Circulation and Mixing of Water Masses of Tatar Strait and the Northwestern Boundary Region of the Japan Sea

Transcription

1 Journal of Oceanography, Vol. 55, pp. 133 to Circulation and Mixing of Water Masses of Tatar Strait and the Northwestern Boundary Region of the Japan Sea STEPHEN C. RISER 1, MARK J. WARNER 1 and GENNADY I. YURASOV 2 1 School of Oceanography, Box , University of Washington, Seattle, Washington 98195, U.S.A. 2 Pacific Oceanological Institute, 43 Baltiyskaya Street, Vladivostok , Russia (Received 4 October 1998; in revised form 10 February 1999; accepted 15 February 1999) The deep waters of the northern portions of the Japan Sea are examined. It is found that the flow regime south of the southern Tatar Strait region is generally cyclonic in the upper ocean, with only weak flows present below depths of a few hundred meters. The Japan Sea appears to be remarkably well-mixed below depths of a few hundred meters, both horizontally and vertically. Based on chlorofluorocarbon measurements, it is concluded that the deep waters of the Japan Sea have been only weakly ventilated in recent decades. Results from a simple box model suggest two possible scenarios for the ventilation of the Japan Sea since the 1930s. In the first scenario, deep ventilation of the Japan Sea was relatively weak, but constant, from the 1930s to the present, with a deep-water residence time of approximately 500 years. In the second scenario, ventilation was relatively vigorous through the mid-1960s, with a deep-water residence time of approximately 100 years; after the mid-1960s, the ventilation of the deep waters stopped. The model results are consistent with the idea that presently the ventilation of the deep water of the Japan Sea is weak or nonexistent. Keywords: Chlorofluorocarbons, deep circulation, dissolved oxygen, Japan Sea, mixing, ventilation. 1. Introduction The northernmost regions of the Japan Sea are generally ice-covered in winter, and for this reason it has been speculated that some of the densest water of the Japan Sea might originate in this area. Makarov (1905) was perhaps the first to suggest publicly that active convection might take place in the northern Japan Sea in winter, and later Sudo (1986) and Martin et al. (1992), among other investigators, suggested that at least some of the deep water of the Japan Sea might result from winter conditions and the formation of sea ice in these areas. In order to examine this possibility, two hydrographic sections were occupied in the northern part of the Japan Sea during a cooperative US-Russian expedition to study the Okhotsk and Japan Seas that took place in April and May of Standard hydrographic variables, nutrients, and dissolved oxygen were measured at all stations, and chlorofluorocarbon (CFC) observations were also collected at many sites. A report on these observations, and a discussion of the inferences that can be made concerning deep water formation, circulation, and mixing in the northern Japan Sea, is the topic of this paper. 2. Data The data used in this work are a subset of a larger dataset obtained during a joint US-Russian expedition in the Okhotsk and Japan Seas that took place in April and May of 1995 aboard the research vessel Akademik Lavrentyev of the Pacific Oceanological Institute, Russian Academy of Sciences, in Vladivostok, Russia. Scientists from the University of Washington and Scripps Institution of Oceanography in the US and the Pacific Oceanological Institute in Russia participated in the cruise. The data were collected using CTD and chemical analysis equipment provided by the University of Washington and the Ocean Data Facility at Scripps Institution of Oceanography. While the work was not formally a part of the World Ocean Circulation Experiment (WOCE), it was a major goal of the expedition to collect data of a quality that met WOCE standards of accuracy and precision (see WOCE Data Handbook, 1994); in general, this goal was achieved. CTD data were collected using a Neil Brown Mark 3 system equipped with a General Oceanics rosette sampler, with liter Niskin Bottles. For casts where the depth was greater than 1000 m, all 24 bottles were used to collect chemical samples. For shallower casts, fewer than 24 bottles were used. All casts were taken to within 15 meters of the bottom of the ocean. Water samples were collected from each Niskin bottle, and a conductivity/salinity determination was made at sea for each sample using a Guildline AutoSal salinometer and UNESCO standard seawater. The results of this analysis were used to adjust the values of conductivity measured by the CTD package. Using this method, it is estimated that the Copyright The Oceanographic Society of Japan. 133

2 values of salinity measured from CTD are accurate to approximately.002 practical salinity units (PSU). The CTD was calibrated in a laboratory prior to and immediately following the cruise, and it was determined that the temperature accuracy was better than.001 C. Dissolved oxygen samples were collected from each Niskin bottle and analyzed using an automated system based on the modified Winkler method (Carritt and Carpenter, 1966); from comparisons with known standards at the time of analysis, it is estimated that the dissolved oxygen samples are generally good to.05 ml/l. Nutrients (dissolved nitrate, nitrite, phosphate, and silicate) were collected from each Niskin bottle and were analyzed using a Technicon autoanalyzer. By comparing duplicate and triplicate samples on several casts with known standards, it was determined that these data are generally good to about 1% of the measured values. In addition to these standard hydrographic parameters, the chlorofluorocarbons CFC-11 and CFC-12 were measured at many stations, using the method outlined in Bullister and Weiss (1988). To avoid contamination from ship-based sources of CFC, a portable analysis van was installed on the ship prior to the cruise, and all CFC analyses were conducted inside of this laboratory. The resulting CFC data were of good quality, with measurement accuracy and precision judged to be 1.2% or.006 pmol/kg, whichever is greater, for both CFC- 11 and CFC-12. The concentrations are reported on the SIO calibration scale (Cunnold et al., 1994), and a sample handling blank of.006 pmol/kg was subtracted from all CFC-11 samples (there was no detectable blank for CFC- 12). 3. Sections As shown in Fig. 1, the two sections discussed in this paper extend across the southern portion of Tatar Strait at approximately 46 N, from Sakhalin Island on the east to Maximova Point on the Siberian coast (Stations ), and from the center of the Japan Sea perpendicular to the Russian coastline between 44 N and 45 N (Stations 177 Fig. 1. Bathymetric chart of the Japan Sea, showing location of stations where data were collected during the 1995 US- Russian Akademik Lavrentyev expedition and limits of ice cover on 3/1/95 and 5/1/95. The ice cover estimates were made by Prof. Seelye Martin of the University of Washington using satellite-based SMMR techniques. Fig. 2. Schematic map of the near-surface circulation of the Japan Sea, from Yurasov and Yarichin (1991). PLS = Primorye/ Liman/Shrenk Current; TC = Tsushima Current; TS = Tsugaru Strait; and SI = Sakhalin Island. 134 S. C. Riser et al.

