!
!
6!
Chapter 2 Climate effect on species distribution: case study in Japan Sea
(This chapter is originally published as: Tu, C.-Y., Tian, Y., & Hsieh, C.-H. (2015).
Effects of climate on temporal variation in the abundance and distribution of the demersal fish assemblage in the Tsushima Warm Current region of the Japan Sea.
Fisheries Oceanography, 24(2), 177–189. http://doi.org/10.1111/fog.12101.)
Introduction
To demonstrate the climate effect on species distributional response, I looked into the demersal fish assemblage of the Tsushima current region of Japan Sea. The data we analyzed comprised the catches and efforts of the Japanese single-trawling fishery, which were derived from the Japan Sea Offshore Bottom Trawl fisheries dataset (JSOBT). Because of the wide-coverage and the target species vary in their geographical affinities and life history traits (Table 2-1), the single trawling data may provide a unique opportunity to examine how the biological characteristics of species influence their response to climate variability.
Previous studies of the ecosystem response to climate variability in the Japan Sea found possible relationships between climate and biological populations (see the summary in Chiba et al. (2008)). The Japan Sea is a semi-enclosed sea, thus the seasonal and interannual variations in its hydrography are presumably affected by basin-scale climatological events (Naganuma 2000, Watanabe et al. 2003). The 1976/77 shift in the plankton biomass in the Japan Sea ecosystem coincided with the Pacific Decadal Oscillation (Chiba and Saino 2002, Chiba et al. 2005). This change to a lower trophic level might have been transferred to upper trophic levels because the
abundances of fishes also exhibited similar decadal variability. A more recent study found that different pattern of abundance fluctuated between the cold- and warm-water species during the regime shift in the late 1980s (Tian et al. 2011), with indicates the importance of the geographical affinity of fishes in response to environmental forcing.
However, most of these studies only focused on changes in abundance. Thus, the climate-driven changes in the geographical distributions of the biological population at interannual and decadal scales are unclear. The geographical distributions of some demersal species appear to reflect the decadal variability in climate (Tian et al. 2008), but no systematic studies have investigated the distributional responses of demersal species at interannual and decadal scales. In addition, the relationships of these responses to the ecological and life history traits of demersal species remain elusive.
In this chapter, I assessed the environmental variability in the Tsushima Current region of the Japan Sea based on local- and basin-scale environmental indicators. I also investigated whether the changes in distribution and abundance were significant at interannual (1972–2002) and decadal scales [between the cold (1977–1988) and warm (1989–2002) periods]. Finally, I determined how well the geographic affinity and life history traits could explain the sensitivity of the responses of species. I anticipated that this approach based on using two temporal scales to assess changes in geographical distribution and abundance would elucidate how demersal fishes have responded to environmental variation in the Japan Sea ecosystem.
Material and methods Fisheries data
To examine the effects of climate on the abundances and distributions of demersal species, I studied 19 species from the fisheries targets that underwent single trawling
doi:10.6342/NTU201802401
!
!
!
8!
during 1972-2002 based on the JSOBT dataset (Misu 1974, Tian et al. 2008, Tian et al.
2011). For further analysis on life history effect and data consistency, I excluded species recorded as a group and the data in offshore area owing to the limited species and discontinuous records in space and time. To generate the distribution map for each month for each species, I summed up the catch and effort in each fishing area (the smallest unit in record is 10’ x 10’ grid) and calculated the catches per united effort (CPUE) by dividing the total catches by the total efforts (number of hauls) in the grid.
In following analysis, I only consider the spawning season of each species because climate variability acting at the ocean surface is more likely to affect demersal species during their spawning (Minami and Tanaka 1992, Rijnsdorp et al. 2009). I used the spatial data averaged across the spawning period to produce the annual map.
For the indices of spatial distribution to each species, I calculated CPUE-weighted median latitude and longitude as the distribution center for each species using annual CPUE map, and a time series for the distribution center for each species was obtained.
