Using outputs from a high-resolution OGCM, seasonal and interannual variations of the Angola Dome (AD) are revisited. Although the AD was previously considered to be one large cold tongue extending from the West African coast, it is shown that two cold domes exist. These two domes have remarkably different mechanisms for their seasonal variation. The weak dome, whose center is located at 6°S, 1°E, develops from May to September owing to the divergence of heat transport associated with upwelling. The strong dome, on the other hand, extends from the west coast of Africa between 20° and 15°S, and develops from April to August by the surface heat flux. The interannual variation of the weak dome is strongly influenced by the Atlantic Niño. An unusual relaxation of easterly wind stress in the central equatorial Atlantic Ocean associated with the Atlantic Niño triggers second baroclinic downwelling equatorial Kelvin waves, which propagate eastward along the equator and poleward along the coast after reaching the African coast as coastal Kelvin waves. Then, downwelling Rossby waves radiate away from the coast and cause significant warming in the weak dome region. The interannual variation of the South Equatorial Undercurrent may be associated with that of the AD; its transport decreases by 0.6 Sv, and its core shifts equatorward by 0.2° when the AD is anomalously weak.
There are several thermal upwelling domes in the World Oceans. One of those in the South Atlantic Ocean is called the Angola Dome (AD), which was originally identified by Mazeika (1967). He showed that the AD is located near 10°S, 9°E. The cold region extends northwestward from the West African coast (Fig. 1) and it is associated with the cyclonic turn of the South Equatorial Undercurrent (SEUC) off the Angola coast, which is also called the Tsuchiya Jet (TJ) in the South Atlantic (Tsuchiya 1986; Peterson and Stramma 1991). However, the existence of the AD remained uncertain because of the coarse observation (Voituriez 1981). Yamagata and Iizuka (1995, hereinafter YI95), using an ocean general circulation model (OGCM), examined the seasonal variation of the AD. They concluded that it is cooled between March and August, and that the surface heat flux plays a major role in its seasonal variation. This is a stark contrast with other domes in the world as they are formed basically by the regional wind-induced upwelling (Umatani and Yamagata 1991; Masumoto and Yamagata 1991; Vinayachandran and Yamagata 1998; Tozuka et al. 2002).
The AD has a large influence on the regional fisheries and climate of the surrounding countries. For example, when an influx of warm water from the equator toward Angola is unusually large, Sardinella aurita, a kind of sardine which likes cold upwelled waters, is repelled southward and is fished in northern Namibia (Binet et al. 2001). Also, when the sea surface temperature (SST) off the Angola coast is anomalously high, the precipitation over Angola tends to become larger (Hirst and Hastenrath 1983; Rouault et al. 2003).
Therefore, understanding the seasonal and interannual variability of the AD is very important from both societal and economical viewpoints. However, no study has been carried out on the interannual variations of the AD partly owing to the sparseness of observational data. One of the main modes of interannual climate variability in the Atlantic that may affect the AD is the Atlantic Niño. It is associated with a positive SST anomaly in the eastern equatorial Atlantic, which is similar to the Pacific El Niño (Zebiak 1993; Carton and Huang 1994) and peaks from austral autumn to austral winter. It is sometimes triggered by a relaxation of the zonal wind in the western equatorial Atlantic, which forces the downwelling equatorial Kelvin wave. After reaching the eastern boundary, it propagates poleward along the West African coast and may trigger the Benguela Niño, that is, the penetration of warm Angolan waters through the Frontal Zone (ABFZ) into much cooler Benguela regions located to the south of the Angola region (Shannon et al. 1986; Lass et al. 2000; Florenchie et al. 2003; Mohrholz et al. 2004).
Using outputs from a high-resolution OGCM, we have investigated the seasonal and interannual variations of the AD. The high-resolution is useful to study details of coastal Kelvin waves, which are resolved more finely in the current model. The present paper is organized as follows. A brief description of the model is given in section 2. Then, seasonal variations of the AD are studied in section 3, because understanding seasonal variations is crucial in understanding interannual variations (Tozuka and Yamagata 2003). In section 4, interannual variations of the AD are discussed for the first time in relation with the Atlantic Niño. Its possible link with interannual variations of the TJ is also discussed. The final section is reserved for a summary and discussion.