3 183). The flow in this region is generally thought to be intensified along the coasts, as can be seen in the schematic diagram shown in Fig. 2, reproduced from Yurasov and Yarichin (1991). A number of investigators have suggested that the circulation of the Japan Sea consists of a three-gyre system, as shown in Fig. 2. However, the flow patterns shown in Fig. 2 are undoubtedly highly seasonally dependent and subject to the influence of eddies and other transients and thus should be viewed as suggestive and not conclusive. Indeed, one purpose of the Lavrentyev expedition was to collect high quality hydrographic and chemical data that could be used to refine the schematic ideas shown in Fig. 2. (a) (b) Fig. 3. (a) Section of potential temperature across Tatar Strait, from data collected on the 1995 US-Russian Akademik Lavrentyev expedition. Station numbers across the top of the figure correspond to station positions shown in Fig. 1. (b) As for (a), but for salinity. (c) As for (a), but for potential density. (d) As for (a), but for dissolved oxygen. (e) As for (a), but for silicate. Circulation and Mixing of Water Masses of Tatar Strait and the Northwestern Boundary Region of the Japan Sea 135

4 3.1 Tatar Strait Sections of potential temperature (θ), salinity (S), potential density (σ θ ), and oxygen and dissolved silicate across Tatar Strait, shown in Figs. 3(a) (e), indicate the degree to which the Japan Sea is approximately a two-layer system: in the spring of 1995, when these measurements were collected, the upper portion of the water column consisted of a 200 m thick layer of relatively warm, fresh water, having low density, low silicate, and high dissolved oxygen. Below this surface layer the temperature and salinity are generally quite well-mixed, with variations in θ, S, and σ θ amounting only to 1.5 C,.03 PSU, and 0.1 sigma-units between a depth of 200 m and the seafloor. The distribution of σ θ across Tatar Strait (Fig. 3(c)) implies that the geostrophic shear over much of the section is weak except near the boundaries, in general agreement with Fig. 2. Perhaps most importantly, (c) (d) Fig. 3. (continued). 136 S. C. Riser et al.

5 Fig. 3(d) shows clearly that the highest values of dissolved oxygen occur at the sea surface; the values at intermediate depths and near the bottom are considerably lower than the values at the sea surface and show no large-scale indication of the presence of recently formed intermediate or deep water, despite the speculation by some investigators that this region is a likely site of wintertime dense water formation in the Japan Sea. In a similar fashion, the silicate data generally decease monotonically from the sea surface to the bottom and provide no indication (i.e., low values of silicate in the (e) Fig. 3. (continued). (a) Fig. 4. (a) Section of potential temperature for the Northwest Boundary section from data collected on the 1995 US-Russian Akademik Lavrentyev expedition. Station numbers across the top of the figure correspond to station positions shown in Fig. 1. (b) As for (a), but for salinity. (c) As for (a), but for potential density. (d) As for (a), but for dissolved oxygen. (e) As for (a), but for silicate. Circulation and Mixing of Water Masses of Tatar Strait and the Northwestern Boundary Region of the Japan Sea 137

6 deeper waters) of recent deep ventilation in this region, although the data were collected only 1 2 months after the disappearance of sea ice that was formed during the winter of (see Fig. 1 for an estimate of the ice coverage). The maximum distance between stations is approximately 30 kilometers for this section, and thus it is possible that some features of the circulation are not resolved in the data shown in Fig. 1; however, the large-scale properties of the flow should be well-defined by these data. 3.2 The northwestern boundary region Stations on this section were occupied in order to examine the characteristics of shallow and deep boundary currents along the Russian coast. As with the Tatar Strait section, the potential temperature (Fig. 4(a)) below depths of a few hundred meters is low enough that the density of seawater is primarily a function of salinity. In the upper ocean, there is an east-west salinity gradient, with lower salinities in the west near the Russian coast (Fig. 4(b)); these (b) (c) Fig. 4. (continued). 138 S. C. Riser et al.

7 low salinities correspond to low values of σ θ (Fig. 4(c)), implying a southward gestrophic shear near the coast. As with the Tatar Strait section, the highest values of dissolved oxygen (Fig. 4(d)) are at the sea surface, and there is no large-scale indication of the presence of anomalously high dissolved oxygen in the subsurface waters. Similarly, the relatively high subsurface values of dissolved silicate at all stations (Fig. 4(e)) offer no evidence of recent dense water formation. 4. Geostrophic Velocities and Transports Geostrophic velocity across Tatar Strait (Fig. 5), computed relative to the seafloor, indicates northward flow along Sakhalin Island, with speeds in excess of 30 cm/sec relative to the bottom. The geostrophic volume transport between the surface and the bottom associated with this flow is approximately m 3 /sec. This flow would appear to be consistent with the northward flow shown schematically by Yurasov and Yarichin (1991) in Fig. 2; however, there is (d) (e) Fig. 4. (continued). Circulation and Mixing of Water Masses of Tatar Strait and the Northwestern Boundary Region of the Japan Sea 139