In addition, the southern and northern boundaries were calculated as the minimum and maximum latitudes, respectively, where a species occurred on the annual CPUE map.
Similarly, the eastern and western boundaries were calculated as the maximum and minimum longitudes, respectively.
Environmental variables
To understand the effects of climate on the Japan Sea ecosystem, I examined local- and basin-scale climate indicators. The local-scale indicator was the 50-meter depth water temperature (wt50m) in the Tsushima Current region. The basin-scale climate indices comprised the Pacific Decadal Oscillation (PDO; Mantua and Hare (2002)), Arctic Oscillation (AO, Thompson and Wallace (1998)), North Pacific Index (NPI,
Trenberth and Hurrell (1994)) and Asian Monsoon Index (MOI, (Hanawa et al. 1988, Watanabe et al. 2003)). In the analysis of the environmental data, I only used the quarterly data that corresponded to the fishery data.
The correlation analysis between the local- and basin-scale climate indicators detected a complex interaction between atmospheric forcing and local water temperature (Fig. 2-2, Table 2-2).
Analysis of environmental effects on the demersal species
To understand the effects of climate on the demersal species, I analyzed the shifts in distribution and abundance separately at both interannual and decadal scales, I used the analytical framework reported by Hsieh et al. (2009b), and outlines of the procedures are given as follows. I defined the cold (1977–1988) and warm (1989–2002) periods based on previous studies of the decadal variability in the Japan Sea ecosystem (Chiba and Saino 2002, Chiba et al. 2005, Tian et al. 2008). In the decadal analyses, I excluded data obtained before 1976 because previous studies (Chiba and Saino 2002, Chiba et al. 2005, Tian et al. 2008) suggest that a decadal shift occurred in 1976/77. In addition, the data period (1972–1976) was too short to make quantitative comparisons.
Thus, I focused on the decadal event. The temperature anomalies suggested that a minor warming may have occurred prior to 1976, but it is likely to have been part of fluctuations on a shorter time scale in the Japan Sea rather than a decadal event (Katoh et al. 2006).
To examine the distributional response at the interannual scale, I performed regression analysis of the environmental variables and distribution center (median latitude/longitude) for each species. I also investigated 1-year and 3-year time-lagged values that represented the delayed environmental effects, which have been observed in
doi:10.6342/NTU201802401
!
!
!
10!
fish species in the North Sea (Perry et al. 2005) and southern California region (Hsieh et al. 2008a). To account for the serial dependency of the time series in regression analyses, I calculated the regression coefficient using the estimated generalized least square (EGLS) method and computed the bootstrapped (1000 times) 95% confidence limits for the hypothesis test (Ives and Zhu 2006). In addition, I examined the boundary of the distribution relative to the environmental variables. I only examined the minimum latitude (5th percentiles) as the southern boundary for cold water species and the maximum latitude (95th percentiles) as the northern boundary for warm water species, respectively, because the Japan Sea is near the southern most limit for most of the cold water species and the northernmost limit for warm water species. In addition, the dynamics at the range boundaries (latitudinal and elevation in terrestrial ecosystem) are expected to be more sensitive to climate variability (Parmesan and Yohe 2003). For the longitudinal boundary, both the minimum and maximum longitudes (with the same calculation as in latitudinal boundaries) were used in analysis. I included abundance in the analysis and examined the partial correlations between the distribution center (or boundary) and environmental variables if a species’ distribution center (or boundary) was significantly correlated with abundance (P < 0.05), because the geographical extent of marine populations may be correlated with their population size (MacCall 1990, Hsieh et al. 2010a). The aim was to separate the possibility that the distributional shift was caused solely by expansion/constriction due to changes in population size.
However, our results were qualitatively the same when abundance was not included as a covariate. For the time series regression analyses, I set the significance level at 5%, without correcting for multiple tests. My analyses aimed to explore potential climate effects, but various climate indices are correlated (Table 2) thus different tests should not be considered to be independent.