2. Model description (OFES)
The OGCM used in this study, which is called OGCM for the Earth Simulator (OFES), is based on the Modular Ocean Model (MOM3; Pacanowski and Griffies 2000) with its code tuned for optimal performance on the Earth Simulator (Masumoto et al. 2004). The computational domain covers a near-global region extending from 75°S to 75°N except for the Arctic Ocean, with a horizontal grid spacing of 0.1°. There are 54 vertical levels with varying distance from 5 m at the surface to 330 m at the maximum depth of 6065 m. The model topography is constructed from the 1/30° bathymetry dataset created by the Ocean Circulation and Climate Advanced Model (OCCAM) project at the Southampton Oceanography Center [obtained through the National Oceanic and Atmospheric Administration (NOAA) Geophysical Fluid Dynamics Laboratory (GFDL)]. The viscosity and diffusivity vary in space, with the coefficients proportional to the cube of the zonal distance between the grids (Smith et al. 2000). The background horizontal biharmonic viscosity and diffusivity are −27 × 109 and −9 × 109 m4 s−1, respectively. For the vertical mixing, the K-profile parameterization boundary layer mixing scheme (Large et al. 1994) is adopted.
Monthly mean wind stress data from the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis data for the period from 1950 to 1999 (Kalnay et al. 1996) are used for the climatological seasonal integration. The surface heat flux is calculated using the monthly mean climatology from the NCEP–NCAR reanalysis data by adopting the bulk formula of Rosati and Miyakoda (1988). Also, the precipitation rate from the same reanalysis data is used to obtain the freshwater flux. In addition to this freshwater flux, the surface salinity is restored to the monthly mean sea surface salinity (SSS) of the World Ocean Atlas 1998 (WOA98; Conkright et al. 1998), with the restoring time of 6 days to include the contribution from the river runoff. In the buffer zones within 3° from the northern and southern artificial boundaries, the temperature and salinity fields are restored to the monthly mean climatological values (WOA98) at all depths. The restoring time at those boundaries is 1 day and it increases up to the infinity in the interior region. The model is integrated for 50 yr from the annual mean temperature and salinity fields (WOA98) without motion. After this spinup process, the model is driven by the daily NCEP–NCAR reanalysis data from 1958 to 2003, with the same restoring boundary condition for SSS.
For comparison, we use subsurface temperature data from the World Ocean Atlas 2001 (WOA01; more information available online at http://www.nodc.noaa.gov) and the assimilation data of Masina et al. (2001). Also, we use the Hadley Centre Sea Ice and SST dataset (HadISST; Rayner et al. 2003) for the sea surface temperature anomaly (SSTA) and the Ocean Topography Experiment (TOPEX)/Poseidon data (see online at http://podaac.jpl.nasa.gov) for the sea surface height anomaly (SSHA).
3. Annual cycle
a. Subsurface temperature and current fields
To study the seasonal march, we have calculated monthly climatology by averaging the model monthly data for 20 yr from 1982 to 2001. Interestingly, we find two domelike cold regions in the subsurface annual mean temperature at a depth of about 50 m (Fig. 1c). Although McClain and Firestone (1993) and Signorini et al. (1999) show the AD in the map of Levitus 100-m temperature, the thermal domes are also prominent at a depth of around 50 m with the maximum geostrophic velocity (see Fig. 4 in Gordon and Bosley 1991). One dome is weaker and centered around 6°S, 1°E, and the other is stronger and centered around 16.5°S, 10.5°E. Those are also captured by Masina et al. (2001) as shown in Fig. 1b. In contrast, the WOA01 climatology (Fig. 1a) does not show clearly those two separate cold regions. Instead, one large cold tongue extends from the West African coast. This difference may be due to significant smoothing and coarse resolution of the WOA01 data. There is a possibility that the so-called AD may be associated with two domelike cold regions, in contrast to the previous image that the AD is one large cold region located near 10°S, 9°E (cf. Mazeika 1967; YI95). Gordon and Bosley (1991) suggested two shallowest peaks of the 12°C isotherm at about 150-m depth. Therefore, we state that the AD is composed of the weak dome and the strong dome at a depth of 54 m.
