An ecosystem model for the North Pacific embedded in a general circulation model

Journal of Marine Systems 25 Ž2000. 159–178 www.elsevier.nlrlocaterjmarsys

An ecosystem model for the North Pacific embedded in a general circulation model Part II: Mechanisms forming seasonal variations of chlorophyll Michio Kawamiya a,) , Michio J. Kishi b, Nobuo Suginohara a a

Center for Climate System Research, UniÕersity of Tokyo, 4-6-1, Komaba, Meguro-ku, Tokyo 153-8904, Japan b Faculty of Fisheries, Hokkaido UniÕersity, 3-1-1, Minato-cho, Hakodate, Hokkaido 041-8611, Japan Received 26 February 1999; accepted 10 January 2000

Abstract Mechanisms forming seasonal variations of the oceanic ecosystem are examined using an ecosystem model embedded in an ocean general circulation model for the North Pacific. The North Pacific can be divided into seven areas based on seasonal variation patterns of chlorophyll. The areal division compares well with that based on observations. For example, a distinct spring bloom can be found in the Bering Sea while the standing stock of phytoplankton is somewhat constant in most of the subpolar gyre throughout the year. It is shown that physical factors determining the modeled seasonal variations are the annual mean vertical flow, vertical mixing, and, less importantly, solar radiation. The annual mean vertical flow can affect the seasonal variation patterns through modifying the relation between the ecosystem and the mixed layer depth ŽMLD.. Various combinations of these factors result in diverse patterns. But it is possible to classify them in terms of the amplitude of the seasonal variation of MLD and the annual mean vertical flow. The results suggest that the difference in physical environments can yield much of the diversity of the observed seasonal variation patterns. q 2000 Elsevier Science B.V. All rights reserved. Keywords: ecosystem model; North Pacific; GCM simulation; chlorophyll; seasonal variation; annual cycles

1. Introduction Temporal variations of the oceanic lower trophic levels of the ecosystem are affected by various factors. Vertical mixing, temperature, light intensity, nutrient composition, and species composition are, )

Corresponding author. Current affiliation: Institut fur ¨ Meereskunde Kiel, Universitat Kiel, Dusternbrooker Weg 20, ¨ D-24105 Kiel, Germany. Tel.: q49-431-597-3973. E-mail address: [email protected] ŽM. Kawamiya..

among others, especially important when we seek the causeŽs. for variations of the oceanic ecosystem regardless of their temporal andror spatial scales. Their effects on the ecosystem are most clearly seen in its seasonal variations. For instance, the vernal blooming, which is a conspicuous feature in some oceanic regions such as the North Atlantic and the Bering Sea, is often associated with changes in strength of vertical mixing Že.g., Sverdrup, 1953; Evans and Parslow, 1985; Fasham, 1995.; the summer maximum of the plankton biomass in the Cali-

0924-7963r00r$ - see front matter q 2000 Elsevier Science B.V. All rights reserved. PII: S 0 9 2 4 - 7 9 6 3 Ž 0 0 . 0 0 0 1 3 - 0

160

M. Kawamiya et al.r Journal of Marine Systems 25 (2000) 159–178

fornia upwelling region has been explained by the fact that the upwelling becomes strongest in that season Že.g., Thomas et al., 1994.; seasonal variations in high nutrient–low chlorophyll ŽHNLC. regions might be strongly controlled by silicate ŽDugdale et al., 1995. or iron ŽMartin and Fitzwater, 1988. concentration. Therefore, understanding of mechanisms forming seasonal variations will lead to a good grasp of essential features of the oceanic ecosystem. Seasonal variations of the oceanic ecosystem show different patterns depending on the geographical position of the ocean. Colebrook Ž1979. demonstrated, by analyzing data in the North Atlantic obtained by the Continuous Plankton Recorder, that timing of the vernal blooming depends on the location. Using chlorophyll data from satellite measurements by a coastal zone color scanner ŽCZCS., Longhurst Ž1995. divided the world ocean into 56 provinces that can be classified into four main domains, and showed typical patterns of seasonal variations for some of the provinces. Banse and English Ž1994. also used CZCS data to demonstrate diverse seasonal variations. Many investigators attribute geographical differences in seasonal variation to physical environments, although some attribute them to differences in chemical andror biological condition. Indeed, Colebrook Ž1979. related timing of the spring bloom to that of the sea surface warming based on in situ data in the North Atlantic. Obata et al. Ž1996., using CZCS data and the data of Levitus Ž1982., also showed that on the global scale the spring bloom is tightly related to the surface warming in the timing. The 56 provinces of Longhurst Ž1995. are based mainly on physical environments. Longhurst Ž1995. adopted the notion that physics determines the overall feature of ecosystem behavior as a premise. However, it has not been thoroughly examined to what extent physical environments can produce the diversity in seasonal variation without invoking complex chemical andror biological factors such as micronutrients and species composition. Fasham et al. Ž1993. and Sarmiento et al. Ž1993. used an ecosystem model embedded in an OGCM to investigate ecosystem dynamics in the North Atlantic. Their main purpose was, however, to identify discrepancies of biomass distributions between model

results and observations and to clarify the causes of the discrepancies. Some studies dealt in detail with the relation between differences in physical environments and seasonal variation patterns of the ecosystem. Evans and Parslow Ž1985. and Fasham Ž1995. demonstrated, using a box model, that dynamics of the mixed layer can explain the difference between regions with and without a large spring bloom. McCreary et al. Ž1996. investigated seasonal variations in the Indian Ocean with an ecosystem model embedded in a 2.5-layer model of the ocean. They showed that seasonal variations in some selected regions are deeply affected by the mixed layer depth ŽMLD.. More recently, Ryabchenko et al. Ž1998. also performed a simulation study for the Indian Ocean using a three-dimensional model and found that the results were basically similar to those of McCreary et al. Ž1996. despite many differences between the models in both physics and biology. In these papers, the focus is on the cause of formation of seasonal variations at some chosen locations, but not on the spatial extent of the region having a certain type of seasonal variation. In this paper, by analyzing the results of the three-dimensional ecosystem model embedded in the OGCM for the North Pacific ŽKawamiya et al., 2000, referred to as KKSa hereinafter., we will identify seasonal variation patterns and their spatial distribution in the model and examine how physical environments regulate seasonal variations of the ecosystem. This paper is organized as follows. In Section 2, the domain is divided into seven areas based on seasonal variation patterns and annual mean abundance of chlorophyll. It will be shown that the pattern and the extent of each area compare well with the observations. The mechanisms forming the contrasts of such patterns in the model are then investigated in Section 3. The factors determining the overall distribution of seasonal variation patterns are discussed in Section 4. Finally, summary and conclusion are presented in Section 5. 2. Areal division All data for the analysis are the monthly averages in the sixth year of the integration; 12 datasets comprise the annual cycle.

