1 Introduction

The Fourth Assessment Report of the Intergovernmental Panel on Climate Change (IPCC-AR4, IPCC 2007) states that “Warming of the climate system is unequivocal, as is now evident from observations of increases in global average air and ocean temperatures, widespread melting of snow and ice, and rising global average sea level”. Levitus et al. (2005) reported that a mean ocean temperature (0–3000 m depth) increased 0.037 °C in the world ocean during 1955–1998. They also indicated that the world ocean is responsible for approximately 84 % of the estimated possible total increase in heat content of the Earth system for 1955–1998. The absorbed heat is confined mainly to the shallower layers in the ocean and increased sea temperature has potential to impact marine pelagic ecosystems. In the North Pacific, as an annually averaged response to global warming, decreased in chlorophyll is expected in correlation with the increased vertical stratification (Sarmiento 2004).

The productivity of the fish stocks may also change due to increased sea temperature and changes in distribution and production of important prey and predator species. Sea temperature is one of the primary factors, together with food availability and suitable spawning grounds, determining the large-scale distribution patterns of fish and shellfish (Drinkwater 2005). Some studies have indicated or speculated on the influence of climate change on fish species. Perry et al. (2005) reported that the distributions of many species, such as Atlantic cod (Gadus morhua) and snake blenny (Lumpenus lampretaeformis), in the North Sea have responded markedly by shifting northwards due to recent increases in sea temperature. They pointed out that species with shifting distributions have a shorter life cycle and smaller body size than non-shifting species. Drinkwater (2005) projected that individual growth rates of Atlantic cod will increase, leading to an overall increase in the total production of Atlantic cod in the North Atlantic in response to increased warming. His scenario based on projections assumes that past trends or relationship will be maintained. He investigated the relationship between sea bottom temperature and the growth of cod and made a projection using the relationship and projected future temperature.

Scenario projections are helpful to address issues and clarify tasks to be investigated. However, if the temporal and spatial variability associated to global warming is different from the past phenomena, the degree of confidence in the projections can be speculative. Moreover, if the amplitude exceeds that of past phenomena, it would become an extrapolation. Therefore, we need a mechanistic model to make future projections of global warming impacts on fish species. Huse and Ellingsen (2008) investigated the effect of global warming on capelin (Mallotus villosus) distribution and population dynamics in the Barents Sea by using a mechanistic model; an individual based model considering biological strategies of capelin. They predicted that capelin distribution will move to the north eastern part of the Barents Sea. The capelin would utilize new spawning areas along Novaya Zemlya start spawning earlier due to the effects of the warming. This kind of mechanistic model allows to identify the key issues and items that require verification.

In this study, we focused on Japanese sardine (Sardinops melanostictus) which is one of the commercially important species in the western North Pacific. The stock of sardine has exhibited dramatic changes in the western North Pacific. The annual catch of sardine around Japan reached levels between 2.2 million and 4.5 million tons in the 1980s, while it has declined drastically in the early 1990s, and the stock of Japanese sardine is currently at a low level. The main spawning of Japanese sardine occurs along central and southern Japan from February to March (Hiramoto 1981; Kuroda 1991; Watanabe et al. 1995). Sardine larvae metamorphose to juveniles while they are transported to the north-east by the Kuroshio (the western boundary current of the subtropical gyre in the North Pacific) and Kuroshio Extension (see Fig. 1). During spring and summer, juvenile and adult fish migrate northwards to their feeding grounds (the Oyashio area) through the mixed water region which is the region between the Kuroshio Extension and the Oyashio Front (Kondo 1980; Hiramoto 1981; Kuroda 1991; Watanabe et al. 1995).

Fig. 1
figure 1

Oceanographic features of the western North Pacific (grey curve arrow, Oyashio; broken curve arrow, Kuroshio and Kuroshio Extension) and spawning regions in the model. Spawning areas are shown with squares (Area-1; 40 cells), triangles (Area-2; 36 cells), circles (Area-3; 55 cells) and crosses (Area-4; 63 cells). Black rectangle around Area-4 shows the location of Tosa Bay, which is main spawning region in the current state. Shaded area in the Kuroshio and Kuroshio Extension shows predicted high-fish-density region during winter to spring (February–May) in the model (Simulation-1) in Section 3. This region is analysis area in Fig. 4