At the decadal scale, to examine the change in the geographical distribution of each taxon from the cold to warm period, I estimated the centroid for each period based on the time series of the distribution centers (the median latitude and longitude were considered simultaneously) by 50% convex-hull peeling (Zani et al. 1998), where all of the distribution centers were weighted equally. This method is robust to the bias caused by potential outliers. I then tracked the movement direction and magnitude of each species from the cold to warm period to determine the decadal variation in the distribution centers. Finally, I used an ANOVA-like randomization test (Hsieh et al.
2008a) to examine whether the shift in centroids between the cold and warm periods was statistically significant for each species.
I then considered whether differences in geographic affinity and life history traits existed between geographically shifting and non-shifting species. I defined the
“geographically shifting species” as those species that exhibited significant distributional shifts at interannual or decadal scales (changes in their distributional domain from the cold to the warm period). For the interannual scale, I analyzed the latitudinal and longitudinal shifts separately. The “geographical affinity” was defined as the cold or warm water group (see Table 2-1) based on studies of a species’
biogeography (Nishimura 1968). The life history traits comprised age at maturation (Am) and asymptotic length (Linf) because these traits were available for all of the species examined. Univariate and multivariate logistic regressions (the latter with stepwise forward selection based on AIC) were used to determine factors that are able to classify the different responses (shift or non-shift) in geographical distributions. The goodness-of-fit was evaluated by AIC, and then Rao's score test (Rao 1948) was used to test whether the value of regression coefficient is significant. Rao test is asymptotically equivalent to the likelihood ratio and Wald test, but it is known to be useful for testing
doi:10.6342/NTU201802401
!
!
!
12!
the improvement of model fit if variables that are currently omitted are added to the model. Therefore, Rao test is often recommended for identifying variable in stepwise forward selection.
Result
The interannual variations in the distribution centers of the demersal species were largely related to environmental variables (Table 2-3). The median latitude and longitude were correlated with environmental variables for more than half of the species, and a limited number of species exhibited distributional shifts only in their boundaries.
The rate of latitudinal shift (center) ranged from –54 to 44 km/°C with an overall average of 5.110 ± 6.617 (standard error) km/°C. Similarly, the longitudinal shift (center) ranged from -45 to 52 km/°C with an overall average of 6.281 ± 6.565 (standard error) km/°C.
Approximately 68% of the species exhibited a significant distributional shift from the cold to the warm period at the decadal scale (Table 2-3), but the movement direction varied among species (Figure 3). Four species in the cold water group exhibited significant poleward shifts, particularly Gadus macrocephalus, Theragra chalcogramma, and Squalus acanthias (Figure 2-3). However, these poleward shifts did not prevail in warm water species. The most significant poleward shift occurred in Glossaodon semifasciatus, whereas Paralichthys olivaceus, Pagrus major, and Evynnis japonica exhibited significant equatorward shifts among the warm water species (Figure 2-3b).
The results of the logistic regression showed that none of the variables we examined (life history traits and geographical affinity) could classify the different distributional responses (shifting v.s. nonshifting) among the species at the interannual
scale (Table 2-4a). At decadal scale, the Linf was significant (P = 0.042) for classifying the distributional responses among the species (Table 2-4a). The results of the multivariate logistic regression were quantitatively the same as the univariate analysis results; thus, only the latter are shown.
Discussion
In this study, I examined the changes in the geographical distribution of the demersal fish assemblage in the Japan Sea and found significant effects of climate. The responses in the distribution in terms of latitudinal and longitudinal shifts at the interannual scale were found in both the cold water and warm water groups. The configuration of the narrow continental shelf in the Japan Sea runs from the northeast to the southwest (Figure 2-1), which may suggest that the overall trend in the along-shore movement is related to environmental variability. However, the rate of distributional shift in the Japan Sea was smaller than that found in a previous study (Poloczanska et al.
