Pacific salmon abundance trends and climate change

Abstract

Irvine, J. R., and Fukuwaka, M. 2011. Pacific salmon abundance trends and climate change. – ICES Journal of Marine Science, 68: 1122–1130.

Understanding reasons for historical patterns in salmon abundance could help anticipate future climate-related changes. Recent salmon abundance in the northern North Pacific Ocean, as indexed by commercial catches, has been among the highest on record, with no indication of decline; the 2009 catch was the highest to date. Although the North Pacific Ocean continues to produce large quantities of Pacific salmon, temporal abundance patterns vary among species and areas. Currently, pink and chum salmon are very abundant overall and Chinook and coho salmon are less abundant than they were previously, whereas sockeye salmon abundance varies among areas. Analyses confirm climate-related shifts in abundance, associated with reported ecosystem regime shifts in approximately 1947, 1977, and 1989. We found little evidence to support a major shift after 1989. From 1990, generally favourable climate-related marine conditions in the western North Pacific Ocean, as well as expanding hatchery operations and improving hatchery technologies, are increasing abundances of chum and pink salmon. In the eastern North Pacific Ocean, climate-related changes are apparently playing a role in increasing chum and pink salmon abundances and declining numbers of coho and Chinook salmon.

climate change, ecosystem regime shift, North Pacific Ocean, Pacific salmon, regime change, regime shift, salmon, salmon abundance

Introduction

Pacific salmon (genus Oncorhynchus) are an important ecological and economic species complex in the North Pacific Ocean. Their natural distribution extends from San Francisco Bay in California, northwards along the Canadian and Alaskan coasts to North American and Russian rivers draining into the Arctic Ocean, and southwards along the Asian coastal areas of Russia, Japan, and Korea (Groot and Margolis, 1991).

Understanding reasons for historical patterns in salmon abundance could help anticipate future climate-related changes. Various researchers (Hare and Francis, 1995; Beamish et al., 1999) have hypothe­sized that the North Pacific Ocean alternates between high and low salmon production regimes. Low-frequency, high-amplitude variations in marine ecosystems involving changes in community composition, species abundances, and trophic structure are known as ecosystem regime shifts (Polovina, 2005). Shifts can be caused by climate factors, or anthropogenic factors including fishing, introduced species, and alterations in river discharge (Jiao, 2009). Integral to understanding climate change effects on salmon is an understanding of how variations in salmon abundance correspond to climate-related ecosystem regime shifts.

Statistical approaches to analysing regime shifts have fallen into two basic types: analyses requiring a priori hypotheses that regime shifts happened during particular years and exploratory or data-driven analyses not requiring a priori hypotheses on regime shift timings (Rodionov, 2004). We took the former approach, testing for differences resulting from regime shifts reported to have taken place approximately in 1947 and 1977 (Beamish and Bouillon, 1993; Francis and Hare, 1994; Hare and Francis, 1995), 1989 (Hare and Mantua, 2000), and 1999 (McFarlane et al., 2000; Peterson and Schwing, 2003).

When considering potential climate change effects on salmon, it must be emphasized that the variable(s) being measured to represent salmon should be appropriate to the scale of the biological or geographical unit being assessed. Variables such as salmon abundance, biomass, distribution, growth, and survival can assist in understanding climate effects, but unfortunately, long time-series are uncommon. In this paper, we rely primarily on commercial catch data, because they are the longest, most consistently gathered data available for Pacific salmon. We regard salmon catch to be an appropriate proxy for their abundance, because our assessment units, taxonomic species in the eastern and western North Pacific, are large and we were examining low-frequency variability.

Our primary goal is to assess Pacific salmon abundance trends and review these from a climate change perspective. A secondary goal is to advise others interested in evaluating climate change effects on salmon of recently updated time-series of catch and hatchery release data. We apply relatively simple statistical approaches to estimates of commercial catch of the five major species that reproduce in both the Asian and the North American continents: pink salmon, O. gorbuscha; sockeye salmon, O. nerka; chum salmon, O. keta; coho salmon, O. kisutch; and Chinook salmon, O. tshawytscha. However, we do not consider Chinook or coho salmon in Asia, because directed fisheries on them are rare, records of their catch are inconsistent, and numbers are low.