Full size image

Japanese anchovy utilize similar feeding grounds, but anchovy and sardine show asynchronicity in their population response to decadal climate variability such as Pacific Decadal Oscillation (PDO, Mantua et al. 1997). Concerning the population alternation between sardine and anchovy, Takasuka et al. (2007b) proposed a simple hypothesis, called the “optimal growth temperature hypothesis”. They showed the optimal growth temperature of Japanese sardine larvae is lower than that of Japanese anchovy (Engraulis japonicus), and hence suggested the stock of Japanese sardine increase and anchovy decrease simultaneously during a cooler climate regime. Overland and Wang (2007) investigated IPCC-AR4 projections and pointed out that the amplitude of PDO is not projected to change drastically during the 21st century but the amplitude of the monotonic trend by the global warming would surpass that of PDO by the middle of the 21st century. Therefore, the effect of global warming would dominate in the future. Takasuka et al. (2007b) also pointed out the possibility that sardine could face collapse in the western North Pacific under a 2–3 °C rise of sea surface temperature (SST) condition due to global warming, because they will not be able to avoid basin-wide scale warming in the long term although fish might migrate to more favorable habitats. However, whether sardine could change spawning areas and periods in order to adapt to environmental changes due to global warming remains an open question. Here, to answer this question, we use a mechanistic model (Okunishi et al. 2009) to conduct numerical experiments with present day and global warming oceanic conditions as predicted form a high-resolution climate model.

2 Model

2.1 Multi-trophic level ecosystem model

We developed a multi-trophic level ecosystem model. This model includes physical, biogeochemical and plankton dynamics, and additionally behavioral ecology of Japanese sardine. The modeling framework is composed of mainly three components; (1) a climate model (coupled atmosphere–ocean model), (2) a lower trophic level ecosystem model, (3) a fish migration model of Japanese sardine (Fig. 2).

Fig. 2
figure 2

Schematic view of the multi-trophic level ecosystem model. Left figure shows the outline of experimental setting in the tow climate scenarios. NO3, NH4, PS, PL, ZS, ZL, POM, DOM and ZP show the compartments with NEMURO, and arrows represent nitrogen silica flow in the central figure. Right figure shows flowchart of the individual-based model structure for Japanese sardine

Full size image

2.2 Climate model

The climate model is a high-resolution setup of the Model for Interdisciplinary Research on Climate (MIROC) version 3.2 (K-1 Model Developers 2004). MIROC3.2 has contributed to IPCC AR4 (IPCC 2007). For physical fields, we used results of two experiments under pre-industrial condition (hereafter Control-RUN) and global warming condition (hereafter CO2-RUN) by the MIROC model (e.g. Sakamoto et al. 2005). The Control-RUN was carried out by fixing the external forcing at the year 1900 by the climate model. The CO2-RUN is a global warming experiment, in which the atmospheric CO2 concentration was increased at the rate of 1 % year−1 from pre-industrial condition. In the present study, the analysis periods are 10 years from the 46th year to the 55th year after spin-up in the Control-RUN, and from the 76th year to the 85th year after spin-up in the CO2-RUN (Fig. 2). This corresponds for the CO2-RUN to a doubling atmospheric CO2 concentration since the pre-industrial condition. This scenario is almost similar to the IPCC SRES A1B scenario (IPCC 2000). Annually averaged SST were predicted to increase by 2 to 3 °C in most regions in the western North Pacific in the CO2-RUN from that in the Control-RUN. Especially, a large SST rise is found along the Kuroshio and Kuroshio Extension due to the increase of volume transport by the Kuroshio and Kuroshio Extension (Sakamoto and Hasumi 2008).

2.3 Lower trophic level ecosystem model

The lower trophic level ecosystem model is NEMURO (North pacific Ecosystem Model for Understanding Regional Oceanography) (Kishi et al. 2007) developed by PICES (the North Pacific Marine Science Organization). This model is a nitrogen- and silica-based model with eleven compartments: nitrate (NO3), ammonium (NH4), small phytoplankton (PS), large phytoplankton (PL), small zooplankton (ZS), large zooplankton (ZL), particulate organic matter (POM), dissolved organic matter (DOM), predatory zooplankton (ZP), particulate silica (Opal), and silicate (Si(OH)4). In the present study, the lower trophic level ecosystem model is forced off-line by ocean physical fields predicted from the ocean part (ocean general circulation model) of the climate model with a horizontal resolution of 0.28° (zonally) x 0.19° (meridionally) (Hashioka et al. 2009). NEMURO was applied to the surface 1500 m in the western North Pacific (about 115E–175°W, 5–65°N). We used adjusted parameters in the NEMURO for the western North Pacific as estimated with a data assimilation method (Ito et al. 2010). Simulated plankton density fields by NEMURO are used as prey density conditions in the fish migration model.