M. Kawamiya et al.r Journal of Marine Systems 25 (2000) 159–178

161

as shown later. Data near the southern artificial wall, 238S–118S, are also excluded. The first two principal modes are depicted in Fig. 1 together with their time coefficients. The first mode ŽEOF 1. shows that the amplitude of seasonal variations is higher at middle and high latitudes. It reaches its maximum phase in April. The peaks in the second mode ŽEOF 2. are located where the mixed layer becomes extremely deep in early spring ŽFig. 5 of KKSa., i.e., off Sanriku, Japan Žthe positive peak., and along the latitudinal line of ; 308N Žthe negative peaks.. The information from the EOF analysis is utilized for areal division. Annual mean abundance of chlorophyll averaged over the upper 150 m ŽFig. 2a. is also exploited to define boundaries because it is highly probable that, even if the seasonal variation patterns look alike, they are formed by a different

Fig. 1. Ža. First two principal modes of EOF analysis applied to model chlorophyll data averaged over the upper 150 and Žb. their time coefficients. In the parentheses in Ža. are the percentages by which the total variance is accounted for. Contour intervals are 0.02. In Žb., the solid line represents the time coefficients for the first mode and dashed line for the second mode.

2.1. Procedure for areal diÕision and description of each area To obtain characteristic seasonal variation patterns and their distribution, empirical orthogonal function ŽEOF. analysis is applied to the chlorophyll Žstrictly equivalent to phytoplankton in the model. data averaged over the upper 150 m. In the analysis, data at latitudes higher than 528N are excluded. This is because too many modes are needed only to explain variations at high latitudes and the amplitude of seasonal variations is much larger at high latitudes

Fig. 2. Annual mean Ža. chlorophyll and Žb. nitrate concentration for the control case averaged over the upper 150 and 100 m, respectively. Contour intervals are 0.02 mgrl and 2 mmolrl, respectively.

M. Kawamiya et al.r Journal of Marine Systems 25 (2000) 159–178

162

Fig. 3. Areal division of the domain based on the model results.

mechanism in a eutrophic, chlorophyll-rich region and an oligotrophic, chlorophyll-poor region. In Fig. 2b, nitrate distribution is depicted for later reference, especially in Section 4.4. Based on the EOF modes and the annual mean abundance, we attain areal division as shown in Fig. 3. The areas are denoted by two letters. The origin of the area name and characteristic features of the area are presented in Table 1. Regions of sharp fronts in Fig. 2a, such as the one along ; 158N, are not included in the division but treated as transition zones except the area KS where the front itself includes modal peaks. Most areas contain at least one of the peaks in either the EOF modes or the annual mean abundance. Also, each boundary beTable 1 Features of the areas shown in Fig. 3 Name of the area

Features

BO

Extremely large amplitude of seasonal variation High positive values in EOF 1 Positive peak in EOF 2 High abundance in annual mean Negative peak in EOF 2 Low abundance in annual mean High abundance in annual mean

SP PR CU KS ST EQ

tween the areas corresponds to a front in either the EOF modes or the annual mean abundance. This division is objective based only on the chlorophyll data in the model. There are distinct differences in seasonal variation pattern among the areas as shown in Fig. 4a, which represents the seasonal variations of chlorophyll averaged over the area and the upper 150 m. It is clearly seen that the area Bering and Okhotsk Sea ŽBO. has the maximum amplitude. The area Perturbed Region ŽPR. has the second largest while the areas Subpolar region ŽSP., Coastal Upwelling region ŽCU., and Kurioshio and Kuroshio Extension ŽKS. show moderate seasonal variations. The areas Subtropical region ŽST. and Equatorial region ŽEQ. indicate only small variations. In Fig. 4b and c, seasonal variations of MLD and nitrate are depicted. PR and KS go through large variation of MLD due to the effect of the Kuroshio in the model, reaching their maximum phase in early spring. The impact of MLD variation can be clearly seen in nitrate, which has high concentration when MLD is large. ST is nitrate depleted all through the year Žhere the phrase ‘‘nitrate depleted’’ means that the surface nitrate concentration is of the order of or less than the value of the half saturation constant for nitrate in the model, 0.03 mmolrl.. KS becomes nitrate depleted from summer to fall but is relatively

M. Kawamiya et al.r Journal of Marine Systems 25 (2000) 159–178

163

enriched in nitrate from winter to spring. The other areas never become nitrate depleted because all of them are located in regions with wind-induced upwelling ŽCoastal and Ekman upwelling., which brings up nitrate from under the euphotic zone. The relation between the seasonal variation of the modeled chlorophyll and the factors such as MLD, upwelling, and nitrate concentration will be discussed in detail in Sections 3 and 4. 2.2. Comparison with obserÕations 2.2.1. Comparison with areal diÕision based on obserÕations The ecological areal division for the North Pacific proposed by Longhurst Ž1995. is redrawn in Fig. 5. This figure is based on observations. He used the figure to classify the seasonal variation of chlorophyll detected by CZCS ŽLonghurst, 1995.. It was also used to describe the spatial variation of parameters needed for estimating primary production from CZCS data ŽLonghurst et al., 1995.. Remarkable resemblance can be found between this figure and the present areal division in Fig. 3. BERS in the former corresponds to BO in the latter, PSAGŽW. and PSAGŽE. to SP, KURO to PR, NPSTŽW. to KS, NPSTŽE. and NPTG to ST, and PNEC together with PEQD to EQ. Although each of the model areas, SP, ST, and EQ, contains two areas of Longhurst Ž1995., similarity can be found between the two of each combination ŽLonghurst, 1995.: the shallow halo-

Fig. 4. Modeled seasonal variations of Ža. chlorophyll averaged over the upper 150 m, Žb. MLD and Žc. nitrate averaged over the upper 100 m, as averaged over the area shown in Fig. 3.

Fig. 5. Ecological areal division of the ocean based on observational facts Žredrawn from Longhurst, 1995.. The North Pacific is picked up from the original figure for the world ocean.

164

M. Kawamiya et al.r Journal of Marine Systems 25 (2000) 159–178

cline found in both NPSTŽE. and NPSTŽW. has an impact on the ecosystem there; in both NPSTŽE. and NPTG, the deep subsurface chlorophyll maxima are formed throughout the year at depths deeper than 100 m, thus indicating a feature of oligotrophic regions; wind-induced upwelling occurring in both PNEC and PEQD carries nutrients to enhance biological activities. The agreement between Figs. 3 and 5 may support reality in the modeled seasonal variations. WARM in Longhurst Ž1995., however, does not have a counterpart in the model result. Although the model result does show a tendency that chlorophyll decreases westward, which is consistent with data, the western equatorial region appears as, rather than a distinct area, a front between the equatorial eutrophic region and the subtropical oligotrophic region. The reason for the absence of WARM may be the model’s inability to reproduce the complex structure of current in the equatorial region. For example, the Equatorial Counter Current in the model does not reach the western region unlike in reality. Because nutrient concentration is lower in the western region, this discrepancy can result in overestimation of nutrient transport from the east to the west, thereby obscuring the boundary between the western oligotrophic region and the eastern eutrophic region. 2.3. Comparison with satellite data In Fig. 6, the modeled chlorophyll concentration in each area is compared with that detected by CZCS. The satellite data are also averaged over each area shown in Fig. 3. The satellite observation provides information averaged over the depths of only 10 m Že.g., Gordon et al., 1982.. Therefore, the modeled seasonal variations are replotted using data averaged over the upper 10 m so that they can be directly compared with the CZCS data. The satellite data at high latitudes for October–February are excluded because they are unreliable ŽYoder et al., 1993.. It is very easy to find differences between the two figures, yet many common features are found: Ø BO has the much larger amplitude than any other area;

Fig. 6. Seasonal variation of chlorophyll averaged over the area shown in Fig. 3 for Ža. model data averaged over the surface 10 m and Žb. CZCS data. The CZCS data at high latitudes for October–February are not shown because they contain large errors.