2013) and it exhibited a high variance. This also suggests that the universal poleward shift detected by a global-scale meta-analysis might not hold at the regional scale (see Perry et al. (2005)). One may argue that a deepening of the vertical distribution is a more universal response to warming than a latitudinal shift for demersal fishes (Dulvy et al., 2008); however, the demersal fishes surveyed during the warming period did not reveal evidence of deepening in the Japan Sea (Kawamura 2009).
The distributional shifts at the decadal scale also exhibited along-shore movement, but the scales varied among different species (Figure 2-3). In contrast to a previous study (Perry et al., 2005), I found that species with a large body size (Linf) were more likely to respond by changing their distribution (Table 2-4). It is unclear why I reached the opposite conclusion, but Morita et al. (2010) provided a possible explanation. Using
doi:10.6342/NTU201802401
!
!
!
14!
a simple bioenergetic model, they showed that the optimal temperature for growth decreases with increasing body size, which indicates that species with large body sizes may be more vulnerable to warming. It is also possible that fishing may affect the species distributional response to climate change. The large species in the Japan Sea, such as G. macrocephalus, S. acanthias, and P. olivaceus, are important targets of fisheries, thus these large species may experience a higher rate of exploitation than other species. These effects of fishing on the spatial distribution of marine fish were also found in the California Current region, where the exploited species were more sensitive to warming (Hsieh et al. 2008a).
I observed variations in the distributional shift among species at both interannual and decadal scales. It is possible that the different physiological requirement of species or even multiple stocks within one species may have restricted their distribution. In the Japan Sea, Arctoscopus japonicus and P. olivaceus are known to comprise two separate stocks. For A. japonicus, the recent catch statistics for all fisheries combined suggest that the northern stock has been depleted since 1980 (Fujiwara et al. 2009) whereas the southern stock remains relatively stable (Matsukura et al. 2014). Similarly, the northern stock of P. olivaceus was listed as unsustainable in a recent stock assessment report (Uehara et al. 2014) whereas the southern stock was listed as sustainable (Nakagawa et al. 2014). This north–south difference may explain the southeastward movement of P.
olivaceus, as shown in Figure 3. In addition to the presence of multiple stocks of certain species in the Japan Sea, other fisheries such as small-scale trawling in the coastal area of Kyoto Prefecture also target the single-trawling species. This multi-fisheries scenario may generate regional differences among species and stocks. These regional differences could also indicate differences in the responses to climate or fishing.
In addition, several other processes can generate various responses in the
distributional shift among species (Chen et al. 2011, Hollowed et al. 2013). In addition to environmental factors and primary production, which are often linked with the recruitment success (e.g., Beaugrand et al. (2003a)), species interactions (e.g.
competition and predation) may also play roles in mediating the response to climate change. For flatfish, the competition for habitat can affect the success of recruitment during the post-larval stages. Some studies indicate that the quantity of available habitat is crucial for juvenile settlement and it also affects recruitment (Gibson 1994, van der Veer et al. 2000). Thus, interspecific density dependence (Hixon and Jones 2005) may become important when suitable habitats are limited. For demersal species in the Japan Sea, predation or cannibalism by adult individuals may also contribute to the mortality during the juvenile stages and affect recruitment (Minami 1986, Tominaga and Nashida 1991). In these conditions, climate is not the only driving force that affects the distributions of species. It would be interesting to investigate how these biotic processes mediate the distributional responses of species to climate variability.
The use of a CPUE map to infer the distributions of species may be one of the limitations of our analyses. The fishing effort distribution for single trawl fisheries remained consistent in space and time (see Appendix I), but it is widely known that the fleet dynamics and fishermen-related factors can also affect the CPUE pattern (Branch et al. 2006). It might be possible to examine the fleet dynamics to approximate the behavior of fishermen as a Levy flight process, which is a stochastic process that is used commonly to study the foraging behavior of predators (Viswanathan et al. 1996). It is often assumed that price is the key factor that affects the behavior of fishermen, but a case study in North Sea fisheries found that the behavior of fishermen was only marginally correlated with the fish price and fishing effort (Marchal et al. 2007). A recent review also suggests that the economic factor is not the only driver of fleet
doi:10.6342/NTU201802401
!