8 Fig. 5. Geostrophic velocity relative to the seafloor for the Tatar Strait section. Solid contours denote northward flow. Fig. 6. Geostrophic velocity relative to the seafloor for the Northwest Boundary section. Solid contours denote flow to the north, and dashed contours denote flow to the south. no corresponding southward flow along the Russian coast present along the western boundary of this section. The geostrophic velocity through the Northwestern Boundary section (Fig. 6) is southward along the Russian coast in the western portion of the section, with geostrophic velocities exceeding 30 cm/sec relative to the bottom and a geostrophic transport of about m 3 /sec. This southward flow along the Russian coast has previously been called the Liman Current (Uda, 1934) and both the Primorskiy Current and Shrenk Current (Yurasov and Yarichin, 1991). However, it appears that the flow regime along the Russian coast as manifested in these two sections is somewhat more 140 S. C. Riser et al.

9 complicated than the scenario depicted schematically by Yurasov and Yarichin in Fig. 2. That the Tatar Strait section shows no southward geostrophic velocity along the Russian coast suggests the possibility of a strong westward component of flow along the Tatar Strait section, which might serve to weaken the Liman Current at the western boundary. In fact, such a scenario is depicted in the schematic circulation from Yurasov and Yarichin shown in Fig. 2. Near 45 N, a westward flow (essentially the northwest recirculation of the northern branch of the Tsushima Current) forms a boundary between the two northernmost gyres of the Japan Sea, and there must be a stagnation point of sorts where this westward flow meets the Russian coast. In the schematic diagram shown in Fig. 2, this point is near 45 N on the Russian coast, but the position of this point is probably seasonally dependent, and if this point were a few degrees farther north then the flow would be approximately consistent with the observed geostrophic velocities shown in Figs. 5 and 6. At any rate, it seems that the Liman Current is probably composed of strong transients driven by seasonal effects and mesoscale eddies, so that the schematic depiction of the flow shown in Fig. 2 is likely to be severely oversimplified. 5. Water Masses and Tracers 5.1 Temperature and salinity The CTD-derived θ/s relation for both sections (Fig. 7) clearly shows the presence of low salinity water along the Siberian coast. The upper water along the coast, especially at stations and , is nearly.7 PSU less saline than water at similar values of S in the more eastern portions of both sections. The origin of this relatively fresh water surely lies to the north of both of the sections and must be related to the freshwater input from rivers in the area, as well as the melting of sea ice in the region. It is the southern section, however, that shows the lower surface salinities near the western boundary. There is evidence of some high salinity intrusions in the upper waters of the more eastern stations of the western boundary section, perhaps related to the northwestern recirculation of the Tsushima Current after it leaves the Japanese coast. For stations farther from the western boundary, the θ/s values for the two sections are nearly coincident. Overall, the degree to which the θ/s curves cross isopleths of σ θ at large angles in Fig. 7 gives the impression that at the largest vertical scales diapycnal mixing processes in the water column predominate over isopycnal mixing. For σ θ > 27.2 (depths below about 150 m), the θ/s curves for all stations coalesce, indicating the degree to which the deeper waters in the region are homogenized. An expanded version of the θ/s relation, shown in Fig. 8, shows only weak regional differences in the properties of the deepest waters: below 1 C, the differences between stations are quite small (<.004 PSU) and are very close to the limit of the accuracy of the salinity measurements. The degree to which the subsurface waters of the Japan Sea are homogenized in salinity can be better quantified by examining Figs. 9(a) and (b); here, the CTD-derived salinity differences have been computed between stations in the Tatar Strait and Northwest Boundary sections and Station Fig. 7. θ/s diagram for the Tatar Strait and Northwest Boundary sections, using CTD data. Numbers on the curves denote station numbers. Data are plotted at the positions of the sample bottles, with contours of σ θ superimposed. Circulation and Mixing of Water Masses of Tatar Strait and the Northwestern Boundary Region of the Japan Sea 141

10 Fig. 8. Expanded θ/s diagram for the deep water of the Japan Sea, using CTD data plotted at the positions of the sample bottles. For comparison, data from Station 184 are also shown. The contour σ θ = 27.3 is superimposed. (a) Fig. 9. (a) The salinity difference S between the value of salinity at Station 184 and individual stations ( ) on the Northwest Boundary section, plotted as a function of σ θ. Approximate depths of the surfaces are shown. (b) As for (a), but for the difference between CTD 173 on the Tatar Strait section and individual stations ( ) on the Northwest Boundary section. 184, located some 500 km to the south of the northwestern boundary section in the central Japan Sea, and Station 173 (see Fig. 1 for the station positions). At depths greater than about 350 m, the difference in salinity at a given σ θ between all stations from the northern regions of the Japan Sea and the central region (i.e., Station 184) is less than.005 PSU, a very small difference that is only slightly above the calibration accuracy of the CTD equipment used to collect the data. 142 S. C. Riser et al.