2.4 Fish migration model

The fish migration model uses individual-based modeling techniques, which was applied to Japanese sardine in the western North Pacific (Okunishi et al. 2009). The individual-based model (IBM) is composed of a bioenergetics sub-model and a Lagrangian transport sub-model. The bioenergetics sub-model estimates growth of sardine and determines swimming directions. The Lagrangian transport sub-model simulates the position of sardine using the ocean current and predicted swimming directions and speeds by the bioenergetics model. Swimming direction during feeding migration was governed by search for the local optimal habitat, which was estimated by the spatial distribution of net growth rate of the bioenergetics sub-model (Okunishi et al. 2009).

We conducted the Control-RUN and the CO2-RUN using the fish migration model forced on predicted sea surface current and sea surface temperature (SSTCM: 0–30 m depth) by the climate model and simulated surface prey density (vertical mean ZS and PL at 0–30 m depth) by NEMURO. The predicted sea temperature from the climate model not including data-assimilation system has a bias as compared with the observed sea temperature. There is a positive (negative) SST bias in the Oyashio region (the mixed water, Kuroshio Current and Kuroshio Extension regions (Fig. 3). We have to correct the SST bias in the fish migration model, because the bioenergetics sub-model has a growth function depending heavily on the absolute sea temperature. In the fish migration model, corrected SST (SSTFISH) was used by subtracting the SST correction term (ΔSST) which is the bias between the 10 years averaged monthly SST (0–30 m depth) in the Control-RUN and the climatological monthly SST of World Ocean Atlas 2005 (Locarnini et al. 2006). Then, SSTFISH = SSTCM + ΔSST, where SSTFISH is SST in the fish migration model.

Fig. 3
figure 3

Anomaly of annual mean SST (0–30 m depth) between the World Ocean Atlas 2005 and the Control-RUN

Full size image

Vulnerability of prey (ontogenetic selection of prey plankton) of sardine at each life stages is the same as that in Okunishi et al. (2009). The fish migration model has a domain of 20–60 °N and 120–180 °E in the western North Pacific. The bioenergetics part allocates the consumption rate C (g prey g fish−1 day−1) over metabolic processes such as respiration R, specific dynamic action S, waste losses due to egestion F and excretion E, and the mass-specific growth rate of an individual fish is defined as:

$$ \frac{1}{W}\frac{{dW}}{{dt}} = \left[ {C - (R + S + F + E)} \right] \cdot \frac{{CA{L_z}}}{{CA{L_f}}}, $$
(1)

where W is the wet weight of the fish (g), t is time (days), CAL z and CAL f are caloric equivalents of prey (cal g prey−1) and fish (cal g fish−1) respectively. The mathematical formulations for these terms are based on Ito et al. (2004). The mathematical formulations in the bioenergetics sub-model are the same as that in Okunishi et al. (2009). To adjust the simulated body size and weight in the Control-RUN to the observed body size at larval stage (Watanabe and Kuroki 1997) and body weight at age 1 from 1973–2002 (Japanese stock assessment report, http://abchan.job.affrc.go.jp), six ecological parameters associated with consumption were changed from Okunishi et al. (2009). In consequence, we used a c  = 0.35 for an intercept of the allometric mass function for consumption and K 11 = K 21 = K 22 = K 31 = K 32 = 0.02 (g-prey m−3) for half-saturation, although Okunishi et al. (2009) used a c  = 0.24, K 11 = 0.03, K 21 = K 22 = 0.05, K 31 = K 32 = 0.5. In Okunishi et al. (2009), there were typos in the equations of egestion (F) and excretion (E). The correct equations are F = ϕC and E = γ(C − F), where ϕ is used 0.1 as the proportion of consumed food egested, γis used 0.16 as the proportion of consumed food excreted.