The two cold domelike patterns in our model results are more clearly seen in vertical sections of temperature (Fig. 2). The weak one is seen along the section of 6°S and 1°E, while the strong one is found along the section of 16.5°S and 10.5°E. The weak dome appears more clearly in the latitudinal section than in the longitudinal section. Also, the weak dome is slightly shallower than the strong dome, and cannot be found at the surface. On the other hand, the strong dome is associated with a cold SST anomaly. Although these domes are found in the upper 300 m as reported by Gordon and Bosley (1991), the seasonal variations are remarkable only above a depth of 120 m. Figure 3 shows the annual march of the simulated AD. The two cold domes exist throughout the year and they are particularly distinguishable from the end of austral winter to spring, in agreement with YI95. The traditional dome seems to correspond to the weak dome, because of the proximity of its location. Since isotherms show domelike patterns off the coast, the strong dome is not just due to coastal upwelling. In addition, the upwelling induced by the wind stress curl is one order of magnitude larger than the coastal upwelling associated with the alongshore wind. The Cape of Fria, which is located to the east of the strong dome, may be partly responsible for introducing the strong cyclonic curl of wind stress in the region.
We see cyclonic gyres associated with these two domes in Fig. 4. Gordon and Bosley (1991) actually reported the Angola gyre in the tropical South Atlantic from the observational data; it was identified as a large cyclonic gyre centered near 13°S, 5°E, and confined for the most part to the upper 300 db. The eastward current around 4°S band, which flows along the northern rim of the weak dome, is the TJ. There is a clear distinction between the Equatorial Undercurrent (EUC) and the TJ. The latter turns southward around 5°E and is called the Angola Current (AC). Although the EUC can affect the AC (Wacongne and Piton 1992), the TJ seems to be a main source of the AC in the annual mean. This is consistent with the recent observation by Mercier et al. (2003), who suggested that the contribution of the EUC to the Gabon–Congo Current is weak. At about 12°S, there is an indication of the South Equatorial Countercurrent (SECC), which is consistent again with the observation by Mercier et al. (2003), despite the fact that the modeled SECC is much weaker (about half) than the observed. The SECC may play an important role to separate the two parts of the AD. We note that the cold northwestward Benguela Current is always found along the southwestern border of the strong dome.
b. Heat budget
The simulated AD is well located within the region of wind-induced Ekman upwelling (Fig. 4). To understand its seasonal variation quantitatively, we consider two artificial boxes A and B in the upper 58 m (Fig. 1a) and calculate the model seasonal cycle of the heat budget and upwelling speed (Fig. 5). The seasonal variations of the heat budget in these boxes are determined by three terms: the surface heat flux, the convergence of the heat transport, and the diffusion. The latter two terms are computed as a whole as difference between the tendency term of heat content and the surface heat flux. Also, we assume that the upwelling in these boxes is composed of the local Ekman pumping and the remote effect that is calculated by subtracting the local Ekman pumping from the vertical velocity. We have confirmed that the results presented here are not sensitive to around a 10-m change in the choice of the bottom of the box.
The heat content in box A, located at the center of the weak dome region (7.5°–4.5°S, 0.5°W–2.5°E), decreases from May to September (corresponding to the austral winter) and increases from October to April (corresponding to the austral summer; Figs. 3 and 5). The divergence of the heat transport related to upwelling seems to dominate the cooling tendency during the austral winter with a peak in August. It is associated more with the local wind-induced upwelling from August to September. From May to July, however, it is mostly due to the remote forcing. Since upwelling to the east of box A is mainly induced by the local wind and the strongest in May (figure not shown), we suggest that the remote forcing is due to upwelling Rossby waves propagating westward. The surface heat flux also shows seasonal variations in accord to the annual march of the solar insolation; it is positive from August to April, playing a major role in the seasonal warming. However, it plays only a minor role in seasonal cooling.
In box B (18°–15°S, 9°–12°E), which covers the strong dome region, the heat content decreases from April to August and increases from September to March (Figs. 3 and 5). This is in phase with the evolution of the weak dome, but the mechanism is different; the surface heat flux plays a more important role in the strong dome. The divergence of heat transport exists all year-round. This is because the upwelling is strong except for the austral summer season and dominated by local wind forcing in this region. Thus, the thermocline is kept shallow at about 30-m depth (Fig. 2). As a result, the seasonal variation of seawater temperature is determined mostly by the surface heat flux. This was already discussed in YI95, where they showed that the AD is driven by the negative surface heat flux over the region (9°–11°S, 8°–10°E). However, we note that only the strong dome is explained by this mechanism.