11 (b) Fig. 9. (continued). Fig. 10. Dissolved oxygen for the Tatar Strait and Northwest Boundary sections as a function of depth. For comparison, data from Station 184 are also shown. At the deepest level where the comparisons were done (σ θ > ), there is a suggestion that the southern portions of the Japan Sea, near Station 184, are slightly more saline than the waters farther north; however, the differences are quite small and further inferences regarding regional differences cannot be made. The subsurface salinities from the Tatar Strait section differ from the salinities from the Northwestern Boundary section by less than.002 PSU, or less than the estimated accuracy in the measurement itself. The implication from these comparisons is that subsurface waters of the Circulation and Mixing of Water Masses of Tatar Strait and the Northwestern Boundary Region of the Japan Sea 143

12 Japan Sea (at least in the regions where data exist) are so homogenized that it is difficult to discern quantitative differences in the water mass properties from region to region in the Sea based on θ/s properties alone, even using the most up-to-date techniques; as a result, an examination of other water mass properties is essential. 5.2 Dissolved oxygen The relatively high values of dissolved oxygen at depth everywhere in the Japan Sea compared to the North Pacific are indicative of enhanced convection and ventilation in the Japan Sea. Profiles of dissolved oxygen measured from Niskin bottle samples (Fig. 10) show that at most stations Fig. 11. Dissolved oxygen for the Tatar Strait and Northwest Boundary sections as a function of σ θ. For comparison, data from Station 184 are also shown. Fig. 12. Per cent saturation of dissolved oxygen as a function of depth for the Tatar Strait and Northwest Boundary sections, computed using the algorithm of Weiss (1970). For comparison, data from Station 184 are also shown. 144 S. C. Riser et al.

13 there is no mid-depth minimum in dissolved oxygen; instead the minimum value occurs at the seafloor. This appears to be the case not just for the northern areas of the Japan Sea, but for areas to the south as well, as can be seen from the dissolved oxygen profile from Station 184 farther to the south in the Japan Sea. A number of investigators (Gamo and Horibe, 1983; Gamo et al., 1986; Chen et al., 1995) have commented previously on the near nonexistence of the middepth O 2 minimum in the Japan Sea and the apparent longterm transient nature of this minimum. The measurements reported here are consistent with those studies. At first glance, it appears that there is a large difference in dissolved Fig. 13. Dissolved silicate for the Tatar Strait and Northwest Boundary sections as a function of depth. For comparison, data from Station 184 are also shown. Fig. 14. Dissolved silicate for the Tatar Strait and Northwest Boundary sections as a function of σ θ. For comparison, data from Station 184 are also shown. Circulation and Mixing of Water Masses of Tatar Strait and the Northwestern Boundary Region of the Japan Sea 145

14 oxygen between Station 184 and all of the stations farther north: at all depths between 200 m and 1700 m, dissolved oxygen values at Station 184 are approximately ml/l higher than the values measured in the Tatar Strait and Northwest Boundary sections, even though the data were collected less than one week apart in time. These discrepancies are especially large in the m depth range. However, when viewed as a function of potential density (Fig. 11), it is clear that nearly all of this north/south difference is due to the deepening of isopycnal surfaces between the northern and more southern portions of the Japan Sea. While it does appear that on deep isopycnal surfaces the more southern areas of the Sea have slightly higher oxygen at depth, the differences are small and possibly transitory. A similar result obtains for the O 2 saturation as a function of σ θ (Fig. 12), which suggests that only isopycnal surfaces with σ θ < (depths above about 150 m) or so were directly ventilated during the winter months of Dissolved silicate The distributions of dissolved silicate with depth and potential density (Figs. 13 and 14) show important analogues with the dissolved oxygen measurements. As a function of depth, the values of silicate increase monotonically from the surface to the bottom, with the Tatar Strait values somewhat higher below 800 m than those in the Northwest Boundary Section. The silicate values at Station 184 are lower than those farther north, especially in the depth range m, suggesting more recent contact with the sea surface. This tendency agrees with the result for oxygen; when viewed on a potential density surface (Fig. 14) it is again clear that the regional differences in the distributions are small for σ θ > Fig. 15. History of the equilibrium concentrations of CFC-11 and CFC-12 for surface waters of the Japan Sea, computed using the solubility relations of Warner and Weiss (1985) and the atmospheric history given in Warner et al. (1996). The temperature and salinity of the surface water have been assumed to be 0.1 C and 34.1 PSU. Also shown is the ratio of CFC-11 to CFC-12 as a function of time Fig. 16. CFC-11 for the Tatar Strait and Northwest Boundary sections as a function of depth. For comparison, data from Station 184 are also shown. 146 S. C. Riser et al.

15 5.4 Chlorofluorcarbons The chlorofluorocarbons CFC-11 and CFC-12 enter the ocean via gas exchange processes at the ocean-atmosphere interface. Using atmospheric models and a knowledge of the industrial production and release of these substances, it is possible to compute their atmospheric concentrations as functions of time dating from the 1930s. The solubility of these CFCs in seawater has been given by Warner and Weiss (1985), and since it is known that the equilibration time between CFCs in the ocean and the atmosphere is relatively short, on the order of 1 month, it is possible to estimate the equilibrium concentrations of CFCs in the upper ocean using the solubility relations. An estimate of the wintertime history of the CFC concentration at the surface of the Japan Sea, shown in Fig. 15, has been made assuming that the surface seawater concentration of CFC is in equilibrium with the atmosphere and that the surface water has the temperature and salinity typical of wintertime conditions (the atmospheric history of CFC has been discussed in more detail in Warner et al. (1996)). For the northern part of the Japan Sea, these sea surface values have been taken to be 0.1 C and 34.1 PSU, typical values for the bottom water of most regions of the Sea, which, based on the oxygen and CFC distributions, must have originated at the sea surface in winter. At any rate, the estimate of the surface water CFC history shown in Fig. 15 is not a very strong function of the choice of these wintertime parameters. The distribution of CFC-11 as a function of depth (Fig. 16) shows a general decrease from the surface to the bottom from observations made on the two sections. The upper 200 m of the water column is nearly well-mixed, and between 1500 m and the bottom there is little variation in the CFC- 11 concentration for most stations. Station 184, in the more southern regions of the Japan Sea, shows (as does oxygen) considerably higher concentrations of CFC-11 between 200 and 1700 m than the values in the northern Sea, but as a function of σ θ (Fig. 17) the differences between the north and south are minimal. CFC-11 profiles collected during the Lavrentyev expedition from other sites in the Japan Sea (Fig. 18; see Fig. 1 for station positions) show clearly the largescale homogeneity of CFC in the deep water of the Japan Sea. The CFC-12 distribution (not shown) shows features similar to those for CFC-11, although at lower concentrations. As is evident from Fig. 15, the ratio of the concentrations of CFC-11 and CFC-12 has changed over time, and as a result this ratio can be useful as a measure of the parcel s age, or the time since it was at the sea surface. There are several caveats associated with such calculations. First, the fact that the subsurface concentrations of CFC are generally less than the sea surface equilibrium values implies that the subsurface water is composed of surface water that has been mixed and diluted by other water having CFC concentrations well below the contemporary sea surface equilibrium values. As a result, the CFC age estimates pertain only to the CFC-bearing component of the water parcel. Second, because an individual seawater parcel may contain components with many different ages, care must be taken in interpreting the ratio-derived age since in general age does not mix in a linear fashion. Third, due to the nature of the ratio calculation, the mixing of CFC-bearing water with waters having Fig. 17. CFC-11 for the Tatar Strait and Northwest Boundary sections as a function of σ θ. For comparison, data from Station 184 are also shown. Circulation and Mixing of Water Masses of Tatar Strait and the Northwestern Boundary Region of the Japan Sea 147