2.5 Key questions and design of experiments

  1. Simulation-1:

    The model was designed to investigate the impact of global warming on spawning grounds and period of Japanese sardine. Firstly, we focused on two key questions as follow:

Under the global warming condition,

  • Would optimal spawning grounds of sardine shift from the present spawning grounds? (Q1)

  • Would the main spawning period and season of sardine change? (Q2)

To elucidate Q1 and Q2, nine simulation experiments of 120-days from spawning day were conducted using the predicted 10 years forcing in the Control-RUN and the CO2-RUN, respectively (Simulation-1). A key point is growth change depending on environmental changes (SST and forage density). In the model, spawning sites were set as shown in Fig. 1 based on historical observation data (e.g. Kuroda 1991; Watanabe et al. 1996; Watanabe et al. 1997). The spawning sites were divided into four regions (Area-1, Area-2, Area-3 and Area-4) for convenience of analysis of the model results. In the current state, the main spawning region is formed in the offshore region in Tosa Bay (e.g. Ishida 2006; Takasuka et al. 2008), which corresponds to Area-4 in the model. While currently, spawning season peaks in February–March, we assumed that individual larvae could hatch from December 1st to April 30th in each spawning points at intervals of one day with the condition that SST is within the optimal range of spawning, i.e., 15–21 °C (Kuroda 1991). The total numbers of hatched larvae in the four spawning regions are the sum of successful hatchings per day over the period December 1st–April 30th. The hatching success rate defines as follow:

$$ HR = {{{NS}} \left/ {{NT}} \right.}, $$
(2)

where HR is the hatching success rate, NS is the number of particle within the range of SST at 15–21 °C at timing of particle release (larvae of hatching success), NT is the total number of setting in each spawning points. The effect of mortality was neglected in Simulation-1 and we used the difference in growth of individuals under various environmental conditions as an indicator of match-mismatch with high prey density and optimal temperature.

  1. Simulation-2:

    We additionally focused on two key questions as follow:

Under the global warming condition,

  • Would the recruitment abundance of sardine change? (Q3)

  • Would fish distributions at the adult stage change during the fishing season (summer)? (Q4)

To elucidate Q3 and Q4, the model was carried out 608 days from the February 1st, year 0 to September 30th, year 1 using the predicted physical and forage conditions in the Control-RUN and the CO2-RUN, respectively (Simulation-2). The spawning points are the same as Simulation-1 in the four regions. The spawning period is only February, because we regard this period important to focus on their response to the global warming in discussing Q3 and Q4 (see Section 4.1). In Simulation-2, fish migration was controlled by the feeding migration only, because fish in the simulations end (September 30th, year 1) are still too young for the spawning migration (e.g. Kuroda 1991). Super-individuals (Scheffer et al. 1995) were used to allow the IBM to represent the population of recruited stock in the western North Pacific. Four million identical individuals are pooled together as each super-individual. Potential population numbers are twenty billion individuals if all larvae would hatch from the spawning points during the spawning period. The internal number in a super-individual is reduced due to mortality. Mortality rate (MR) was regarded to be dependent on fish size. Moreover, we considered the effect of growth-selective predation by piscivorous predators on larvae based on Takasuka et al. (2007a). Based on Kuroda (1991) and Kawai (1987), mortality rate of standard growth individuals (MRS) was assumed as follows:

  • MRS = 0.075 day−1 at the early larvae stage (body length: <8 mm),

  • MRS = 0.01 day−1 at the late larvae stage (body length: 8–50 mm),

  • MRS = 0.001 day−1 at the juvenile and adult stages (body size: >50 mm).

We assumed that individuals having a high-growth-rate (0.6 > mm day−1) have a low-mortality-rate (MR = 0.5 MRS), those having a low-growth-rate (0.4 < mm day−1) have high-mortality-rate (MR = 1.5 MRS) and those having a standard-growth rate (0.4–0.6 mm day−1) have standard-mortality-rate (MR=MRS) during the larvae stage. During the juvenile and adult stages, non-growing individuals (growth rate = 0 mm day−1) have a high-mortality-rate (MR = 5 MRS), normal growing individuals (growth rate > 0 mm day−1) have a standard-mortality-rate (MR=MRS).