Why is the AD associated with two domelike cold regions? One possibility is the SECC mentioned in the previous section. Another possibility is two cores in the local wind stress curl. We have calculated the heat budget of the area located between the weak dome and the strong dome. Although the surface heat flux plays an important role there as in the strong dome region (box B), the local Ekman upwelling is not so strong. Therefore, the region between box A and box B is less cold.
4. Interannual variations
a. Relation with the Atlantic Niño
We focus here on the interannual variation of the weak dome. Figure 6 shows the interannual variation of the upper 58-m heat content anomaly (HCA) and the SSHA in box A. The anomaly field is defined as deviation from the monthly mean climatology. Also shown is the Atlantic Niño index (ATL3), which is calculated by averaging quantities over 3°S–3°N, 20°W–0°E, and captures strong interannual fluctuations associated with the Atlantic Niño (Zebiak 1993). To confirm the validity of using the ATL3 in the present model, we have conducted EOF analyses on the model SSTA in the equatorial Atlantic (figure not shown). The first (EOF) mode, which explains 27% of the total variance, captures the variability associated with the Atlantic Niño. As is clear from Fig. 6, the interannual variations associated with the AD and the Atlantic Niño are relatively well reproduced in the model. The lagged correlation between HCA in box A and SSTA in ATL3 shows the maximum of 0.49 when the latter leads the former by 1 month. The correlation between SSHA in box A and SSHA in ATL3 shows the maximum of 0.44 without a lag. Since these values are significant above 99% confidence level based on the nonparametric test, we may conclude that the AD is related with the Atlantic Niño.
Figure 7 shows the subsurface temperature at a depth of 54 m from May to July, for each year from 1982 to 2001. As shown in Fig. 8, there are two seasonal peaks in the standard deviation of interannual variation of HCA of the AD. One is in February and the other is in June. The high variability in February appears to be due to the meridional dipole mode, which is known to develop in this season. We focus here on the peak in June to discuss the relation between the AD and the Atlantic Niño. In particular, we discuss the warm event in 1999 (Mohrholz et al. 2004). This is because we can compare our model results with in situ data from the R.V. Poseidon cruise (John et al. 2004). The OFES captures the intense intrusion of the warm Angola Current into the northern Namibian waters (Figs. 9a–d). This is in good agreement with the in situ observation of Mohrholz et al. (2001) that describes the horizontal distribution of the temperature at 50-m depth from the same cruise data (Figs. 9e,f). As seen in Fig. 7, the AD is warmer in May–July (MJJ) 1999. To understand the mechanism in detail, we have calculated time series of the anomalous heat budget and upwelling speed in box A (Fig. 10). The anomalous rate of change of heat content is positive from January to June with a peak in May. Because the AD starts to cool normally during this season, this anomalous warming is rather unusual. The unusual event is due mainly to the divergence of the heat transport related to downwelling excited remotely. This is further supported by Fig. 11, which shows interannual variations of SSHA along the zonal band of box A (along 6°S). The positive SSHA associated with downwelling in box A may be traced back to the coastal area. The second baroclinic Rossby wave may be responsible for this process. This is because its propagation speed of about 0.15 m s−1 corresponds to the long Rossby wave speed calculated from the internal wave phase speed given in Table 1.
The anomaly in the decomposed vertical speed in box C, which covers a coastal region (7.5°–4.5°S, 9°E to the West African coast), is then examined (Fig. 11c). The strong downwelling from March to April 1999 is clearly due to the remote effect. It is linked with the propagation of downwelling coastal Kelvin waves. The propagation is a regular seasonal phenomenon (YI95; Schouten et al. 2005), but the case in 1999 is unusually strong. To understand more details, we have decomposed the vertical speed at the coastal point (5.5°S, 11°E) into contributions from five normal vertical modes of coastal Kelvin waves, by calculating vertical structure functions from the potential density profile. The reconstructed vertical speed from March to April 1999 at a depth of 58 m is mainly explained by the second baroclinic coastal Kelvin wave (Fig. 12 and Table 1).
The second baroclinic coastal Kelvin wave we have identified is traced back to the equator. To demonstrate this, the interannual variations in SSHA and the zonal wind stress anomaly along the equator are shown in Fig. 13. We find the unusual relaxation of the trade wind from February to March in 1999, which may excite the downwelling equatorial Kelvin wave in the central equatorial region near 20°W.