16 Fig. 18. CFC-11 as a function of depth for stations distributed over the Japan Sea. Station locations are shown in Fig. 1. Fig. 19. pcfc-11 age for the Tatar Strait and Northwest Boundary sections as a function of depth. For comparison, data from Station 184 are also shown. little or no CFC can yield unpredictable results. Warner et al. (1996) have dealt with these potential problems for the North Pacific and have shown that, in most situations, quite reasonable interpretations of the CFC age are possible under a variety of concentrations and mixing conditions. Considering the caveats discussed above, at least two estimates of the CFC age are available, as discussed in detail by Warner et al. (1996). As can be seen in Fig. 15, the ratio of the CFC-11 and CFC-12 concentrations can be used from the early 1940s to the 1970s to infer the CFC age of a seawater sample. This method (the ratio age ) makes the assumption that the CFC-bearing component of the sample 148 S. C. Riser et al.

17 Fig. 20. pcfc-11 age for the Tatar Strait and Northwest Boundary sections as a function of σ θ. For comparison, data from Station 184 are also shown. preserves its sea surface CFC-ratio (determined from the equilibrium values) after it leaves the sea surface and mixes with other water. After the late 1970s, however, the ratio became nearly constant and is less useful for age estimation. A second estimate of age can be constructed from examining the pcfc age, by estimating the partial pressure of the CFC- 11 component as pcfc-11 = [CFC-11]/F, where [CFC-11] is the measured concentration and F is the solubility of CFC- 11 as a function of temperature and salinity given by Warner and Weiss (1985). By computing the partial pressure of the CFC-11 bearing component in this manner, and then comparing this partial pressure to the history of the atmospheric CFC-11 concentration, the age (the pcfc-11 age ) can be estimated. As noted, a weakness of the ratio age determination is that the atmospheric ratio, and thus the equilibrium oceanic ratio, has not changed substantially since the 1970s. As a result, it is difficult to use the ratio age to date samples ventilated in the past 20 years; since it appears that much of the Japan Sea above 1000 m has been ventilated since the early 1970s, we have chosen to employ the pcfc-11 age in this analysis. Values of the pcfc-11 age for samples from the Tatar Strait and northwestern boundary regions of the Japan Sea, as functions of depth and σ θ (Figs. 19 and 20), show that only very small regional differences exist in the age of the waters on various density surfaces of the Japan Sea, in agreement with the distributions of other properties discussed previously. It is clear from the CFC age distribution that some component of the waters at all depths of the Japan Sea has been ventilated within the past 40 years or so, with the waters at depths above 300 m having been ventilated during the previous few winters. On the other hand, there is no evidence in the samples collected during the 1995 cruise of recent ventilation (within the past years) of the deeper waters of the Japan Sea, in either the Tatar Strait and Northwest Boundary regions or in more southern areas of the Sea (see Fig. 18). This strongly suggests that, based on the CFC distribution, it can be concluded that during the winter of no deep water was formed in the Tatar Strait and northwestern boundary regions of the Japan Sea, and furthermore, that over much of the Japan Sea, including the northern boundary region, ventilation of the water below 2000 m has been relatively weak since at least the 1960s. 6. Mixing Properties in the Northern Japan Sea The remarkable homogeneity of deep water mass properties of the Japan Sea on density surfaces suggests that it is worthwhile to examine the nature of mixing in more detail. The evolution of dissolved oxygen following a seawater parcel can be written do dt = κ H 2 O + κ V 2 O z 2 + R O = M + R O, 1 () where t is the time, 2 is the horizontal Laplacian operator, κ H and κ V are horizontal and vertical diffusivities, and z is the vertical coordinate. R O is the biological consumption rate of oxygen, and M is the sum of the horizontal and Circulation and Mixing of Water Masses of Tatar Strait and the Northwestern Boundary Region of the Japan Sea 149