!
!
16!
dynamics (van Putten et al. 2012). Although I neglected the influence of fleet dynamics on the fishing effort and CPUE in the present study, this issue may be addressed in future studies.
Conclusion
In this study, I provide quantitative evidence of shifts in the distributions and abundances of the demersal fish assemblage in the Japan Sea in response to climate variation. The decadal variations in distributions were explained largely by the asymptotic length, which may suggest that warming has greater negative effects on larger fishes, thereby indicating the possible effects of fishing activities. Not every kind of responses can be classified by life history traits or geographical affinity in our study.
Thus, my findings support previous studies, which showed that life history traits or geographical affinity alone, are not sufficient for interpreting the responses of species to climate variation (Hsieh et al. 2008a, Hsieh et al. 2009b). Furthermore, it would be difficult to project the responses of species to future climate change based on any single factor. More studies of the interactions among species and the biological nonlinear amplification of environmental effects will be necessary to obtain a better understanding of the effects of climate on marine ecosystems.
Table 2-1 Life history traits (age at maturation Am and asymptotic length Linf) and geographical affinity of the target species in the Japan Sea single-trawl fisheries (Tian et al., 2011). The asymptotic lengths are mostly compiled from Ogata (1980) with some exceptions (see note).
*1 FishBase (Froese and Pauly 2013)
*2 It is the main species of "other cold water flounders" (Pleuronectidae spp.) group from JSOBT dataset as in Tian et al. (2011)
*3 Narimatsu et al. (2007)
*4 Also spawn during March-May, but the peak of CPUE is in September-November.
*5 Use maximum length to represent asymptotic length because Linf is unavailable.
*6 Fujioka et al. (1990) 2 Theragra chalcogramma Walleye pollock Cold water 100-500 3 56.1 Dec-Mar
3 Pleurogrammus azonus Arabasque
greening Cold water <200 2 43.5 Sep-Nov
4 Arctoscopus japonicus Japanese
sandfish Cold water 300-500 2 27.8 Dec-Mar 5 Squalus acanthias Piked dogfish Cold water 150-180 10 124.0*1 Feb-May 6 Glyptocephalus stelleri Witch flounder Cold water 200-300 2 58.9 Jan-Apr
7 Hippoglossoides dubius Flathead
flounder Cold water 150-500 5 55.8 Feb-Apr 8 Pleuronectes herzensteini Brown sole Cold water 30-130 2 28.2 Feb-May 9 Microstomus achne*2 Slime flounder Cold water 50-400 3 71.5*1 Feb-Apr
10 Hippoglossoides pinetorum
Pointhead
flounder Warm water 150-190 2 37.0 Jan-Mar 11 Eopsetta grigorjewi Shotted halibut Warm water <140 2 40.8 Feb-Mar
12 Tanakius kitaharai Willowy
flounder Warm water 80-150 2 28.0 Dec-Jan*3
13 Glossanodon
semifasciatus Deepsea smelt Warm water <200 1 25.5*1 Jan-Sep 14 Paralichthys olivaceus Bastard halibut Warm water <150 2 80.7 Mar-Jul 15 Pagrus major Silver seabream Warm water <100 3 54.4 Apr-Jul
16 Evynnis japonica Crimson
seabream Warm water 30-130 2 34.0*1 Jul-Sep 17 Dentex tumifrons Deepsea snapper Warm water <200 2 41.5*1 Sep-Nov*4
seabream Warm water 30-130 2 34.0*1 Jul-Sep 17 Dentex tumifrons Deepsea snapper Warm water <200 2 41.5*1 Sep-Nov*4