Ø KS reaches its maximum in March and declines to its minimum in late summer or early fall; Ø PR has the larger amplitude than SP, although they are at similar latitudes; Ø PR has the larger amplitude than KS, although they both go through large variation of MLD ŽFig. 4b. compared with that in the other regions; Ø CU reaches its maximum in summer, although the concentration is much lower in the model; Ø ST and EQ experience very small variations. The agreement is satisfying considering the simplicity of the present ecosystem model and the fact that the areal division is totally based on the model chlorophyll data.

M. Kawamiya et al.r Journal of Marine Systems 25 (2000) 159–178

3. Mechanisms forming seasonal variations of each area Having seen that the seasonal variations in the model are not unrealistic, mechanisms forming them are examined with emphasis on the similarities listed in Section 2.3. 3.1. Bering and Okhotsk Sea (BO) The seasonal variation of the vertical chlorophyll profile in BO is shown in Fig. 7a. Here, instead of

Fig. 7. Ža. Modeled seasonal variation of chlorophyll at 568N, 1778E, which is located in BO in Fig. 3. Contour intervals are 0.25 mgrl. Žb. Light dependence of photosynthesis at 568N, 1778E. More specifically, isopleth of w Ir Iopt xexpw1y Ir Iopt x in Eq. Ž2. of KKSa averaged over the day cycle. Contour intervals are 0.05.

165

taking an average over the area, a specific point Ž568N, 1778W. in BO is chosen. This is because the averaging process sometimes obscures characteristic features of interest, especially in the analysis for balance of terms, which will be made below. Even if a mechanism forming seasonal variation is the same, the depth to which strong vertical mixing can reach or the thickness of the euphotic layer can be different depending on the specific point; if we take an average over a certain area, the emerging balance of terms can be different from that at each point. At this chosen point, a distinct chlorophyll maximum appears in May at shallow depths Ž; 10 m.. After its disappearance, the relatively weak subsurface chlorophyll maximum ŽSCM. is formed at the deeper depths, being clearest in June. In winter, concentration becomes lower and independent of depth. The high concentration in May has supporting evidence not only from the CZCS data ŽFig. 6b., but also from in situ data of Smith and Vidal Ž1984. who observed the extremely high chlorophyll concentration Ž; 10 mmolrl. in May at several stations. To answer the question as to how the bloom is excited in the model, the terms in the governing equation of chlorophyll ŽEq. A16 of KKSa. are averaged over a month and plotted in Fig. 8a,b. The balance is shown for April and May because these months correspond to, respectively, immediately before and at the maximum phase of the bloom. In Fig. 8c, the same balance is shown for a point in SP for later comparison. In April, the contribution of vertical mixing is so large that net primary productivity ŽNPP. is balanced by grazing and vertical mixing at the depths where NPP takes the maximum value; vertical mixing is playing a significant role in sweeping chlorophyll away from the layer most suitable for photosynthesis. In contrast, the contribution of vertical mixing is very small in May; the overall balance is held between NPP and grazing. This qualitative difference in term balance between April and May means that the large imbalance between NPP and grazing is produced by surface warming and subsequent weakening of vertical mixing; the imbalance works to increase chlorophyll because it is generated by disappearance of a term acting against NPP. The bloom ceases when grazing catches up with the chlorophyll increase.

166

M. Kawamiya et al.r Journal of Marine Systems 25 (2000) 159–178

The summer SCM is formed mainly by the effect of light. This is made evident in Fig. 7b where the light dependence of photosynthesis Žw IrIopt xexpw1 y IrIopt x in Eq. Ž2. of KKSa. averaged over the day cycle is plotted. Due to photoinhibition, the most favorable light condition occurs at the depth of ; 40 m during summer. This depth is very close to the location of the SCM in Fig. 7a, suggesting that the light condition is most responsible for the formation of the SCM. The small differences in maximum depth between Fig. 7a and b are due to other factors, among which ammonium inhibition is the most important. The summer SCM after blooming is also found in the observation ŽSambrotto and Goering, 1983., although it is hard to judge whether the formation mechanism is the same or not. Note that photoinhibition limits phytoplankton growth only near the surface; at the depth where light condition is favorable, other processes such as grazing limit phytoplankton growth Žcf. Fig. 8b,c.. In Fig. 7a, the extreme value exists only in May while in Figs. 4a and 6a, it is also seen in April in BO. This is because the maximum is reached in April at some other points where the retreat of the mixed layer occurs earlier than at this point Ž568N, 1778W.. The variations in Figs. 4a and 6a result from the integration over the grids each of which takes the maximum value at different time in April or May. The duration of high chlorophyll concentration is not long enough in the model compared with the CZCS data. The reason for this disagreement is not clear. It may be that shallowing of the mixed layer occurs as late as in June at some locations in reality. Another possibility is succession of species caused by the variation of silicate concentration, which is not incorporated in the model; a bloom of dinoflagel-

Fig. 8. Balance among the selected terms in the governing equation of chlorophyll for Ža. April and Žb. May at 568N, 1778E. Žc. As in the above two but for March at 488N, 1558W. The data are averaged over a month. The terms for photosynthesis and respiration are combined as NPP for simplicity. Thick solid line represents NPP, dashed line grazing, dotted line mortality, dash dotted line vertical advection, and thin solid line vertical diffusion. Horizontal scaling is different between the two plots to enlighten the relative contribution of each term.

