Doppler Current Pro- filer Workshop at the National Oceanographic Data Center, radiative forcing and other large-scale changes. A greater May 14–15, 1992. [Available online at http://ilikai.soest. understanding of how these biases have influenced our hawaii.edu/sadcp/intro.html.] estimates of equatorial ocean circulation is necessary to ——, R. Lukas, J. Sadler, and K. Wyrtki, 1983: Equatorial un- fully understand the extent to which the EUC influences dercurrent disappears during the 1982–83 El Niño. Science, basin-scale and global climate, and thus what to expect 222, 1121–1123, doi:10.1126/science.222.4628.1121. Gove, J. M., M. A. Merrifield, and R. E. Brainard, 2006: Temporal from anthropogenic and natural climate variations. Using variability of current-driven upwelling at Jarvis Island. J. Geo- a similar approach but with vastly more data, one may phys. Res., 111, C12011, doi:10.1029/2005JC003161. propose improvements to the global network of oceano- Izumo, T., 2005: The equatorial undercurrent, meridional over- graphic observation systems, improve the usefulness of turning circulation, and their roles in mass and heat exchanges equatorial data by accounting for the sampling biases ex- during El Niño events in the tropical Pacific Ocean. Ocean posed here, and increase confidence in model assessments. Dyn., 55, 110–123, doi:10.1007/s10236-005-0115-1. Johnson, G. C., M. J. McPhaden, and E. Firing, 2001: Equato- Finally, understanding the mechanisms and time scales rial Pacific Ocean horizontal velocity, divergence, and for the apparent sinusoidal meander of the EUC could upwelling. J. Phys. Oceanogr., 31, 839–849, doi:10.1175/ illuminate better ways to understand, model, and predict 1520-0485(2001)031,0839:EPOHVD.2.0.CO;2. the behavior of the EUC along with its relationship to ——, B. M. Sloyan, W. S. Kessler, and K. E. McTaggart, 2002: atmospheric forcing. For example, apparent ;10% var- Direct measurements of upper ocean currents and water prop- erties across the tropical Pacific Ocean during the 1990s. Prog. iations in EUC strength observed from the equator that Oceanogr., 52, 31–61, doi:10.1016/S0079-6611(02)00021-6. are actually due to the meandering of the EUC away Karnauskas, K. B., and A. L. Cohen, 2012: Equatorial refuge amid from the mooring (perhaps due to temporal variations in tropical warming. Nat. Climate Change, 2, 530–534, doi:10.1038/ cross-equatorial wind stress or anomalous meridional nclimate1499. currents associated with tropical instability waves) clearly ——, R. Murtugudde, and A. J. Busalacchi, 2010: Observing the Galápagos–EUC interaction: Insights and challenges. J. Phys. have the potential to obfuscate the expected linear re- Oceanogr., 40, 2768–2777, doi:10.1175/2010JPO4461.1. lationship between zonal wind stress, the ocean’s zonal ——, C. J. Johnson, and R. Murtugudde, 2012: An equatorial ocean pressure gradient, and EUC transport. Such investi- bottleneck in global climate models. J. Climate, 25, 343–349, gations should be process oriented, drawing from more doi:10.1175/JCLI-D-11-00059.1. extensive ADCP datasets, autonomous platforms such as Kessler, W. S., M. J. McPhaden, and K. M. Weickmann, 1995: Forcing of intraseasonal Kelvin waves in the equatorial Pacific gliders, and high-resolution numerical models. Ocean. J. Geophys. Res., 100, 10 613–10 631, doi:10.1029/ Acknowledgments. The authors are deeply indebted 95JC00382. Lukas, R., 1986: The termination of the equatorial undercurrent in to Prof. Jan Witting of the Sea Education Association the eastern Pacific. Prog. Oceanogr., 16, 63–90, doi:10.1016/ (SEA) in Woods Hole, Massachusetts, for generously 0079-6611(86)90007-8. providing the SEA dataset and valuable guidance over Lyman, J. M., G. C. Johnson, and W. S. Kessler, 2007: Distinct 17- the course of this project. The authors thank the NSF and 33-day tropical instability waves in subsurface obser- Physical Oceanography program (OCE-1233282) and vations. J. Phys. Oceanogr., 37, 855–872, doi:10.1175/ JPO3023.1. the WHOI Academic Programs Office for funding. Madden, R. A., and P. R. Julian, 1971: Detection of a 40–50 day oscillation in the zonal wind in the tropical Pacific. J. Atmos. , REFERENCES Sci., 28, 702–708, doi:10.1175/1520-0469(1971)028 0702: DOADOI.2.0.CO;2. Atlas, R., R. N. Hoffman, J. Ardizzone, S. M. Leidner, J. C. Jusem, McPhaden, M. J., and Coauthors, 1998: The Tropical Ocean– D. K. Smith, and D. Gombos, 2011: A cross-calibrated multi- Global Atmosphere observing system: A decade of progress. platform ocean surface wind velocity product for meteorological J. Geophys. Res., 103, 14 169–14 240, doi:10.1029/97JC02906.

Unauthenticated | Downloaded 10/01/21 09:37 PM UTC DECEMBER 2014 C O R R I G E N D U M 2871

CORRIGENDUM

WILLIAM R. LESLIE Department of Geology, Oberlin College, Oberlin, Ohio, and Woods Hole Oceanographic Institution, Woods Hole, Massachusetts

KRISTOPHER B. KARNAUSKAS Woods Hole Oceanographic Institution, Woods Hole, Massachusetts

JAN H. WITTING Sea Education Association, Woods Hole, Massachusetts

In Leslie et al. (2014), the author list as shown on the title page was incomplete. Jan H. Witting, who was noted in the acknowledgments section, should have been listed as a full coauthor, as shown above. We regret any inconvenience this has caused.

REFERENCE

Leslie, W. R., and K. B. Karnauskas, 2014: The Equatorial Undercurrent and TAO sampling bias from a decade at SEA. J. Atmos. Oceanic Technol., 31, 2015–2025, doi:10.1175/JTECH-D-13-00262.1.

Corresponding author address: Kristopher B. Karnauskas, Woods Hole Oceanographic Institution, 360 Woods Hole Rd., MS 23, Woods Hole, MA 02544. E-mail: [email protected]

DOI: 10.1175/JTECH-D-14-00187.1

Ó 2014 American Meteorological Society Unauthenticated | Downloaded 10/01/21 09:37 PM UTC