18 vertical (or isopycnal and diapycnal) mixing terms. The operator d/dt denotes the change following a marked seawater parcel. If time τ is measured relative to the time a seawater parcel left the sea surface, and the change in dissolved oxygen is estimated as AOU, the apparent oxygen utilization (i.e., the difference between the in situ oxygen concentration and the wintertime saturation value at the sea surface), then Eq. (1) becomes AOU τ This can be rearranged to yield M + R O. ( 2) λ = M R O = AOU R O τ 1. 3 ( ) Here λ denotes the ratio of mixing to biological consumption of oxygen. The rate of oxygen consumption is generally a poorly-known parameter; however, recently Chen et al. (1996) have made an experimental estimate of R O with relatively small error bounds, for both the upper and deep layers of the Japan Sea, finding that R O =.12 ±.007 ml/l/ yr at depths of m, and R O =.026 ±.001 ml/l/yr below 2000 m. As Chen et al. point out, the isolated nature of the deep waters of the Japan Sea compared to deep waters elsewhere in the world ocean make the estimation of R O somewhat more straightforward than elsewhere in the world ocean, with correspondingly higher confidence in the quality of the estimates of R O than elsewhere. The shallower value is generally similar to upper ocean values from other regions of the world ocean, as summarized by Martin et al. (1987), while the deep value of R O for the Japan Sea is somewhat higher. Estimates of R O at depths between those given by Chen et al. have been made by fitting the Chen values to a function of the form R O = R 1 exp(γz)+ R 2 (where γ, R 1, and R 2 are constants), a slight modification to the suggestion of Wyrtki (1962). By using this parameterization for R O, it is possible to estimate λ using Eq. (3) and the data shown in Figs. 11 and 19. The results for each station, shown in Fig. 21, suggest that in general λ takes on values between approximately 1.5 and 3 for the upper ocean, and values in a relatively narrow range between 1 and 1.5 below a depth of about 500 m. Thus, in the upper ocean it would appear that the dissolved oxygen concentration changes preferentially by mixing, while in the deeper portions of the Japan Sea changes in the dissolved oxygen concentration following a parcel occur more due to biological consumption. This would appear to be consistent with the CFC age estimates shown in Fig. 19, since in the upper ocean mixing in winter must be relatively vigorous, resulting in ventilation to depths of several hundred meters each year; as a result, a seawater parcel must have its dissolved oxygen value reset to the surface saturation value relatively frequently. In the deeper portions of the Japan Sea, where, based on the CFC age estimates, ventilation has not occurred in several decades and mixing is clearly less vigorous, the dissolved oxygen concentration at a parcel changes more by steady biological consumption over a Fig. 21. The ratio λ = M/R O for the Tatar Strait and Northwest Boundary sections as a function of depth computed as described in the text. For comparison, data from Station 184 are also shown. 150 S. C. Riser et al.

19 relatively longer period of time. In estimating λ in Eq. (3), it is clear that the resulting values represent the average value of λ between the sea surface and any depth. For a parcel at depth z, the value of R O (z) is used in estimating λ. However, in actuality the value of R O experienced by the parcel varies along its path as it mixes and changes depth. This is one likely cause of the relatively large variation in the estimated values of λ in the upper ocean; parcels in this region of the water column probably experience large and frequent changes in depth, and a single value of R O cannot adequately parameterize the consumption of oxygen following these parcels. For the deeper water, however, it can be hypothesized that parcels sank relatively rapidly during some ventilation event and then remained at the observed level of the column over a long period of time, so that the value of R O chosen based on the observed depth is truly representative of the degree of biological consumption experienced by the parcel over most of its lifetime. Since in general there are no sources or sinks for CFC in the ocean below the sea surface, it is not possible to construct a model for CFC evolution following a particle that is analogous to Eq. (1). However, it is possible to use the CFC and dissolved oxygen observations together in a simple box model to attempt to discern the recent history of ventilation in the Japan Sea. Consider a simple 2-layer ocean, with a lower layer that is sometimes in contact with the atmosphere. Suppose that properties of the shallow layer such as dissolved oxygen and CFC are known at some specific times, and the history of some of the properties of the deeper layer (dissolved oxygen in this case) is also known. It is assumed that mass in the lower layer is conserved, so that fluid entering via ventilation must be accompanied by the discharge of a similar volume of fluid into the upper layer, with this discharge perhaps having different properties than the ventilated fluid. According to this simple scenario, the evolution of a tracer φ in the lower layer as a function of time is given by ( ) d ( V O φ )= T V ( φ V φ )+ V O R φ + κ A Oφ dt z. ( 4) Here V O is the volume of the lower layer, T V is the transport of ventilated water into the lower layer, φ V is the concentration of the tracer in the ventilated fluid, R φ is the production of φ by nonconservative processes, κ is the vertical eddy diffusivity, and A O is the area of the upper surface of the deep layer. The first term on the right represents the advective transport of tracer in the ventilated fluid into the deep layer; the second is the production or decay of tracer by nonconservative processes; and the third is the diffusion of tracer into lower layer from the upper layer. If the volume of the deep layer is taken to be constant, then Eq. (4) becomes dφ dt = αφ ( V φ )+ R φ + κ φ H z, ( 5) where α = T V /V O is the fraction of water in the deep layer ventilated in a unit time, and H is the scale depth of the deep layer, given by A O /V O. Fig. 22. Dissolved oxygen and the vertical gradient of dissolved oxygen as a function of time at a depth of 2000 m averaged over the Japan Sea, from Riser and Warner (1999). The numbers in parentheses denote the number of stations that went into the estimate. Lines connecting the points are a cubic spline fit to the observations. Error bars denote the standard error of the estimate. The vertical gradient was computed using Eq. (9). The solid line denotes the value of R O, the dissolved oxygen consumption rate, from Chen et al. (1996). Circulation and Mixing of Water Masses of Tatar Strait and the Northwestern Boundary Region of the Japan Sea 151