Abundance proxies and general approach

To index salmon abundance, we used recently revised datasets maintained by the North Pacific Anadromous Fish Commission, the international organization responsible for keeping official records of commercial catch and hatchery production in the North Pacific Ocean (Irvine et al., 2009), which we updated to include preliminary 2009 data. Kasahara (1963) reviewed the early (pre-1961) portion of the catch time-series. Because some Asian statistics were inaccurate and some conversion factors were chosen rather arbitrarily, he cautioned that some estimates were “very rough”. North American catches were more reliable, in part because most salmon were caught within a few miles of the coast. Eggers et al. (2005) and Irvine et al. (2009) updated these time-series, providing additional background information. Estimates since 1961 are generally considered more accurate and precise than earlier estimates.

We followed the approach of Beamish et al. (2004), who assumed the major marine effects would happen in the first ocean year for young Pacific salmon. For our analyses, we aligned regime periods with the year of ocean entry for all species and assumed all pink and coho salmon returned to spawn in the year following spring when they had entered the ocean, whereas sockeye, chum, and Chinook salmon returned 2 years after going to sea (Table 1).

Table 1.

Five regime periods examined (equivalent to ocean entry years) and corresponding catch (return) years for sockeye (Sox), chum, Chinook (Cn), pink, and coho salmon.

		Catch years
Regime	Regime periods	Sox, Chum, Cn	Pink, Coho
1	1925–1946	1925–1948	1925–1947
2	1947–1976	1949–1978	1948–1977
3	1977–1988	1979–1990	1978–1989
4	1989–1998	1991–2000	1990–1999
5	1999–2009+	2001–2009+	2000–2009+

Statistical approach

We assessed whether patterns of abundance for each salmon species varied among regimes within the entire North Pacific, the western North Pacific, and the eastern North Pacific. Salmon abundance in the western North Pacific was indexed by the aggregate commercial catch of Russia, Japan, and the Republic of Korea; in the eastern North Pacific by the combined catches of the United States and Canada; and the entire North Pacific by the sum of all of these catches.

If salmon are more or less abundant in one regime than another, the slope of the line of abundance vs. time, or the mean abundance, or both, likely differs between regimes. First we computed linear regressions for each salmon species, regime (1–5, Table 1), and region (North Pacific, western North Pacific, and eastern North Pacific). Let ln(C_t) be the natural log of catch in millions of salmon in year t for the sequence of N years in each regime:

where m is the slope of the line and b the y-intercept.

The null hypothesis that positive (or negative) correlations between ln(C_t) and t were not greater (or less) than zero was tested using the t-statistic (one-tailed test, α = 0.1). Pyper and Peterman (1998) found that the t-distribution provided a better balance between the effective sample size (N*) and Type 1 error rates (i.e. concluding a correlation to be significant when it is not) than the Chi-squared distribution.

We were concerned about a lack of independence among values within each catch time-series, reasoning that catches could be influenced by catches from earlier years and that this might reduce the effective degrees of freedom substantially below the number of years in the time-series N (Botsford and Lawrence, 2002). We used the approach recommended by Pyper and Peterman (1998; p. 2133, final paragraph) to compute N*, constraining N* to the maximum of the number of years N. N* − 1 effective degrees of freedom (df_ef) were used to evaluate the significance of the one-tailed tests.

In the second approach, we determined whether 95% confidence intervals around mean abundance indices overlapped among various regimes. When intervals did not overlap for a particular set of regimes, we concluded that abundances in these regimes were different. Except for pink salmon, regime means were the arithmetic mean of ln-transformed annual catches. For pink salmon, because of their 2-year life cycle and consistent differences between abundances in odd- and even-numbered years, we calculated the regime means as the average of each successive 2-year average catch before determining confidence limits.

Results

After computing df_ef, many instances remained where the probability (p) of correlations was greater (or less) than zero (Table 2). Table 3 summarizes these results by species within reported regimes and compares mean abundance indices among regimes.

Table 2.

Statistical results from analyses of historical catch time-series by species and regions separated into reported regimes.

