The propagation of density-compensated (warm/salty or cool/fresh) spiciness anomalies in the North Pacific thermocline is investigated using Argo profiles for the period 2001–2008. A cool/fresh spiciness anomaly on 25 < σθ < 25.5 kg m−3 isopycnals appears in the eastern subtropical North Pacific at 120°W–150°W in 2003–2004 with a salinity anomaly of about −0.15 PSS-78. This spiciness anomaly migrates southwestward, and arrives in the western tropical North Pacific at 145°E–175°W in 2008 with the salinity anomaly decreasing to about −0.043 PSS-78. Two warm/salty anomalies are observed to propagate along the same path from 2003 to 2005, and after 2005. The propagation path and speed of the anomalies are in good agreement with advection by the mean geostrophic current. In the course of propagation, the anomalies are diffused and are subject to high frequency injection of spiciness anomalies, especially in the eastern subtropical North Pacific.
 The subduction and propagation of density-compensated temperature and salinity water-mass perturbations, referred to as spiciness anomalies, from the mid-latitude to the equatorial Pacific have been hypothesized to play an important role in decadal variability in the Pacific [e.g., Gu and Philander, 1997]. While modeling studies have demonstrated the propagation of spiciness perturbations from the subtropics to the equator [Pierce et al., 2000; Nonaka and Xie, 2000; Yeager and Large, 2004; Fukumori et al., 2004], observational evidence has been more limited. The basin-wide network of XBT observations has tracked subducted temperature anomalies from the subtropics to tropics on decadal timescales [e.g., Schneider et al., 1999; Zhang and Liu, 1999; Luo and Yamagata, 2001], but only along repeat hydrographic lines [e.g., Suga et al., 2000] and at the Hawaii Ocean Time-series station [Lukas, 2001; Lukas and Santiago-Mandujano, 2008] could decadal spiciness anomalies on isopycnals be separated from undulations of the depth of isopycnals. The recent advent of the Argo network of drifting buoys allows, for the first time, the basin-wide description of spiciness signals. Using these observations, we describe the propagation and attenuation of spiciness anomalies in the thermocline from their source in the eastern subtropical North Pacific to the central- and western tropical North Pacific.
 We use two gridded products of temperature and salinity on isopycnal surfaces derived from Argo data. Also, we have verified spiciness anomalies described here in individual temperature and salinity profiles. Argo profiles for the period 2001–2008 are obtained from the US Global Argo Data Assembly Center and linearly interpolated onto standard levels and isopycnal surfaces. These interpolated profiles of potential temperature, salinity, potential density, isopycnal depth, and stratification are horizontally gridded using two different methods. Method I averages data for each month and within 3° latitude × 3° longitude bins, typically at least 50 profiles (Figure 1a) and then forms annual averages using weights dependent on the number of observations for each month. Method II employs a variational interpolation technique (K. Lebedev, in preparation, 2010; see also documentation at http://apdrc.soest.hawaii.edu/projects/Argo/data/Documentation/gridded-var.pdf) that minimizes the misfit between the irregularly distributed original data and the interpolated, monthly and annual averaged fields. The interpolated data are smoothed by twice applying a spatial filter with half-power cutoffs of 10° latitude and longitude.
 Montgomery potential [e.g., Cushman-Roisin, 1994] and geostrophic velocities on isopycnal surfaces are calculated from the bin-averaged hydrography using as reference level the mean 2001–2008 surface dynamic topography that is based on the mean surface dynamic topography [Maximenko and Niiler, 2005; Maximenko et al., 2009] and AVISO (Archiving, Validation and Interpretation of Satellites Oceanographic data) sea level anomalies [Ducet et al., 2000]. Results are robust if the 1000-m depth velocity field from Argo floats is used as a reference instead.
3. Mean State
 The depths of 25 < σθ < 25.5 kg m−3 isopycnals (Figure 1b) that connect the subtropical and tropical Pacific [Johnson and McPhaden, 1999] are characterized by a bowl shape in the subtropical gyre. The winter outcrop line of σθ = 25.25 kg m−3 is located between 30°N and 35°N from 140°E to 160°W and sharply turns northeastward east of 160°W (black dashed line in Figure 1b). The water subducted east of 160°W flows southward between 120°W and 150°W (white contours in Figure 1b) and then westward in the North Equatorial Current (NEC) between 9°N and 15°N to the western boundary. This subduction pathway is the focus of the present study. Potential vorticity is roughly conserved along this pathway except near the outcrop region (Figure 1c).