20 Based on Eq. (5), it is possible to write evolution equations for dissolved oxygen and CFC-11 of the form do dt = α( O V O)+ R O + κ O H z ( 6) for dissolved oxygen (where R O is the oxygen consumption rate, as in Eq. (1)), and dc dt = α( C V C)+ κ C H z ( 7) for CFC-11. Note that R C is zero, as in general there are no subsurface sources or sinks of CFCs. Using data from a number of historical archives, it is clear (Fig. 22) that dissolved oxygen at a depth of 2000 m in the Japan Sea has been decreasing at a nearly steady rate since the early 1930s. Gamo et al. (1986) was the first to report this well-known change in deep properties of the Japan Sea, and since then this trend has been confirmed by a number of other investigators. From Fig. 22, it would appear that, for dissolved oxygen, do dt R O, ( 8) where R O is the rate of consumption of dissolved oxygen given by Chen et al. (1996). Although this balance is only approximate, it would appear to be accurate to within the errors inherent in estimating the dissolved oxygen concentration, its time derivative, and R O as functions of time. (There is a hint in Fig. 22 that perhaps the rate of dissolved oxygen decrease at 2000 m was slightly less than R O from the 1930s through about 1950, and equal or greater than R O after this time, but this is only a qualitative impression.) This approximate balance implies that, in terms of Eq. (5), the advection of ventilated water into the deep layer and the diffusion of oxygen from shallower layer are both small compared to consumption. The vertical gradient of dissolved oxygen at 2000 m can be estimated as do dz O ( 1500 m ) O 2000 m 1000 m ( ) ( 9) as shown in Fig. 22. Taking H = 900 m (the volume of the Japan Sea below 2000 m divided by the area of the Japan Sea at 2000 m, as determined from the ETOPO5 bathymetric dataset), and κ = 1 cm 2 /sec (probably an upper limit on its value), the diffusion term in Eq. (6) is estimated to be approximately ml/l/yr, a value more than an order of magnitude smaller than R O, which is seemingly consistent with Eq. (8). Based on the conditions in 1995 when the Lavrentyev data were collected, we estimate that O V O ~ 3 ml/l at 2000 m, so that for consistency with Eq. (8), we require that α should be somewhat less than.01 ml/l/yr on the average between the 1930s and the present; this upper limit on α would appear to be consistent with the deep water CFC ages shown in Fig. 19. Since there is no long-term time series of CFC-11 measurements in the deep water of the Japan Sea (few published data exist except for the 1988 profile reported by Tsunogai et al. (1993)), an analysis of Eq. (7) must proceed differently than for Eq. (6); instead of examining the changes in CFC-11 in the deep water over time, we are forced to make inferences about the past history using the contemporary observations and assumptions about conditions in the past. To do this, we integrate Eq. (7) iteratively in time to find C i C i 1 + α( C Vi C i 1 )+ κ C H z i 1 t, 10 ( ) where t is the time step, taken to be 1 year. The subscript i denotes the value after the i-th time step. In order to evaluate Eq. (10), we have taken C = 0 in 1934 and assumed that C Vi was equal to the sea surface saturation CFC-11 values for each year as shown in Fig. 15. The CFC-11 concentrations in the deep layer of the box model have been studied using Eq. (10), for several choices of the ventilation and diffusion parameters, as shown in Fig. 23. With relatively weak but continuous ventilation of the deep water since the 1930s (α = 0.002/yr), and CFC-11 vertical diffusion defined analogous to Eq. (9) with κ = 0, the predicted value of CFC-11 in the deep water of the Japan Sea in 1995 is about 20% less than the values observed in both 1988 and in 1995 during the Lavrentyev expedition. Increasing κ to 0.1 cm 2 /sec with α = 0.002/yr yields estimates for CFC-11 that are only slightly larger than the results for κ = 0. Note that, since the vertical gradient of CFC-11 is not known before 1995, the value of the gradient observed in 1995 as defined by Eq. (9) is used for estimating the diffusive term in all years in this calculation. Because of the lower subsurface values of CFC-11 prior to 1995, this is likely to overestimate the actual value of the vertical gradient of CFC-11 somewhat. An evaluation of Eq. (10) with α = 0.002/yr and κ = 1 cm 2 /sec, again estimating the vertical gradient in Eq. (10) from the value measured in 1995, greatly overestimates both the 1998 and 1995 observations. It thus appears plausible that both the 1988 and 1995 CFC- 11 observations could be fit reasonably well with a choice of α = 0.002/yr and κ approximately 0.1 cm 2 /sec. A more accurate estimate of the vertical gradient for years before 1995 (probably smaller than the 1995 value) would simply imply a that a larger value of κ would be necessary in order to be equivalent to the α = 0.002/yr curves shown in Fig. 23. For 152 S. C. Riser et al.