3 Results

3.1 Forage conditions for sardine larvae

In the Control-RUN, zooplankton biomass had a range (25–125 mg mg-wet m−3) in the Kuroshio and Kuroshio Extension similar with the observed zooplankton biomass (50–100 mg-wet m−3) by Ikeda and Motoda (1978). In the CO2-RUN, ZS increased by 7.5 % in the Kuroshio and Kuroshio Extension (Fig. 4), which are nursery regions of Japanese sardine, during winter to spring. The increased temperature caused a reduction of the nutrient supply from the deeper layer and hence decreased PL by 14 % in comparison with the Control-RUN. However, it caused an increase of regeneration of PS from NH4 by 13 % compared with the Control-RUN since regeneration of PS depends on temperature raise. This increase of production of PS resulted in the increase of the production and the biomass of ZS by 15 and 7.5 %, respectively. These changes resulted in favorable forage conditions for sardine larvae under the global warming condition.

Fig. 4
figure 4

Schematic view of nitrogen flow fluxes and averaged stock sizes in the CO2-RUN in the surface layer (0–30 m depth) in the Kuroshio and Kuroshio Extension, which are the high-fish-density region during winter to spring in the Simulation-1 (shown in Fig. 1). Significant variations of nitrogen flow fluxes between the CO2-RUN and the Control-RUN (increase or decrease more than 10 %) show values above flux arrows. NO3, NH4, PS, PL, ZS, ZL, POM, DOM and ZP are the compartments defined with NEMURO. Stock size values are shown near each compartment as μM-N unit in the Control-RUN. Values between brackets give the variation of stock sizes between the CO2-RUN and the Control-RUN

Full size image

3.2 Hatching rates and growth at the larval and juvenile stages (Simulation-1)

In the Control-RUN, the main spawning regions were formed in Area-4 (Table 1), where hatching success rates were between 86.3–96.9 % with the maximum in January. The hatching success rate decreased according to the latitude since the water temperature becomes too cold for the successful hatching. This result is consistent with the observations that the main spawning regions are formed in the Tosa Bay during winter (e.g. Ishida 2006; Takasuka et al. 2008). In Area-3, the hatching success rate is high in April and consistent with observed high occurrence of spawning during spring (Takasuka et al. 2008). But, the hatching success rate is also high in December in Area-3, although the observation data shows that egg production is low in December in Area-3 (Kubota et al. 1999). In the CO2-RUN, the hatching success rates decreased considerably in Area-4 (31.5–78.0 %) compared with those in the Control-RUN and the main spawning region was shifted to Area-3 (62.7–94.0 % with the maximum in January). The main spawning grounds shift from Areas-3 and 4 in the Control-RUN to Area-2, 3 and 4 with most favorable condition in Area-3 in the CO2-RUN, i.e., a latitudinal shift to the north.

Table 1 Hatching success rate (%) in the four spawning regions in each cohort with spawning month
Full size table

In Fig. 5, we show the predicted averaged weights of the high-hatching success rate cohorts (more than 50 %) at the larval and juvenile stages (120 days) issued from predefined spawning areas. The weight is used as an indicator to compare growth rate under present day and global warming in Simulation-1 (Fig. 5). In the most northern spawning area (Area-1), there is no cohort reaching the minimum threshold of 50 % hatching rate. In Area-2, averaged weights after 120 days in the CO2-RUN were larger than in the Control-RUN. Similar results are predicted for cohorts hatching in Area-3, excepted for the month of April, and in the most southern area (Area-4) all cohorts of control-run have higher growth rate than in the CO2-RUN. To explain these changes in growth rates, we investigated the temperature and forage conditions encountered by the February-spawned cohort from Area-3 and 4 (Figs. 6 and 7). Larvae cohort hatching in February in Area-3 found much favorable temperature feeding conditions and slightly higher forage (small zooplankton) biomass during their first 2 months of live in the CO2-RUN than in the Control-RUN (Fig. 6). Logically, these different conditions led to a higher proportion of low weight larvae in the Control-RUN, comparatively to the CO2-RUN. In the southern region (Area-4), the situation is different with larvae and juvenile fish encountering more favorable feeding temperature in the Control-RUN than in the CO2-RUN (Fig. 6), but the forage density is still higher in CO2-RUN during the first 2 months. The result is a higher frequency of low-weight-fish in the CO2-RUN but still a good proportion of medium-weight-fish. Additionally, fish weight distributions for Simulation-2 that includes mortality rates are also shown on Figs. 6 and 7. Since mortality rates increases with lower growth rates and conversely decreases with higher growth rates, the logical result when considering % frequencies is a decrease in frequency of low-weight-fish and an increase in frequency of high-weight-fish.