 In the eastern subtropics, spiciness gradients along isopleths of the mean Montgomery potential are large (color in Figure 2). In addition, the vertical salinity gradient in the upper ocean is unstable [Yeager and Large, 2007], and the base of the deep mixed layer associated with Eastern North Pacific Subtropical Mode Water formation [Hautala and Roemmich, 1998] is close to σθ = 25 kg m−3 (Figure 1d). These conditions favor the generation of spiciness anomalies in the upstream portion of the subduction path by anomalous advection across salinity/temperature gradients [Schneider, 2000] and diapycnal mixing below the mixed layer [Johnson, 2006; Yeager and Large, 2007], in addition to surface fluxes into the mixed layer [Bindoff and Mcdougall, 1994].
4. Propagation of Spiciness Anomalies
 Spiciness anomalies occur in the outcrop region at and north of 30°N, in the western subtropical gyre, the subduction path, the shadow zone in the eastern Pacific around 10°N, and at the equator. In the subduction path that follows the mean gyre circulation from the outcrop region in the eastern subtropics toward the western tropics, two warm/salty anomalies and a cool/fresh anomaly persist for several years, following the mean advective path (Figure 3). Results are similar for bin-averaged and interpolated datasets, and we mainly present results using the interpolated dataset.
 A cool/fresh anomaly propagates clockwise around the subtropical gyre from 2003 to 2008 (Figures 2 and 3). This anomaly is first detected at 18°N–30°N in the eastern subtropical North Pacific in 2003–2004 (Figures 3a–3b), although observations there are quite sparse in 2003 (not shown). This spiciness anomaly migrates southwestward along the mean Montgomery potential, and approaches the western boundary at 9°N–15°N in 2008 (Figures 3c–3f). When passing Hawaii around 2005–2006 it is consistent with observations at the Hawaii Ocean Time-series station (22°45′N, 158°W) of a freshening for 25 < σθ < 26 kg m−3 isopycnals of approximately 0.03 PSS-78 in 2005–2006 compared to 2004 and 2007 [Lukas and Santiago-Mandujano, 2008].
 Warm/salty spiciness anomalies are observed before and after the cool/fresh spiciness anomaly. Although propagation of both signals can be seen for 3–4 years only, the paths of these warm/salty anomalies are consistent with the path of the cool/fresh anomaly. In 2003, warm/salty waters are at 130°W–180° in the eastern subtropics (Figure 3a), propagate toward the southeast, and disappear in 2006 at the western boundary (Figure 3d). In 2005, a warm/salty spiciness anomaly is first detected at 30°N–40°N near the outcrop region (Figure 3c), and migrates southwestward to 20°N in 2008 (Figure 3f).
 The propagation speed of the spiciness anomalies matches mean advection speeds. We seed a passive tracer at the position of the 2008 cool/fresh spiciness anomaly (circles in Figure 3f) and estimate its upstream positions from the time averaged, geostrophic velocity on the isopycnals. This tracer location coincides with the cool/fresh spiciness anomaly from 2004 to 2007 (Figures 3b–3e). In particular in 2005 and 2007, the positions of tracer and spiciness anomalies show excellent agreement. This backtracking also suggests that the waters that make up the cool/fresh spiciness anomaly were subducted around 1999–2000, when Argo float density there was too low to investigate associated surface processes.
 Averaging salinity anomalies between the 1.6 and 3.4 m2 s−2 isopleths of the mean Montgomery potential confirms that the warm/salty and cool/fresh spiciness anomalies are primarily advected by the mean circulation (Figure 4), although the speed of the mean circulation is slower than the propagation of the spiciness signals in the upstream potion of the path. This Hovmoller diagram indicates a range of speeds from ∼1 cm s−1 in the eastern subtropics to ∼10 cm s−1 in the western tropics.