M. Kawamiya et al.r Journal of Marine Systems 25 (2000) 159–178

167

Weather Station ŽOWS. Papa ŽClemons and Miller, 1984., which is near the chosen point. The reason for the absence of blooming can be seen in Fig. 8c, where the water column term balance at this point is depicted for March. Even in this month when vertical mixing is most vigorous here, the contribution of vertical mixing is small at the maximal NPP depth compared with that in BO ŽFig. 8a,b.. Thus, the grazing term can be balanced with NPP almost all through the year, thereby suppressing the bloom. In turn, inhibition of strong vertical mixing is caused by strong stratification due to the shallow halocline. Fig. 10 shows the modeled and the observed salinity section along 1798W in March. In the model the strong halocline exists in SP at the depth of 40 m. This halocline persists throughout the year ŽFig. 9b. and prevents winter convection caused by the surface cooling. The suppression of mixed layer deepening by the shallow halocline has been also pointed out in many

Fig. 9. Modeled seasonal variation of Ža. chlorophyll, Žb. salinity at 488N, 1558W, which is located in SP in Fig. 3. Contour intervals are 0.1 mgrl for chlorophyll and 0.1 psu for salinity.

lates often takes over that of diatoms after silicate becomes depleted, thereby making the apparent duration of the bloom longer ŽTsunogai and Watanabe, 1983.. 3.2. Subpolar region (SP) The seasonal variation of chlorophyll is shown in Fig. 9a for the point, 488N, 1558W, located in SP. No pronounced blooming can be found here despite that this region is located near BO, where chlorophyll concentration takes the extremely high value in April or May almost everywhere. This uniformity of chlorophyll concentration was also observed at Ocean

Fig. 10. Salinity section along 1798W in March for Ža. the model and Žb. the observation ŽLevitus and Boyer, 1994.. The arrow in Ža. indicates the location of the boundary between SP and BO. Contour intervals are 0.2 psu.

168

M. Kawamiya et al.r Journal of Marine Systems 25 (2000) 159–178

observational studies Že.g., Tabata, 1964., although the observed shallow halocline lies at deeper depths Ž; 100 m, see Fig. 10b.. Obstruction of the bloom by the shallow winter mixed layer in the subarctic Pacific is also suggested in other modeling studies ŽEvans and Parslow, 1985; Fasham, 1995.. The halocline exists all over the area SP, but it suddenly disappears at the boundary between SP and BO. This is clearly seen in Fig. 10a: north of ; 548N where the boundary is located, the halocline loses its intensity. This distinct difference in salinity structure induces the contrast in the seasonal variations of SP and BO. The tendency of the halocline to disappear toward the north is also seen in the observation ŽFig. 10b.. The pronounced SCM in summer seen in Fig. 9a is formed by the same mechanism as in BO. This is easily inferred from the fact that the two areas are both nitrate-rich so that the only process to strongly inhibit phytoplankton growth near the surface is photoinhibition. The SCM is also found at OWS Papa ŽClemons and Miller, 1984., although it is much weaker. 3.3. Perturbed region (PR) Fig. 11a shows the seasonal variation of chlorophyll at 388N, 1578E. As in BO, the conspicuous blooming is found in April, and then the SCM is formed. The high chlorophyll concentration in spring is detected by CZCS ŽFig. 6b. and is also found by Imai et al. Ž1988., who compiled in situ data from 1973 to 1984 taken by the quarterly routine observation around Japan. Resemblance between PR and BO is also retained even if we look into the mechanism shaping the seasonal variations. Examining the term balance in PR Žnot shown. reveals that the same scenario as in BO can be applied: surface warming causes imbalance between NPP and grazing, thus resulting in rapid phytoplankton growth. Fig. 11b shows the seasonal variation of chlorophyll at a location in EQ. This will be referred to later ŽSection 3.7. but is placed here to save some space. Care must be taken even though the modeled seasonal variations look very similar to the satellite observation ŽFig. 6.. The deep mixed layer in March, which is the cause of the pronounced blooming in the model, is formed due to inability of the OGCM to reproduce the separation of the Kuroshio Žsee

Fig. 11. Modeled seasonal variation of chlorophyll at Ža. 388N, 1578E and Žb. 08, 1398W, which are located in PR and EQ in Fig. 3, respectively. Contour intervals are 0.1 mgrl.

KKSa.. In reality, however, the mixed layer in March is also deep in this area ŽFig. 5 of KKSa., although somewhat shallower than that in the model. In spite of this coincidence, real oceanographic conditions in this area are much more complicated: the Kuroshio separates from the coast of Japan further south; interactions among warm core rings detached from the Kuroshio, the Oyashio intrusions, and heat and fresh water exchanges between the atmosphere and the ocean determine the surface density structure in this area Že.g., Talley et al., 1995.. These features cannot be expressed in this coarse resolution model. Therefore, the processes of mixed layer deepening in the model and the real world differ, yet the results are not totally dissimilar with respect to the mixed layer deepening. It is probable that the real blooming

M. Kawamiya et al.r Journal of Marine Systems 25 (2000) 159–178

169

is also excited by this deep mixed layer, although the impact of mixed layer shallowing is much larger in the model. Because the area PR is formed through the inability of the OGCM, we cannot put any confidence in its spatial extent and so forth. Much finer resolution, ; 1r88, is needed to faithfully reproduce the detachment of the Kuroshio ŽHurlburt et al., 1996.. However, it is widely accepted that the physical and biological conditions in this area are distinctly different from the surrounding region ŽTalley et al., 1995; Kasai et al., 1997.. We believe that the area PR would be distinguished even if we make areal division based only on data. The reason for the existence of the summer SCM is also explained by photoinhibition, as in BO and SP.

3.4. Coastal Upwelling region (CU) Chlorophyll concentration shows the modest seasonal variation in CU ŽFig. 12a for the point at 368N, 1258W.. The feature that chlorophyll concentration is high in early summer is found in the CZCS data although concentration is much lower in the model ŽFig. 6.. This feature is indirectly supported by Chelton et al. Ž1982., who showed that in situ data of the zooplankton biomass in this area take their maximum value in summer. Due to the absence of strong vertical mixing and the abundant nutrient carried by the wind-induced upwelling, chlorophyll concentration closely traces the variation of the light condition ŽFig. 12b., which provides the most favorable condition in June. The best condition in June is brought about by the long duration of daytime. Intense light cannot be the reason: light intensity is well above the optimal value Ž Iopt in Eq. Ž2. of KKSa. even in winter because Fig. 12b shows that the light condition is optimal at the subsurface all through the year. Contrary to the model results, the high chlorophyll concentration in summer is often associated with the seasonal variation of upwelling intensity in observational studies Že.g., Mann and Lazier, 1991., although it is still a hypothesis that the upwelling intensity determines the chlorophyll seasonal variation ŽThomas et al., 1994..

Fig. 12. Ža. Modeled seasonal variation of chlorophyll at 368N, 1258W, which is located in CU in Fig. 3. Contour intervals are 0.1 mgrl. Žb. Light dependence of photosynthesis as in Fig. 7b but for the above location.

Chlorophyll concentration in this area is much lower than it is in the CZCS observation. This is perhaps attributable to the parameter choice. It is known that neritic species have more vigorous photosynthetic rates than oceanic ones ŽParsons et al., 1984.. In the model, the photosynthetic rate characteristic for oceanic species is applied to the whole domain and this value must be too small for the coastal upwelling region. Though the concentration is low, the modeled amplitude of the seasonal variation relative to the annual mean abundance is rather large Žamplitudermean ratio is ; 0.4 as calculated from Fig. 4a.. The model result is indicative in demonstrating that the effect of light alone can generate such a large annual variation.