To examine the above possibility, we have calculated Dn, which is called an equivalent forcing depth (e.g., Gill 1982):
where An(z) is the vertical structure function, H is the bottom depth, and Hmix is the mixed layer depth. We note that the wind stress can efficiently excite the vertical mode for which Dn is small. Since Dn at 0°, 20°W is the smallest for n = 2 (Table 1), the second mode must be forced efficiently. This is in agreement with the observational data (du Penhoat and Treguier 1985; Verstraete 1992) and the OGCM results (Illig et al. 2004). The above confirms that the second baroclinic downwelling equatorial Kelvin wave forced by the unusual relaxation of zonal wind from February to March 1999 in the central equatorial region near 20°W have propagated eastward along the equatorial waveguide. After reaching the eastern coast from March to April 1999, the equatorial Kelvin wave may have propagated along the West African coast as coastal Kelvin wave. The signal then may have propagated westward from the coast to the AD region as long Rossby waves. The interannual variation of the AD in years such as 1984, 1988, and 1995 may be explained in this way. In contrast, variations in years such as 1992, 1994, 1997, and 2000 when the AD was anomalously strong seem to be explained as a mirror image of the above.
b. Possible links with the Tsuchiya Jet
Tsuchiya (1986) showed that the TJ exists in the southern Atlantic, which is an eastward current in the thermostad layer of almost uniform potential density under the thermocline between 3° and 6°S (Mercier et al. 2003; Bourles et al. 2002). As it flows eastward, its core rises from about 250- to 150-m depth and shifts poleward from 3°S near the western boundary to 4.5°S at 26°W (Schott et al. 2004; Brandt et al. 2006). These features are well reproduced by the present model (Fig. 4). The TJ may influence the equatorial condition by partially blocking the equatorward heat transport, and carrying water eastward into the dome. McCreary et al. (2002) recently suggested a link between the TJ in the Pacific and the off-equatorial eastern upwelling region. Also, Kessler (2002) showed that the TJ shifts equatorward owing to the extremely shallow thermocline of the Costa Rica Dome in the eastern Pacific (cf. Umatani and Yamagata 1991). Thus, it is of interest to investigate a possible link between the interannual variation of the TJ and that of the AD in the Atlantic.
We have compared the simulated TJ with the observational data by Schott et al. (2004; Fig. 14). The simulated TJ has a core at a depth of about 120 m with a maximum speed of 0.25 m s−1 at 4°S. Although it is somewhat shallower, weaker, and closer to the equator than the observed one, the OFES is reasonably successful in reproducing the TJ. The simulated TJ is also in good agreement with other observations (Bourles et al. 2002; Mercier et al. 2003).
To clarify relationship between the AD and the TJ, we adopt a composite method. Five typical warm AD years (1984, 1986, 1988, 1995, and 1999) and five typical cold AD years (1992, 1993, 1994, 1997, and 2000) are selected based on the upper 58 m HCA in box A. Here, we focus on conditions at 10°W as the TJ reaches the western rim of the AD at this longitude. Figure 15 shows that the core of the TJ in the warm years shifts equatorward by 0.4° than that in the coldest years. These results are reliable above the 90% significance level when based on the nonparametric test. This meridional migration may be interpreted in the following way. The shallower thermocline associated with the cold AD causes the thermostad layer to stretch vertically, which in turn causes the TJ to move poleward to conserve its potential vorticity. If the distance between the 26.3 and 26.7 isopycnals at 3.5°S stretches from 130 to 150 m in Fig. 15, the core must shift poleward by 0.5° to conserve its potential vorticity ( f /h). The mean transport of the TJ, which is integrated for the core above 0.1 m s−1 from May to July, is 2.6 Sv (1 Sv = 106 m3 s−1). It decreases by 0.6 Sv at above 97% significance level in the warm years, while that in cold years increases by 0.7 Sv. This may be linked with the difference in the vertical velocity in the AD; the upwelling in the cold AD years is 1.7 times stronger that that in the warm AD years. The above interesting feature, extending westward to about 25°W, supports the intriguing result of McCreary et al. (2002). More recently, Furue et al. (2007) have suggested that off-equatorial upwelling in the southeastern Pacific is important for the dynamics of the southern TJ.
Most years used in this composite analysis are the Atlantic Niño years (1984, 1988, 1995, and 1999) for the warm AD composite and the Atlantic Niña years (1992, 1994, 1997, and 2000) for the cold AD composite. Therefore, we suggest that the interannual variation of the TJ may be related to the Atlantic Niño/Niña at least through variations of the AD as discussed in the previous section. However, the warm AD in 1986 and the cool AD in 1993 cannot be explained by the above link; those appear to be more related to local wind-induced upwelling.