Fig. 5
figure 5

Predicted averaged weights of the 9 years case simulations in the three predefined spawning regions where hatching success is above 50 % (a: Area-2, b: Area-3, c: Area-4). Each line shows a difference of spawned month

Full size image
Fig. 6
figure 6

Averaged temperature (a, b) and ZS density (c, d) on the fish location during 120-days and histogram of weight of Japanese sardine after 120-days (e, f) in the Control-RUN and the CO2-RUN in Area-3 in the February-spawned cohorts. Thin lines show averaged values each simulation year and thick lines show 9 years averaged values in a, b, c and d. Grey bands a, b show the range of optimal temperature of feeding consumption of sardine (16–18 °C). a, b, c show results in Simulation-1. d and e show results both in Simulation-1 and Simulation-2

Full size image
Fig. 7
figure 7

Averaged temperature a, b and ZS density c, d on the fish location during the 120-day simulations and histogram of weight of Japanese sardine after 120-days (e, f) in the Control-RUN and the CO2-RUN in Area-4 in the February-spawned cohorts. Thin lines show averaged values each simulation year and thick lines show 9 years averaged values in a, b, c, d. Grey bands (a, b) show the range of optimal temperature of feeding consumption of sardine (16–18 °C). a, b, c show results in Simulation-1. d and e show results both in Simulation-1 and Simulation-2

Full size image

3.3 Geographical distribution of the adult sardine (Simulation-2)

In Simulation 2, averaged hatched larvae were twelve billion individuals in the Control-RUN and thirteen billion individuals in the CO2-RUN. After the twenty-month simulations, averaged abundance of recruited stock decreased down to 4 % in the Control-RUN and 3 % in the CO2-RUN. There was no significant difference between survival rate in the Control-RUN and the CO2-RUN (Welch’s t-test, p < 0.05). There was also no significant difference between averaged weight and histogram of weight in the Control-RUN and that in the CO2-RUN at the juvenile and adult stages in Simulation-2 (Welch’s t-test, p < 0.05) (Fig. 8). Simulated average weights in both climate scenarios have similar values in comparison with observed data, which the range of averaged weights are 30–69 g at Age-1, 49–99 g at Age-2 from 1973–2002 (Japanese stock assessment report, http://abchan.job.affrc.go.jp). In the CO2-RUN, averaged temperature on the fish location was higher (by 0.7–1.7 °C) than that in the Control-RUN from February to April at Age-0 (Fig. 9a). Averaged forage density in the fish location with the CO2-RUN was higher than in the Control-RUN from February to March and June at Age-0 (Fig. 9b). Main fish location in the feeding ground, where frequency of appearance is up to 5 % in one degree latitude, was from 33°N to 40°N in the Control-RUN, and from 36°N to 46°N in the CO2-RUN during June and August, Age-1 (Fig. 10). During this northward migration period, averaged geographical distribution of fish moves from one to two degrees northwards under the global warming condition due to rising temperature.

Fig. 8
figure 8

Averaged weight and histogram of weight of Japanese sardine in the simulation-2

Full size image
Fig. 9
figure 9

Averaged temperature a and averaged forage density (PL + ZS) b on the fish location in the Simulation-2. Significance levels of the differences between mean values at each year were tested with the Welch’s t-test. Grey belts show significant differences (Welch’s t-test, p < 0.05)

Full size image
Fig. 10
figure 10

Histogram of north latitude on the location of Age-1 fish in June a, July b and August c in the Control-RUN and the CO2-RUN. Parenthetical values in the legend are averaged north latitude

Full size image

With currents, temperature and prey plankton density determine the movement of fish in their feeding ground. The migration algorithm allows to fish to move to the place providing optimal growth, thus allowing the fish to adapt to changing conditions under the scenario of increasing CO2. We defined optimum habitats (OH) as the regions where over 10 % of occurrence rate was observed in the optimum temperature of feeding consumption (16–18 °C) during August, Age-1. 62 % (91 %) of Fish in the Control-RUN (in the CO2-RUN) are located within the OH (Fig. 11). The latitude ranges of the OH in the western North Pacific were 36.3–42.6°N in the Control-RUN and 39.4–46.5°N in the CO2-RUN, respectively. Geographical distributions of Japanese sardine can be altered directly through climate-induced changes in temperature. Temperature changes shift Japanese sardine in the western North Pacific to around 1–2° northwards. The southernmost part of the OH in the CO2-RUN was the downstream site of the Oyashio (the Oyashio and the Mixed Water region) (Fig. 11b).