170

M. Kawamiya et al.r Journal of Marine Systems 25 (2000) 159–178

3.5. Kuroshio and Kuroshio Extension region (KS) The point 288N, 1418E is chosen to represent KS and its seasonal variation is plotted in Fig. 13a. From January to March, the chlorophyll maximum is found at the relatively shallow depths Ž0–60 m. with rather high concentration; from April to December, it is formed at the deeper depths Ž; 100 m. with lower concentration. The high concentration in winter or early spring is also seen in the CZCS data ŽFig. 6b. and the observation of Imai et al. Ž1988.. Longhurst Ž1995. suggests that in NPSTŽW. of his province division ŽFig. 5., which corresponds to KS of this study, the relatively high chlorophyll concentration is retained at the near-surface levels Ž0–50 m. in late winter and the deeper SCM is found in other seasons. The model calculation shows a good agreement with his description based on observations. The deep SCM is maintained by a mechanism different from that in BO, etc. Fig. 13b Žtop. shows the light dependence of photosynthesis. It can be seen that the most favorable light condition occurs at the depths far shallower than the SCM. This is because nitrate is depleted where the light condition is optimal. Fig. 13b Žmiddle. shows the nutrient dependence in the governing equation of chlorophyll ŽwNO 3r NO 3 q K NO 4xexp yC NH 4 4 q NH 4r NH 4 3 q K NH 4 4 in Eq. Ž2. of KKSa.. It is clearly seen that in summer nutrient environment is ill-conditioned at the depth of the optimal light condition. The SCM forms at the depths where the product of the light and the nutrient dependence takes the maximum value Žcompare the bottom plot of Fig. 13b with Fig. 13a.. In other words, the SCM settles down at the depths where the best combination is found between the light and the nutrient condition. The ultimate cause of the difference in mechanism for SCM formation between KS and BO is the wind-induced vertical flow: KS lies in the region with Ekman downwelling, while BO in the region with Ekman upwelling. It can also be seen that from January to March, nitrate depletion is relaxed due to deep vertical mixing. Phytoplankton growth is thus enhanced, resulting in high concentration from late winter to early spring. It is noteworthy that the role of vertical mixing is opposite compared with that in the case of BO and others, where vertical mixing acts to reduce

Fig. 13. Ža. Modeled seasonal variation of chlorophyll at 288N, 1418E, which is located in KS in Fig. 3. Contour intervals are 0.02 mgrl. Žb. Dependence of photosynthesis on Žtop. light and Žmiddle. nutrient at the same location. More specifically, isopleth of Žtop. w Ir Iopt xexpw1 y Ir Iopt x and Žmiddle. wNO 3 rNO 3 q K NO 34xexpyC NH 4 4qNH 4 rNH 4 q K NH 44 in Eq. Ž2. of KKSa. These variables are dimensionless. The product of the two is also plotted in the bottom plot.

M. Kawamiya et al.r Journal of Marine Systems 25 (2000) 159–178

171

biological activities through light limitation. On the contrary, here in KS where nitrate is exhausted during warm seasons, mixed layer deepening enhances photosynthesis by transporting nitrate to the surface layer. The deep mixed layer can be found also in the observation ŽFig. 5 of KKSa.. 3.6. Subtropical region (ST) Fig. 14a depicts the seasonal variation of chlorophyll at 268N, 1498W. The deep SCM is formed throughout the year with modest maxima in March and May. In Fig. 14b, the phytoplankton growth term is decomposed as in Fig. 13b. This figure shows that the SCM is retained by the same process as in KS, i.e., the balance between the light and the nutrient condition. The moderate variation is caused by the change in intensity of light and vertical mixing: in winter, vertical mixing can reach ; 100 m depths, thereby stimulating biological activities; in summer when light is strong, chlorophyll concentration is enriched because light can penetrate more deeply so that the two conditions mentioned above can be balanced at the deeper depths, where more abundant nitrate is available for photosynthesis. The maximum in summer is not in June, when light is strongest, but in May. This is because in May, there still remains the effect of vertical mixing bringing nitrate up to the surface layer, so that nutrient concentration at the depths of SCM is slightly higher than in June. The seasonal variation of chlorophyll in the subtropical Pacific has been extensively observed at Station ALOHA Ž238N, 1588W. by the Hawaii Ocean Time-Series ŽHOT. Program since October 1988. Winn et al. Ž1996. provided the integrated chlorophyll abundance averaged over 1988 through 1993, which is shown in Fig. 15 together with the corresponding model result. Near the sea surface ŽFig. 15a., chlorophyll abundance is fairly constant in the model while it takes the minimum value in summer in the observation. The observed summer low value is due to increase in Nrchlorophyll ratio caused by photoadaptation, as explained by KKSa. The seasonal variation near the sea surface cannot be reproduced by the model without the photoadaptation process. On the other hand, the model and the obser-

Fig. 14. Ža. Modeled seasonal variation of chlorophyll at 268N, 1498W, which is located in ST in Fig. 3. Contour intervals are 0.01 mgrl. Žb. Dependence of photosynthesis on light and nutrient as in Fig. 13b but for the above location.

vation show the common feature at the deeper depths ŽFig. 15b. that the integrated chlorophyll increases in

172

M. Kawamiya et al.r Journal of Marine Systems 25 (2000) 159–178

have a systematic seasonal variation. This view is supported by the CZCS data ŽFig. 6b.. SCM is formed also in EQ by the same mechanism as in BO, etc.: the depth of the SCM corresponds to that of the optimal light condition; the equatorial upwelling brings up plenty of nitrate in this area. Small-amplitude short-term variations can be found. They are caused by the variation of vertical mixing, which is, in turn, induced by that of temperature and wind stress. The chlorophyll maximum occurs when vertical mixing is weak at the depth of SCM and the minimum when it is strong; vertical mixing acts to interfere with phytoplankton growth in this nutrient replete area.

