Using datasets of sea surface temperature (SST), surface heat flux, upper ocean thermal data, and climatological temperature and salinity profiles, we try to detect ‘reemergence’ areas of winter SST anomalies in the world's oceans, and describe characteristics of these areas in terms of mixed layer depth (MLD), annual mean heat flux and properties of waters formed in winter mixed layer. Eventually, seven reemergence areas are found: four in the Northern Hemisphere and three in the southern Hemisphere. All areas have a large seasonal variation of MLD, and are the regions where annual mean heat fluxes are relatively small except for two regions in the Northern Hemisphere. In the viewpoint of water properties, it is found that these areas correspond to the mode water formation regions: subtropical mode waters of the Indian Ocean, South Pacific, South Atlantic, North Pacific and North Atlantic, and North Pacific central and North Atlantic subpolar mode waters.
 The purpose of the present study is to detect reemergence areas in the world's oceans and further to describe their characteristics in terms of mixed layer depth (MLD), annual mean surface heat flux, and properties of waters formed in the winter mixed layer. The remainder of this paper is organized as follows. In section 2, various kinds of datasets used are described. In section 3, reemergence areas detected in the world's oceans are described. In section 4, characteristics of reemergence areas are described. Section 5 gives our summary and remarks.
 In order to examine seasonal change of MLD and properties of waters formed in the winter mixed layer, we use gridded climatologies of temperature and salinity profiles of World Ocean Atlas 1998 [Boyer et al., 1998] (WOA98). In addition, to observe a reemergence process in a subsurface layer, the upper ocean thermal data of White  are also used. Since these oceanic data are given at the standard depths, the interpolated data of 10 m by an Akima's scheme [Akima, 1970] were prepared.
 The above datasets have different grid scales and data periods. However, in the present study, we did not unify both the grid scales and the data periods.
3. Reemergence Areas of Wintertime SST Anomalies
 So far, it is considered that mixed layer develops to the deepest depth in March (September) in the Northern (Southern) Hemisphere. However, month when the MLD is deepest naturally depends on oceanic condition and heat exchange between the ocean and the atmosphere [Takeuchi and Yasuda, 2003]. Therefore, in the present study, first we investigated the month showing the deepest MLD using WOA98. Here, the definition of MLD is the depth having the density greater by 0.125 sigma-theta than that at the sea surface. Eventually, it is found that deepest MLDs appear in February or March (August or September) in almost all areas of the Northern (Southern) Hemisphere oceans (not shown here).
 Next, using five SST datasets, lag correlation analysis for SST anomalies is made at each grid. The reference SST anomalies are those in the month having deepest MLD. In order to robustly detect reemergence areas, we adopt the following criterions. (1) Lag correlation coefficient has a minimum prior to reach a maximum. (2) The maximum of lag correlation coefficients exceeds a 99% significance level, the value of which depends on the data length of dataset used. (3) At least, three of five SST datasets satisfy the above two criterions. Therefore, reemergence areas detected in the present study are considered to be very conservative, since the above criterions are severe.
 As a result, we could detect seven reemergence areas as shown in Figure 1: four areas in the Northern Hemisphere and three in the Southern Hemisphere. Among them, five areas are situated around the pole-ward and western side of the subtropical gyre, and two areas are located in the northern subtropical gyre in the North Pacific and the subpolar gyre in the Northern Hemisphere. In order to clearly show behaviors of lag correlation coefficient and seasonal change of MLD, we set seven reemergence areas as follows: Region IO (55°E–73°E, 25°S–39°S) in the Indian Ocean, Region SP (159°E–179°W, 31°S–43°S) in the South Pacific, Region SA (15°W–29°W, 31°S–43°S) in the South Atlantic, Region n-NP (157°E–179°W, 37°N–45°N) and Region s-NP (141°E–159°E, 27°N–33°N) in the northern and southern North Pacific, and Region n-NA (19°W–35°W, 41°N–53°N) and Region s-NA (45°W–65°W, 33°N–39°N) in the northern and southern North Atlantic. These regions are set conservatively within the reemergence areas shown in Figure 1.
Figure 2 shows behaviors of lag correlation coefficient and seasonal change of MLD at seven areas. It is found that there are two types of reemergence areas in terms of behavior of lag correlation coefficients. That is, in Regions IO, SP, SA, n-NP and n-NA, lag correlation coefficients take maxima around one year later. It is also found that in these areas the second peak tends to appear around two years later. On the other hand, in Regions s-NP and s-NA, they reach maxima around 7 to 8 months later, and become smaller at one year later.
 All areas show a large seasonal variation in MLD. The annual difference of MLD amounts to 100 m or greater. That is, it can be said that large annual difference of MLD is necessary condition for reemergence of winter SST anomalies, as already pointed out by Timlin et al. .