5. Summary and discussion
Using outputs from the high-resolution OGCM, the seasonal cycle of the oceanic conditions of the AD is revisited. The high-resolution (0.1° in both the latitude and longitude) of the present model allowed us to suggest a new interpretation of the AD. In particular, two cold domes are identified in the Angola basin. One is the weak dome, whose center is located at 6°S and 1°E, and the other is the strong dome, which extends from the west coast of Africa between 20° and 15°S.
The weak dome develops from May to September. The divergence of the heat transport associated with upwelling tends to reduce the heat content of the weak dome adiabatically. The upwelling is not only due to the local wind stress curl, but also due to the remote forcing, which is forced to the east of this dome by local wind curl. In contrast, the cooling associated with the surface heat flux plays only a minor role. The strong dome develops from April to August. The divergence of the heat transport associated with locally forced upwelling tends to reduce the heat content of the surface layer in this area throughout a year. This is why the thermocline of this region is basically shallow. In contrast to the weak dome, the seasonal variation of the surface heat flux thus plays an important role in introducing the seasonality in the strong dome.
The interannual variation of the AD is affected by the Atlantic Niño, and the warm event in 1999 has been discussed in detail. It is shown that the second baroclinic downwelling equatorial Kelvin wave forced by the unusual relaxation of zonal wind from February to March 1999 in the central equatorial region (about 20°W) propagates eastward along the equatorial waveguide. After reaching the eastern coast, it propagates along the coast from March to April 1999 as a form of coastal Kelvin wave. This effect propagates westward off the West African coast, and reaches the AD as a form of long Rossby wave. The interannual variation of the AD in years such as 1984, 1988, and 1995 shows the similar variation. In contrast, years such as 1992, 1994, 1997, and 2000 are close to a mirror image.
The warmer AD in early 1986 cannot be explained by the above mechanism, because the Atlantic Niño is not observed and the peak of warming in the AD is somewhat earlier. This event may be related to the strong negative event of the meridional mode, which is characterized by cold (warm) SST anomalies in the north (south) of the equator (Chang et al. 1997). The meridional mode peaks from austral summer to austral fall and is formed thermodynamically by the wind–evaporation–SST feedback (Xie and Philander 1994). Since the AD region is identified as the southern pole of this mode, the oceanic dome and the meridional mode are expected to interact with each other through variations of the subtropical high. Further studies in this direction are needed.
The core of the TJ in the warm AD years shifts equatorward by 0.4° more than in the cold AD years. This may be explained by the fact that the shallow thermocline associated with the strong AD causes the thermostad layer to stretch vertically, which in turn causes the TJ to move poleward to conserve its potential vorticity. Also, its transport in the warmest years is decreased by 0.6 Sv, while that in coldest years is increased by 0.7 Sv. This feature extends to the western area of about 25°W. It is interesting to see how these variations of the TJ influence the meridional heat transport and the horizontal circulation in the Atlantic. Further studies are necessary to investigate the possible link with the TJ.
The above new results are based on the OGCM outputs. Because of lack of systematic, high-density oceanic measurement in the southeastern tropical Atlantic, it is rather difficult to compare the above result with the observations to a full extent. However, we expect that observational efforts such as the Pilot Research Moored Array in the Tropical Atlantic (PIRATA; Servain et al. 1998) will improve the situation and enhance our understanding of the climate variability in the AD region.
We thank J. R. E. Lutjeharms, J. P. McCreary, and M. Rouault for stimulating discussions. The OFES was run on the Earth Simulator of Japan Agency Marine-Earth Science and Technology. Constructive comments provided by the editor and three anonymous reviewers helped us to improve the earlier manuscript. The present research is supported by the Japan Society for Promotion of Science through both Grant-in-Aid for Scientific Research (A) and the 21st century COE grant for the “Predictability of the Evolution and Variation of the Multi-scale Earth System: An integrated COE for Observational and Computational Earth Science.” Doi is also supported by the Sasakawa Scientific Research Grant for doctoral students from the Japan Science Society.
Corresponding author address: Takeshi Doi, Department of Earth and Planetary Science, Graduate School of Science, University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-0033, Japan. Email: email@example.com