4. Discussion 4.1. Synthesis of the experiment

Fig. 15. Chlorophyll abundance integrated over Ža. 0–50 m and Žb. 100–175 m. The model results Žsolid line. are taken from the point 268N, 1498W, which is located in ST in Fig. 3. The observation Ždashed line with error bars. indicates the averaged values over 1988 through 1993 obtained in the HOT Project ŽWinn et al., 1996.. The error bars show the standard deviation.

summer, though the modeled values are out of the range of error bars. Winn et al. Ž1996. state that the increase reflects the change in phytoplankton biomass and that it is brought about by the change in light intensity, which is also the cause for the modeled variation. The modest maximum from summer to fall at ; 70 m depth is formed due to nutrients regenerated within the euphotic zone, while the stronger maximum below is formed by uptaking the nitrate transported from under the euphotic zone by diffusion. 3.7. Equatorial region (EQ) As shown in Fig. 11b, where chlorophyll concentration at 08, 1398W is plotted, EQ does not seem to

In Section 3, the mechanism shaping the seasonal variation pattern in each area was examined by selecting a specific point as representative of the area. A unified view of what determines the distribution of the pattern is, however, still not very clear. This subsection deals with factors that draw the boundaries among the areas. So far we have seen that the factors such as light, nutrient, and vertical mixing are combined in various ways to cause the diversity of seasonal variations. It is possible, however, to comprehensively classify the seasonal variations in terms of physical environments, which also define the extent of the area. Obviously, the seasonal variation of MLD should be included in the physical factors for the classification: the large variation of MLD always leads to the chlorophyll variation as shown in Section 3. To find other physical factors, it is helpful to remember that the ecosystem behavior shows remarkable differences between areas with replete nutrient and those with limited nutrient. Influences of nutrient concentration can be summarized as follows. Ø Abundance of phytoplankton: phytoplankton is more abundant in the eutrophic region than in the oligotrophic region. Ø Response of phytoplankton to changes in MLD: in the eutrophic region mixed layer deepening sup-

M. Kawamiya et al.r Journal of Marine Systems 25 (2000) 159–178

presses photosynthesis because vertical mixing causes chlorophyll dilution, while in the oligotrophic region it stimulates photosynthesis because vertical mixing causes nutrient supply. Ø Mechanism forming SCM: in the eutrophic region SCM is located at the depth of the most favorable light condition, while in the oligotrophic region it is retained at the depth where the light and the nutrient condition are best combined. Ø Vulnerability of phytoplankton to changes in MLD: in the eutrophic region, phytoplankton is subject to strong vertical mixing even for somewhat shallow MLD because SCM is formed at the depth of the optimal light condition, unlike in the oligotrophic region where it is formed at deeper places. The second and fourth item are obviously related to the pattern of the seasonal variations. Thus, nutrient concentration is a key to define the pattern. Nutrient concentration is, in turn, tightly related to vertical velocity, which means that vertical velocity should be one of the physical factors for the classification.

173

Fig. 16 shows the relation among the areal division, vertical velocity, and seasonal variation of MLD. It is clearly seen that the vertical velocity and the MLD seasonal variation define the boundaries, except for EQ where the strong Ekman divergent flow spreads out nitrate lifted up by the upwelling and the deep MLD ŽFig. 5 of KKSa. is also acting to enrich nitrate on both sides of the equator. Considering the fact that the areal division is made based only on the model chlorophyll data and no information from the physical variables is utilized, the agreement between the boundaries and the contour lines is significant. Based on Fig. 16, the seven areas can be classified as shown in Fig. 17. Indeed, the areas sorted in the same quadrant show similar seasonal variations. Resemblance between BO and PR has been, for example, already pointed out in Section 3.3; minor differences between the two areas are attributable to other various factors such as differences in light and temperature. This diagram is demonstrating that the diverse seasonal variations in chlorophyll are created

Fig. 16. Relation among the areal division, vertical velocity, and MLD seasonal variation. The red contour lines represent 0 mrday for the annual mean vertical velocity at the depth of 100 m. The blue contour-lines represent 100 m for the amplitude of MLD seasonal variation Ždifference between the maximum and the minimum MLD.. The red letter ‘‘U’’ indicates upwelling, while ‘‘D’’ downwelling. The blue letter ‘‘H’’ indicates to the regions where the amplitude of the MLD variation is larger than 100 m.

174

M. Kawamiya et al.r Journal of Marine Systems 25 (2000) 159–178

Fig. 17. Classification of the areas shown in Fig. 3. The vertical axis shows the amplitude of the seasonal variation of MLD, while the horizontal axis represents vertical velocity.

basically by a rather simple combination of the ambient physical processes. The processes described so far are essentially vertical. Importance of horizontal processes can be found in the following aspects: Ø determination of MLD distribution through horizontal heat transport; Ø determination of vertical velocity through divergence and convergence of horizontal flow; Ø future application of the model to problems such as response of the ecosystem to climate variations, where changes in horizontal heatrnutrient transport near the surface and at the intermediate depths are important. As for the first item, we do not discuss in detail the significance of horizontal heat transport on MLD distribution because it is out of the scope of this paper. But it is obvious that the heat transport by the Kuroshio is deeply involved in forming the region of deep winter mixed layer in the western North Pacific ŽFig. 5 of KKSa.. Furthermore, the horizontal transport is of potential importance in that the nitrate distribution is considerably modified by the Ekman transport if an inappropriate parameter value is chosen, as will be explained in Section 4.4. 4.2. SCM formation in the model SCMs in this model are formed through either the photoinhibition process Žin the eutrophic region. or

the balance between light and nutrient limitation Žin the oligotrophic region.. Several other mechanisms, however, have been proposed by many researchers. An excellent review on SCM by Longhurst and Harrison Ž1989. and references therein give the following possibilities: differential phytoplankton sinking rate at a density discontinuity; behavioral aggregation; physiological changes in the biomass to chlorophyll ratio; enhanced growth near the nutricline; and differential grazing by herbivores. None of these can be entirely excluded as a cause for SCM. Probably, SCM in nature is a result of their mixture. The mechanisms forming the SCMs in the model are undoubtedly greatly simplified, although they are presumably participating in the formation of real SCM. However, exact reproduction of SCM is out of the scope of this study. What should be examined is relevance of neglecting the processes pointed out above to the seasonal variations. From this point of view, it is unlikely that any of the neglected processes can seriously affect the model’s response to the two key factors determining the areal division, that is, the annual mean upwelling and the MLD seasonal variation.

4.3. Possible effects of short-term Õariation of wind stress This model is forced by monthly mean wind stress field, and the effect of short-term variation of wind stress is not taken into account. As Ridderinkhof Ž1992. showed, its effect on MLD is important especially in warm seasons when the mixed layer is shallow, while it is not in cold seasons; MLD in cold seasons is largely determined by convection process, which is included in the model although in a crude fashion. In this study, mixed layer variation is playing an important role in some areas as shown in Section 3, with the key process being its shallowing in early spring. In other words, the essence is the difference in MLD between winter and spring. Forcing the model with daily wind stress would deepen MLD from late spring trough fall. However, considering the fact that MLD in winter is generally much larger

M. Kawamiya et al.r Journal of Marine Systems 25 (2000) 159–178

175

4.4. SensitiÕity experiments Robustness of the conclusion derived from the above experiment Žreferred to as the control case hereinafter. may be examined by sensitivity experiments. Two parameters are chosen for the experiments. One is the half saturation constant for nitrate Ž K NO .. It is chosen because the value adopted in the 3 control case is very close to the lower limit of acceptable values ŽHarrison et al., 1996.. An experiment with a larger value will be instructive. The other parameter chosen is the photosynthetic rate Ž Vmax .. Because it has been proved that the balance Žor imbalance. between photosynthesis and grazing is critical to dynamics in this model, it is of interest to see consequence of changing parameter values associated with the two key processes, photosynthesis and grazing. The photosynthetic rate is one of such parameters. The main focus will be on the

Fig. 18. Annual mean Ža. chlorophyll and Žb. nitrate concentration as in Fig. 2 but for the case with the increased half saturation constant.

than that in warm seasons, it is unlikely that the impact on MLD difference between winter and spring would be very large. Consequently, the behavior of the ecosystem model would be qualitatively unchanged. We do not insist that it is meaningless to include short-term variation of wind stress. Its inclusion will result in deepening of the mixed layer, and will enhance phytoplankton growth in the oligotrophic regions and suppress it in the eutrophic regions. However, in a model to investigate a climatological state as the present one, it is extremely difficult to include the short-term variation excluding phenomena characteristic to a certain specific year. In the present state of the art, the method to force a climatological model with the effect of short-term variation is not established and can be itself a theme of investigation.