Figure 3 shows a depth-time diagram of lag correlation coefficients of temperature anomalies in Regions n-NP, n-NA, s-NP and s-NA for the reference temperature averaged from the sea surface to 20 m in the month of the deepest mixed layer. The upper ocean thermal data from 1980 to 2002 are used in calculation. This figure clearly shows that the anomalous water formed in the winter deepest mixed layer is capped by the shallow seasonal thermocline in warming season, and then subsurface water is entrained due to the mixed layer deepening in the next cooling season. In Regions n-NP and n-NA, reemergence signals at the sea surface appear from approximately 7 months later to 13 months later: naturally, this is the same result as that shown in Figure 2. On the other hand, reemergence signals appear only from 6 months later to 8 months later in Region s-NP, and only from 8 months later to 9 months later in Region s-NA. That is, in Regions s-NP and s-NA, the waters formed in the winter deepest mixed layer tend to be refreshed in next winter. The reason why the reemergence signals disappear in the next mid winter in Regions s-NP and s-NA will be clarified in terms of annual mean heat flux in these areas.
4. Characteristics of Reemergence Areas in Terms of Annual Mean Heat Flux and Water Properties
Figure 4 shows a diagram of annual mean heat flux versus annual difference of MLD. Here the result of SOC surface flux [Josey et al., 1998] is shown. The results using the other two datasets also gave almost the same results, although data distributions were scattered more compared with Figure 4. We can point out that the data in Regions IO, SP, and SA in the Southern Hemisphere distribute within about ±20 W/m2, and those in Regions n-NP and n-NA in the Northern Hemisphere do within −20 to −100 W/m2. On the other hand, in Regions s-NP and s-NA, they distribute within the values of −50 to −200 W/m2. Regions s-NP and s-NA are situated in the areas east of large continents, where outbreak of the cold and dry air frequently occurs in winter. That is, it can be said that in Regions s-NP and s-NA, although winter SST anomalies reemerge in early fall, SST anomalies are completely refreshed due to a large amount of heat loss from the ocean. On the other hand, since heats gained by the oceans in Regions IO, SP, SA, n-NP and n-NA are consumed in the next cooling season as much as almost the same amount, winter SST anomalies can reemerge in the next winter.
 Next, we examine characteristics of properties of waters formed in the winter mixed layer. Figure 5 shows a temperature-salinity (T-S) diagram of waters formed in winter mixed layer, in which water properties of known mode waters are also drawn. This diagram clearly shows the wasters formed in the winter mixed layer in Regions IO, SP, SA, n-NP, n-NA, s-NP and s-NA correspond to subtropical mode waters (STMW) of the Indian Ocean (IOSTMW [Toole and Warren, 1993]), South Pacific (SPSTMW [Roemmich and Cornuelle, 1992]), South Atlantic (SASTMW [Provost et al., 1999]), North Pacific central mode water (NPCMW [Suga et al., 1997]), North Atlantic subpolar mode water (NASPMW [McCartney and Talley, 1982]), and subtropical mode waters of the North Pacific (NPSTMW [Masuzawa, 1969]) and North Atlantic (NASTMW [Worthington, 1959]), respectively. In geographical view, it is also found that these seven regions correspond well to known formation areas of mode waters [see Hanawa and Talley, 2001]. That is, it can be said that at least the formation regions of mode waters mentioned above are the reemergence areas.
5. Summary and Remarks
 In the present study, we tried to detect reemergence areas of winter SST anomalies in the world's oceans. Resultantly, seven areas were found: four in the Northern Hemisphere and three in the Southern Hemisphere. All areas corresponded to the regions where winter MLD develops to some degree and has a large annual difference of MLD. The two of seven areas were those where winter SST anomalies reappeared not in winter but fall, and the rest five regions were those where reemergence occurred in the next winter. In terms of annual mean heat flux, the latter five areas were the regions showing relatively small annual mean heat flux, while the former two areas were those showing negatively large annual mean heat flux. In terms of water masses, it was found that these reemergence areas corresponded to the mode water formation areas: IOSTMW (Region IO), SPSTMW (Region SP), SASTMW (Region SA), NPSTMW (Region s-NP), NASTMW (Region n-NA), NPCMW (Region n-NP), and NASPMW (Region n-NA).
 Previous authors [e.g., Alexander and Timlin, 1999] have reported much wider reemergence areas in the North Pacific: latitudinal belt along 40°N (see their Figure 2). Further Coëtlogen and Frankignoul  have pointed out that the reemergence could also occur in the remote area due to the advection by the strong current such as the Gulf Stream. In the present study, as mentioned in earlier section, we set rather severe criterions to firmly and robustly detect the reemergence areas. Therefore, the areas detected in the present study are very conservative.
 The authors wish to express their sincere thanks to members of Physical Oceanography Group at Tohoku University for their fruitful discussion. Two anonymous reviewers gave useful comments. This study was done as part of the 21st Century Center-Of-Excellence (COE) Program, ‘Advanced Science and Technology Center for the Dynamic Earth (E-ASTEC)’, at Tohoku University.