 An alternative hypothesis for the propagation of spiciness anomalies is a direct response due to anomalous advection by higher baroclinic mode Rossby waves, which propagate in the direction of and at a slower speed than a current [Liu, 1999]. However, anomalous velocities of large-scale, higher baroclinic modes are weak, and cannot account for the observed magnitude of spiciness signals. Furthermore, the spiciness anomalies are not accompanied by co-propagation of pressure anomalies on the same isopycnals (not shown). We therefore conclude that the propagation path and speed of the spiciness anomalies are explained better by the mean flow.
 The spiciness anomalies are modified along their path, with amplitude gradually decreasing with time (Figure 4). The amplitude of the salinity (potential temperature) anomaly of the cool/fresh spiciness signal decreases by about 70% from about −0.15 PSS-78 (−0.5°C) in 2004 to about −0.043 PSS-78 (−0.14°C) in 2008 (Figure 3). In addition, intra-annual variability is superposed, especially in the first three to four thousand kilometers from the outcrop line (Figure 4). This region is favorable for generation of spiciness anomalies on subducted isopycnals as mentioned in Section 3. Of particular note is a rapid shift from cool/fresh to warm/salty anomalies in winter 2007, likely a manifestation of spiciness anomaly generation on the subducted isopycnals.
5. Summary and Discussion
 The unprecedented temporal and spatial coverage of temperature and salinity observations by the Argo array allows observational analysis of the basin-wide, coherent propagation of spiciness anomalies (Figures 3 and 4). In particular, a cool/fresh spiciness anomaly can be tracked for 6 years from the eastern subtropical North Pacific in 2003 to the western tropical North Pacific in 2008 (Figure 2). The propagation path and speed of the spiciness anomaly are explained by the advection by the mean geostrophic current. In the course of propagation, the anomalies are attenuated and are subject to high frequency injection of spiciness anomalies, especially in the eastern subtropics.
 These observations raise further questions regarding the attenuation of spiciness anomalies and their downstream impact. An important process for the attenuation is temporal variability of velocity fields, which act to diffuse spiciness anomalies from the mean path. Indeed, a part of the cool/fresh anomaly overshoots the path to the shadow zone in the eastern Pacific around 5°N–10°N (Figures 2 and 3). This overshoot is consistent with the modeling result that intra-annual variability of the velocity field enhances tracer transports from the subtropics to equator through an interior path [Fukumori et al., 2004].
 The downstream impact of spiciness signals provided the initial motivation for this study. At the western boundary, the NEC bifurcates into the northward Kuroshio and the equatorward Mindanao current. Spiciness anomalies are expected to split into northward and equatorward components while being further attenuated by western boundary current mixing, consistent with the disappearance of the warm/salty anomaly in the western boundary region in 2006 (Figure 3d). The equatorial portion either enters the Indian Ocean via the Indonesian Throughflow [Stammer et al., 2008] or affects the temperature and salinity characteristics of the source waters of equatorial upwelling [Fukumori et al., 2004] and impacts tropical climate [Schneider, 2004]. Along the equator the variance of spiciness is large, but dominated by fluctuations of the strong spiciness front in water properties between southern and northern hemispheres [Johnson and McPhaden, 1999; Wang et al., 2004]. The quantification of the impact of the extratropical spiciness anomalies on the equatorial circulation is a difficult but an important future task. The anomalies in Figures 3 and 4 and the observation of freshening trend of −0.012 PSS-78 yr−1 for 25 < σθ < 26 kg m−3 isopycnals at the Hawaii Ocean Time-series station between 1991 and 2005 [Lukas and Santiago-Mandujano, 2008] suggest continuous forcing of the equatorial region by spiciness anomalies originating from the subtropical North Pacific.
 We thank two anonymous reviewers for comments that helped to improve the manuscript. This research was supported in parts by NSF OCE06-47994 and OCE05-50233, the Office of Science (BER), U.S. Department of Energy grant DE-FG02-07ER64469, and NASA NNX08AR49G. Additional support was provided through sponsorship of research activities at the International Pacific Research Center (IPRC) by the Japan Agency for Marine-Earth Science and Technology (JAMSTEC), by NASA through grant NNX07AG53G, and by NOAA through grant NA17RJ1230. This is IPRC contribution 670 and SOEST publication 7890. IPRC Argo products are public at http://apdrc.soest.hawaii.edu/projects/argo/.