Fig. 19. Annual mean Ža. chlorophyll and Žb. nitrate concentration as in Fig. 2 but for the case with the halved photosynthetic rate.

176

M. Kawamiya et al.r Journal of Marine Systems 25 (2000) 159–178

results of EOF analysis for the experiments because they are the most relevant to the theme of this paper, namely the relation between chlorophyll seasonal variation and physical environments. Those nitrate and chlorophyll fields that will be shown in the following sections should be compared with Fig. 2. 4.4.1. Half-saturation constant for nitrate In the sensitivity experiment, K NO 3 is set to 3.0 mmolrl. This is two orders of magnitude larger than the value adopted in the control case, but is still within the range of acceptable values. Values of this order are often reported for the eutrophic region Žcf. Parsons et al., 1984.. Surface nitrate ŽFig. 18b. is increased as expected, especially in the southern part of the subtropical gyre. This is primarily due to the Ekman transport from the equatorial region. As a result, the oligotrophic region becomes fairly rich in nitrate. On the other hand, the chlorophyll distribution is not changed drastically although the value is slightly decreased and the maximum corresponding to the area UW is somewhat obscured ŽFig. 18a..

EOF analysis is applied in the same way as in the control case, and the result is shown in Fig. 20a. The resultant patterns are not modified drastically compared with those in the control case ŽFig. 1.. A distinct difference from the control case is disappearance of the peaks forming KS. This difference can be understood when the mechanism for the seasonal variation in KS is recalled: the key factor is the response of phytoplankton to the enhanced nitrate transport in winter and the reduced nitrate after spring. In this case, however, significance of the nitrate added by the winter transport is substantially reduced due to the increased nitrate. This experiment demonstrates that ability of phytoplankton to deplete nitrate after spring is essential for the existence of KS. 4.4.2. Photosynthetic rate Vmax is halved in this experiment. The chlorophyll field ŽFig. 19a. becomes remarkably flat compared with the control case. With the halved photosynthetic

Fig. 20. First two principal modes of EOF analysis as in Fig. 1a but for the case with Ža. the increased half saturation constant for nitrate and Žb. the halved photosynthetic rate. In the parentheses are the percentages by which the total variance is accounted for. Contour intervals are 0.02.

M. Kawamiya et al.r Journal of Marine Systems 25 (2000) 159–178

rate, the photosynthesis process cannot compete with nitrate transported by the Ekman flow from the adjacent eutrophic regions. Consequently, as can be seen in Fig. 19b, the entire domain becomes nitrate rich whereby the chlorophyll field has little spatial variation. Fig. 20b depicts the result of the EOF analysis. Mode patterns of EOF analysis are similar to those in the control case ŽFig. 1.. Unlike in the case of the larger K NO 3 , the peak in the EOF 2 forming KS remains although its amplitude and its proportion are decreased. The reason for the survival of KS is that phytoplankton can still use up the surface nitrate in summer owing to the small K NO 3 . Thus, the effect of reducing Vmax appears mainly in the annual mean distribution. With the small Vmax , nitrate in the eutrophic regions can penetrate into the oligotrophic region crossing the gyre boundaries, which is caused by the Ekman transport.

5. Summary and conclusion We have investigated mechanisms forming the seasonal variations of the oceanic ecosystem in the North Pacific by using the ecosystem model embedded in the OGCM. The North Pacific can be divided into seven areas on the basis of the seasonal variation patterns of chlorophyll in the model. The factors shaping each pattern are nutrient concentration which is in turn determined mainly by vertical velocity, vertical mixing, and less importantly light condition. KS and ST in Fig. 3 both lie within the downwelling region but can be separated due to strong vertical mixing in KS from winter to early spring; chlorophyll concentration in KS shows the maximum value in March as vertical mixing enriches the surface nutrient. The other areas are related to the upwelling region, but show different seasonal variations depending on dynamics of the mixed layer: BO and PR, which experience the large seasonal variation of vertical mixing bear blooming when the surface warming leads to the shallow mixed layer; on the other hand, SP, CU, and EQ go through only moderate variations caused by the variation of light condition and the modest change in vertical mixing. The areal division shown in Fig. 3 corresponds well to that by Longhurst Ž1995. based on observa-

177

tions ŽFig. 5.. Also, the seasonal pattern of each area is reasonably compared with that obtained by satellite measurements and in situ observations. Furthermore, most of the mechanisms forming the variations turn out to have some observational basis. There are, of course, differences between the model and the observations. For example, the modeled high chlorophyll concentration in BO does not last long compared with the CZCS data, and the modeled chlorophyll concentration is much lower than that in the observations. Chemical andror biological factors not incorporated in the present model may be important in these points. Furthermore, this study does not exclude a possibility that some other mechanisms are functioning even in the region where the model result is found to match the data. For example, the model cannot deny the importance of iron for suppression of blooming in SP. However, the overall agreement on the seasonal variation patters of chlorophyll and their spatial distribution between this simple model and the observations strongly supports the notion that the ambient physical conditions can account for a significant portion of the diversity of the seasonal variation patterns.

Acknowledgements We would like to thank T. Sugimoto, I. Koike, M. Kawabe, M. Takahashi, and I. Yasuda for helpful comments. Thanks are extended to the members of the ocean-modeling group at CCSR, University of Tokyo, for stimulating discussions. M. Watanabe, F. Saito, and H. Mizukami helped prepare the manuscript. We also appreciate the comments of the three anonymous reviewers, which contributed greatly to the improvement of the manuscript.

References Banse, K., English, D.C., 1994. Seasonality of coastal zone color scanner phytoplankton pigment in the offshore oceans. J. Geophys. Res. 99, 7323–7345. Chelton, D.B., Bernal, P.A., McGowan, J.A., 1982. Large-scale interannual physical and biological interaction in the California Current. J. Mar. Res. 40, 1095–1125.