21 Fig. 23. The history of CFC-11 in the deep layer of the box model of the Japan Sea, computed using Eq. (10) for various choices of α (yr 1 ) and κ (cm 2 /sec), as described in the text. The symbol * denotes the estimated value of CFC-11 at 2000 m in 1977, extrapolated from the data of Tsunogai et al. (1993), and the symbol denotes the typical value at 2000 m from the 1995 Lavrentyev expedition. The error bar denotes the estimated sum of measurement and extrapolation errors in the 1988 sample and the standard error of the mean estimate from 9 samples collected during the 1995 Lavrentyev expedition. higher values of α, the model solutions tend to greatly overestimate the deep CFC-11 concentration, with the apparent errors increasing with increasing values of α. For α = 0.01/yr, κ = 0, the results greatly overestimate both the 1988 and 1995 concentrations, with even larger discrepancies for α = 0.01/yr and κ = 1 cm 2 /sec. The box model thus shows that, as an average from the 1930s to the present, only about 0.2% of the deep water of the Japan Sea has been directly ventilated per year, yielding an estimate for the residence time of the water below 2000 m in the Japan Sea of about 500 years. The pcfc-derived ages of the deep water shown in Figs. 19 and 20 imply that the the CFC-bearing component of the water below 2000 m in the Japan Sea contains mostly water ventilated through the 1960s, and the model solutions shown in Fig. 23 indicate that constant, relatively weak ventilation from the 1930s to the present time can plausibly explain the 1988 and 1995 observations. But the rather striking constancy of the pcfc-11 age in the water below 2000 m (values near 35 years) suggests the need to examine a second, time-dependent ventilation scenario, with ventilation set to a relatively high value prior to the mid-1960s and a lower value thereafter. In this second scenario, the ventilation has been set to a relatively large value (α = 0.01/yr) through the mid-1960s, and then set to zero in the succeeding years. With κ = 1 cm 2 /sec, the model distribution shows a small cusp in the distribution of CFC-11 near 1965, when the direct ventilation was assumed to cease. Prior to this time, the distribution was mainly controlled by ventilation, due to the relatively high value of α, while after this time the distribution was controlled only by vertical diffusion. The results from this model scenario agree well with both the 1988 and 1995 CFC-11 observations and imply a deepwater residence time of only 100 years before the mid- 1960s. The average rate of deep water ventilation in this second case is about α = 0.005/yr between the 1930s and 1995, or a deep-water residence time of about 200 years, somewhat smaller than the constant ventilation case but with α still small enough to be consistent with Eq. (8). The results from the second scenario are perhaps more consistent with the deep-water residence times estimated by Tsunogai et al. (1993) and Gamo and Horibe (1983), using other tracers, than the case with α = 0.002/yr over the entire period. Using the results of the two most likely choices for α and κ, a history of the pcfc-11 age for the Japan Sea at 2000 m has been constructed, as shown in Fig. 24. Here the model concentrations shown in Fig. 23 have simply been converted to pcfc-11 ages by using the CFC history shown in Fig. 15, the age algorithm given by Warner et al. (1996), and CFC- 11 solubility function provided by Warner and Weiss (1985). The results are displayed as ventilation year versus year, where the ventilation year can be interpreted as the year in which the deep water would appear to have been ventilated based on its pcfc age. As an example, based on the pcfc- 11 age in 1970 the Japan Sea deep water would have appeared to have been ventilated in 1949, for the α =.002/ yr, κ = 0.1 cm 2 /sec case. The pcfc-11 ages derived from the 1988 and 1995 measurements, also shown in Fig. 24, are both closer to the time-dependent ventilation curve (α = 0.01/ yr, κ = 1 cm 2 /sec, with ventilation cessation in 1965) than the constant ventilation case; however, given the paucity of Circulation and Mixing of Water Masses of Tatar Strait and the Northwestern Boundary Region of the Japan Sea 153

22 Fig. 24. The apparent year that the deep waters of the Japan Sea were ventilated, as a function of year, for two possible ventilation scenarios, computed as described in the text. The computation was carried out for years after 1955 only; for earlier times the pcfc- 11 age estimates, and thus the ventilation year estimates, were somewhat unstable due to the very small concentrations of CFC-11 present. The jumps in the curves are due to the fact that time was sampled in one-year bins. Symbols superimposed on the curves denote the ventilation years estimated from the pcfc-11 ages in the 1988 (Tsunogai et al., 1993) and 1995 observations. observations and uncertainties in the model, it is probably impossible to discriminate between the relative likelihood of these two ventilation scenarios. Yet the history of dissolved oxygen in the deep Japan Sea (Fig. 22) implies that at some point earlier in this century the deep ventilation was much stronger than it is presently: the values of oxygen near 6 ml/l in the early 1930s provide clear evidence that some type of deep ventilation was active over a substantial portion of the Japan Sea during, or just prior to, that time (or, that biological consumption of dissolved oxygen was considerably less than in recent times). For example, by integrating Eq. (6) in time (yielding a result analogous to Eq. (10)), it is found that in order to increase the present dissolved oxygen concentration in the deep water by 1 ml/l (essentially returning it to its value in the 1930s), it would be necessary for α to be as high as 0.02/yr for the next 50 years. Such a value is 4 10 times higher than the time-averaged values of α in the best solutions shown in Fig. 23. Thus, it seems unlikely that deep ventilation can be adequately characterized by α = 0.002/yr throughout this century, and for this reason perhaps the α(t) scenario depicted in Figs. 23 and 24 is a more accurate depiction of the history of deep ventilation over the Japan Sea during the past 7 decades. 7. Discussion The observations reported in this paper can be used to argue that it is unlikely that the deep waters of the Tatar Strait or northwestern boundary region of the Japan Sea were directly ventilated by the atmosphere during the winter of , a conclusion based largely on the deep water distribution of dissolved oxygen, CFC-11, and CFC-12 in this area; none of the indicators that might be used to provide evidence of recent ventilation of the waters below depths of a few hundred meters showed characteristics consistent with recent deep water formation in this region. This conclusion has been drawn from a dataset consisting of observations at 20 stations, with an average spacing between stations of 35 kilometers. Thus, it would seem that the possibility that ventilation did occur in the northern Japan Sea in the winter of but was not captured in the data due to undersampling cannot be dismissed out of hand. It is useful to examine this idea. If V is the volume of the Japan Sea below 2000 m and µ is the fraction of this volume that is ventilated over some time interval τ, then the volume transport T of ventilated water over this time interval is T ~ µv/ τ. The transport T is presumably supplied by some deep flow with dimensions W (the width), H (the thickness), and velocity U, for a transport of WHU. Equating these two expressions for the transport and solving for U, it is found that U µv HW τ. 11 ( ) If we take µ ~ 1% (suggested as an extreme value by the model solution shown in Fig. 23), V ~ km 3 (taken from the ETOPO5 bathymetric data), H ~ 200 m, W ~ 35 km (the typical horizontal and vertical distances between chemical samples in the sections), and τ ~ 3 months (the winter season), then it is found that U ~ 5 cm/sec. A flow of this magnitude could reasonably be present between stations 154 S. C. Riser et al.