Fig. 11
figure 11

Probability distributions of the optimum temperature for feeding (16–18 °C) of Japanese sardine in the Control-RUN a and the CO2-RUN b in August. Contour intervals are 10 %

Full size image

4 Discussion

4.1 Change in spawning areas and periods

Simulated growth curves in the Control-RUN predicted optimal growth during January to February in Area-4 (Fig. 5c). This timing and spawning ground are consistent with current observations suggesting that Japanese sardine have optimized their spawning strategy under the present oceanographic conditions, by selecting the most favorable place and season for larvae to grow up. However, this advantage was projected to diminish by the effect of global warming. The growth rate during the larval stage directly influences their mortality through predation processes; faster-growing larvae are less vulnerable to predation (e.g. Takasuka et al. 2003, 2004). We can conclude that low-weight-fish at 120 days in the model are abortive individuals. Therefore, it may be acceptable to increase the mortality rate of the low-growth fish in Simulation-2 (Fig. 7e, f). The model results suggest that February-spawned fish in Area-4 have a low recruitment rate under the global warming condition. This result suggests the possibility that sardine in the current main spawning region will be placed at a disadvantage for their recruitment and may face a population collapse under the global warming condition. However, in Area-2 and Area-3, fish having a more favorable temperature condition in the CO2-RUN mature faster than those in the Control-RUN (Fig. 5a, b). Under global warming, the number of hatching larvae increases by 3 to 15 % in the sum of Area-2 and Area-3 (Table 1). Thus, sardine could move their spawning ground toward more north regions, and succeed in recruitment.

Average of monthly egg abundance in February as a percentage of annual abundance was more than 50 % between 1978 and 1992 (Watanabe et al. 1996). Peak spawning month (February) of sardine historically has not changed as far as we know. Regulation of gonadal cycle in teleosts involves a complex interaction of temperature and photoperiod (Peter and Crim 1979). Day length may be one of the important factors determining the spawning period of Japanese sardine. Therefore, it seems presumable to suppose that sardine will respond to the global warming condition with a strategy of changing the spawning regions instead of changing spawning period, although the model results suggest that some cohort of sardine would obtain a favorable temperature condition if they change the spawning period.

The direct effect of temperature would be a trigger to determine the spawning regions. Therefore, it is important to focus on the response to global warming of February-spawned cohorts in the model for discussing change in recruitment abundance and geographical distributions of Japanese sardine. Then, we discuss these subjects in Section 4.2 and 4.3 from the results in February-spawned fish in Simulation-2.

4.2 Change in recruitment abundance

In Simulation-2, model results suggest that the abundance of recruited stock has statistically the same level under the global warming condition in comparison with the current state. Fish in the larval stage, which have a high mortality rate, are projected to be exposed to more favorable forage condition under the global warming condition than that in the current state (Fig. 9). Around the Kuroshio Extension, increasing temperature with global warming leads to high ZS biomass rather than decreasing ZS biomass nutrient with strengthened stratification in the model (Fig. 4). This result is at least partly consistent with the observational result by Nakata et al. (2001). They reported that the spring biomass of small copepods has a positive correlation with SST in February in the slope water off the Pacific coast of central Japan near the Kuroshio. Cumulative mortality between the egg stage and age 1 recruitment is thought to be responsible for eventual population fluctuations of the sardine (Watanabe et al. 1995). The model result gives us an expectation that the sardine populations will increase under the global warming condition. However, the abundances in both climate scenarios are similar. Consumption rate is consistently maintained at more than 70 % of the maximum consumption rate by the effect of prey-density-dependence in this forage-density level in the model. It means that fish growth and mortality have a low sensitivity for forage condition in the model and the model results depend heavily on temperature condition on the fish. Sardine shifted the spawning ground northwards regions (Area-2, Area-3) (Table 1) and a large number of fish in the Area-3 were exposed the favorable temperature condition for feeding consumption in the CO2-RUN (Fig. 6b) especially during February to April (Fig. 9a). Therefore, the rising temperature in the CO2-RUN did not cause the decline in the abundance of recruited stock.