178

M. Kawamiya et al.r Journal of Marine Systems 25 (2000) 159–178

Clemons, M.J., Miller, C.B., 1984. Blooms of large diatoms in the oceanic, subarctic Pacific. Deep-Sea Res. 31 Ž1., 85–95. Colebrook, J.M., 1979. Continuous plankton records: seasonal cycles of phytoplankton and copepods in the North Atlantic Ocean and the North Sea. Mar. Biol. 51, 23–32. Dugdale, R.C., Wilkerson, F.P., Minas, H.J., 1995. The role of a silicate pump in driving new production. Deep-Sea Res. 42, 697–719. Evans, G.T., Parslow, J.S., 1985. A model of annual plankton cycles. Biol. Oceanogr. 3 Ž2., 327–347. Fasham, M.J.R., 1995. Variations in the seasonal cycle of biological production in subarctic oceans: a model sensitivity analysis. Deep-Sea Res. 42, 1111–1149. Fasham, M.J.R., Sarmiento, J.L., Slater, R.D., Ducklow, H.W., Williams, R., 1993. Ecosystem behavior at Bermuda Station S and Ocean Weather Station INDIA: an observational analysis. Global Biogeochem. Cycles 7 Ž2., 379–415. Gordon, H.R., Clark, D.K., Brown, J.W., Evans, R.H., 1982. Satellite measurements of the phytoplankton pigment concentration in the surface waters of a warm core Gulf Stream ring. J. Mar. Res. 40, 491–502. Harrison, W.G., Harris, L.R., Irwin, B.D., 1996. The kinetics of nitrogen utilization in the oceanic mixed layer: nitrate and ammonium interactions at nanomolar concentrations. Limnol. Oceanogr. 41, 16–32. Hurlburt, H.E., Wallcraft, A.J., Schmitz, W.J., Hogan, P.J., Metzger, E.J., 1996. Dynamics of the KuroshiorOyashio current system using eddy-resolving models for the North Pacific ocean. J. Geophys. Res. 101, 941–976. Imai, M., Ebara, S., Kawashima, K., Kubo, N., Sato, N., Moriyama, E., 1988. Seasonal variation of chlorophyll-a in the seas around Japan. Oceanogr. Mag. 38, 23–32. Kasai, H., Saito, H., Yoshimori, A., Taguchi, S., 1997. Variability in timing and magnitude of spring bloom in the Oyashio region, the western subarctic Pacific off Hokkaido, Japan. Fish. Oceanogr. 6, 118–129. Kawamiya, M., Kishi, M., Suginohara, S., 2000. An ecosystem model for the North Pacific embedded in a general circulation model: Part I. Model description and characteristics of spatial distributions of biological variables. J. Mar. Syst. 25, 129–157. Levitus, S., 1982. Climatological atlas of the world ocean. NOAA Prof. Rep. 13. Technical report, Natl. Oceanic and Atmos. Admin. Levitus, S., Boyer, T.P., 1994. World ocean atlas 1994, NOAA Atlas NESDIS vol. 4 NODC, Washington, DC. Longhurst, A., 1995. Seasonal cycles of pelagic production and consumption. Prog. Oceanogr. 36, 77–167. Longhurst, A.R., Harrison, W.G., 1989. The biological pump: profiles of plankton production and consumption in the upper ocean. Prog. Oceanog. 22, 47–123. Longhurst, A., Sathyendranath, S., Platt, T., Caverhill, C., 1995. An estimate of global primary production in the ocean from satellite radiometer data. J. Plankton Res. 17, 1245–1271. Mann, K.H., Lazier, J.R.N., 1991. Dynamics of Marine Ecosystems. Blackwell, Boston, 466 pp. Martin, J.H., Fitzwater, S.E., 1988. Iron deficiency limits phyto-

plankton growth in the northeast Pacific subarctic. Nature 331, 341–343. McCreary, J.P., Kohler, K.E., Hood, R.R., Olson, D.B., 1996. A four-compartment ecosystem model of biological activity in the Arabian Sea. Prog. Oceanogr. 37, 193–240. Obata, A., Ishizaka, J., Endoh, M., 1996. Global verification of critical depth theory for phytoplankton bloom with climatological in situ temperature and satellite ocean color data. J. Geophys. Res. 101, 20657–20667. Parsons, T.R., Takahashi, M., Hargrave, B., 1984. Biological Oceanographic Processes. 3rd edn. Pergamon, Oxford. Ridderinkhof, H., 1992. On the effects of variability in meteorological forcing on the vertical structure of a stratified watercolumn. Cont. Shelf Res. 12, 25–36. Ryabchenko, V.A., Gorchakov, V.A., Fasham, M.J.R., 1998. Seasonal dynamics and biological productivity in the Arabian Sean euphotic zone as simulated by a three-dimensional ecosystem model. Global Biogeochem. Cycles 12, 501–530. Sambrotto, R.N., Goering, J.J., 1983. Interannual variability of phytoplankton and zooplankton production on the southeast Bering Sea shelf. In: Wooster, W.S. ŽEd.., From Year to Year. pp. 161–179, A Washington Sea Grant Publication. Sarmiento, J.L., Slater, R.D., Fasham, M.J.R., Ducklow, H.W., Toggweiler, J.R., Evans, G.T., 1993. A seasonal three-dimensional ecosystem model of nitrogen cycling in the North Atlantic euphotic zone. Global Biogeochem. Cycles 7 Ž2., 417–450. Smith, S.L., Vidal, J., 1984. Spatial and temporal effects of salinity, temperature and chlorophyll on the communities of zooplankton in the southeastern Bering Sea. J. Mar. Res. 42, 221–257. Sverdrup, H.U., 1953. On conditions for the vernal blooming of phytoplankton. J. Cons., Cons. Perm. Int. Explor. Mer 18, 287–295. Tabata, S., 1964. Main physical factors governing the oceanographic conditions at ocean station ‘‘P’’ in the northeast Pacific Ocean. PhD thesis, Univ. of Tokyo. Talley, L.D., Nagata, Y., Fujimura, M., Iwao, T., Kono, T., Inagake, D., Hirai, M., Okuda, K., 1995. North Pacific intermediate water in the KuroshiorOyashio mixed water region. J. Phys. Oceanogr. 25, 475–501. Thomas, A.C., Huang, F., Strub, P.T., James, C., 1994. Comparison of the seasonal and interannual variability of phytoplankton pigment concentrations in the Peru and California Current system. J. Geophys. Res. 99, 7355–7370. Tsunogai, S., Watanabe, Y., 1983. Role of dissolved silicate in the occurrence of a phytoplankton bloom. J. Oceanogr. Soc. Jpn. 39, 231–239. Winn, C.C., Campbell, J.R., Christian, R.M.L., Hebel, D.V., Dore, J.E., Fujieki, L., Karl, D.M., 1996. Seasonal variability in the phytoplankton community of the North Pacific Subtropical Gyre. Global Biogeochem. Cycles 9, 605–620. Yoder, J.A., McClain, C.R., Feldman, G.C., Esaias, W.E., 1993. Annual cycles of phytoplankton chlorophyll concentrations in the global ocean. Global Biogeochem. Cycles 7 Ž1., 181–193.