The intensified wind by amplified Aleutian Low Pressure in the wintertime causes the low sea surface temperature in the east of Japan from winter to early spring. This condition increase the biological production on the sardine migration route owing to the enhanced nutrient supply due to the deepening of the winter mixed layer depth (Yasuda et al. 1999), hence supporting good recruitment of Japanese sardine through higher food availability (Yatsu et al. 2005). Previous studies imply that the increasing temperature by the warming causes a decreasing MLD and strengthened stratification in the Kuroshio and Kuroshio Extension, leading to a decrease of abundance of recruited stock under the global warming condition, contrary to the model prediction. Therefore, the processes that can either increase or decrease small-zooplankton biomass in spring in the Kuroshio and Kuroshio Extension with global warming are key mechanisms that need to be investigated in details.

Another important issue is temperature dependency. Although larvae successfully hatched with the optimal temperature for spawning both in the Control-RUN and the CO2-RUN, averaged temperature on the fish location in the CO2-RUN is significantly higher (0.7–1.7 °C) than that in the Control-RUN at the larval stage (Fig. 9b). Averaged temperature in February during early larvae was 16.6 °C in the Control-RUN and 18.1 °C in the CO2-RUN (Fig. 9a). In the model, this temperature difference of 1.5 °C is estimated to induce only food consumption difference of 1.5 %, and hence the survival rate was not significantly changed in the CO2-RUN compared with that in the Control-RUN. However, this result strongly depends on the relationship between temperature and growth rate. Moreover, this relationship is based on observation data that showed a large degree of scatter (Takasuka et al. 2007b) and therefore we need more field studies clarifying whether this temperature increase due to projected global warming will cause significant differences in the growth and survival rates of larval or not.

4.3 Change in geographical distributions

In the CO2-RUN, the acceleration of Oyashio (Sakamoto et al. 2005) prevents the increase in SST in this region under the global warming condition. Therefore, the SST anomaly between the CO2-RUN and the Control-RUN in the downstream site of the Oyashio (near 40°N, 147°E) is lower (<2 °C) than that in the whole Oyashio region in August (Fig. 12a). This will not cause a significant change (shifting northwards) of fishing grounds around Japan in summer. Ito et al. (2006) pointed out that the acceleration of the Oyashio was found in observed data (1960–2002) of southernmost position of the First Branch of the Oyashio, which is characterized by the position of the 5 °C isotherm at 100 m depth (Okuda 1986), and they also pointed out that temperature in the water mass of the mixed water region had a tendency to decrease except for SST. There is a cooling trend in the mid-latitude North Pacific during 1955–1998 despite the increase in mean temperature in the world ocean (Levitus et al. 2005). These previous studies are consistent with the results in the model, such as the fishing grounds around Japan could sustain under the global warming condition.

Fig. 12
figure 12

Averaged SST in the Control-RUN in August (contour lines) and the anomaly of averaged SST between the CO2-RUN and the Control-RUN in August (color contour). White line area shows the range of optimal temperature of feeding consumption of sardine (16–18 °C) the Control-RUN a. Green shaded areas show the geographical distribution of optimum temperature of feeding consumption in the CO2-RUN b, the A1B c and B1 d scenarios of the IPCC-AR4 in August. Monthly averaged SST for periods 2076–2086 of MIROC-Hi model results were used to estimate this areas in the A1B and B1 scenarios. These scenarios SST corrected the model bias by using the anomaly SST between the averaged monthly SST for periods 1980–1999 and the climatological monthly SST of World Ocean Atlas 2005 (Locarnini et al. 2006). Contour lines intervals are one degree Celsius

Full size image

Additionally, we compared the geographical distribution of optimum temperature for feeding (16–18 °C) of Japanese sardine in the Control-RUN (Fig. 12a), the CO2-RUN (Fig. 12b), the A1B (Fig. 12c) and B1 (Fig. 12d) scenarios of the IPCC-SRES (IPCC 2000). We used monthly averaged SST for periods 2076–2086 of MIROC-Hi model results with the A1B and B1 scenarios (IPCC AR4 Model Output website; http://www-pcmdi.llnl.gov/ipcc/about_ipcc.php). The difference of geographical distribution of optimum temperature between the A1B and B1 is not so pronounced, although they have different greenhouse gas emission scenarios.