	Sockeye salmon				Chum salmon				Pink salmon
Regime	Linear equation	r-value	p-value	dfef	Linear equation	r-value	p-value	dfef	Linear equation	r-value	p-value	dfef
North Pacific
1	y = −0.0138 x + 3.773	−0.348	0.131	10.3	y = −0.0024 x + 3.8119	−0.069	0.439	5.4	y = 0.0013 x + 5.1465	0.035	0.462	8.0
2	y = −0.0007 x + 3.2941	−0.024	0.460	17.2	y = 0.0011 x + 3.6644	0.053	0.397	24.6	y = −0.0051 x + 4.8414	−0.339	0.087	15.6
3	y = 0.0302 x + 3.7354	0.532	0.034	10.6	y = 0.0493 x + 3.8244	0.913	0.011	3.9	y = 0.0083 x + 4.9907	0.309	0.153	11.0
4	y = −0.0734 x + 4.4845	−0.701	0.078	3.7	y = 0.0176 x + 4.397	0.302	0.277	4.1	y = 0.0249 x + 5.3234	0.562	0.036	9.0
5	y = 0.0526 x + 3.6114	0.800	0.047	3.3	y = 0.0081 x + 4.5163	0.326	0.179	8.0	y = 0.0271 x + 5.4142	0.751	0.064	3.4
All	y = 0.0062 x + 3.3542	0.392	0.146	7.2	y = 0.011 x + 3.4898	0.699	0.055	4.3	y = 0.0053 x + 4.8604	0.390	0.186	5.4
Western North Pacific
1	y = −0.0458 x + 2.8134	−0.593	0.073	5.4	y = 0.0012 x + 3.363	0.024	0.479	5.1	y = −0.0028 x + 4.7458	−0.052	0.444	8.0
2	y = 0.0045 x + 1.9721	0.069	0.435	6.1	y = 0.0064 x + 3.301	0.224	0.132	24.6	y = −0.0023 x + 4.4146	−0.108	0.367	10.4
3	y = 0.0106 x + 1.3793	0.220	0.235	11.0	y = 0.0544 x + 3.5298	0.902	0.007	4.0	y = −0.0174 x + 4.4515	−0.601	0.054	6.3
4	y = 0.0108 x + 1.7971	0.200	0.277	9.0	y = 0.0049 x + 4.2048	0.073	0.441	4.6	y = 0.0524 x + 4.5038	0.764	0.019	5.6
5	y = 0.0433 x + 1.935	0.622	0.063	5.3	y = 0.0116 x + 4.2604	0.415	0.117	8.0	y = 0.0492 x + 4.6964	0.694	0.080	3.7
All	y = −0.0063 x + 2.2754	−0.299	0.181	9.4	y = 0.0135 x + 3.075	0.731	0.029	5.2	y = 0.0018 x + 4.5013	0.138	0.338	9.6
Eastern North Pacific
1	y = −0.0059 x + 3.351	−0.152	0.239	22.1	y = −0.0063 x + 2.7013	−0.264	0.101	23.0	y = 0.0037 x + 4.0678	0.135	0.333	10.7
2	y = 0.0001 x + 2.8803	0.004	0.492	29.0	y = −0.0111 x + 2.3639	−0.279	0.150	13.8	y = −0.0117 x + 3.741	−0.328	0.116	13.0
3	y = 0.032 x + 3.6349	0.536	0.035	10.2	y = 0.0301 x + 2.4404	0.350	0.129	10.3	y = 0.0356 x + 4.1205	0.663	0.042	6.0
4	y = −0.0842 x + 4.4235	−0.720	0.072	3.6	y = 0.0576 x + 2.6866	0.817	0.021	4.4	y = −0.0071 x + 4.7675	−0.153	0.326	9.0
5	y = 0.0563 x + 3.3904	0.713	0.074	3.6	y = −0.0052 x + 3.0219	−0.091	0.401	8.0	y = −0.0059 x + 4.7764	−0.195	0.307	7.1
All	y = 0.0095 x + 2.9331	0.471	0.128	5.7	y = 0.0051 x + 2.3375	0.310	0.233	5.9	y = 0.0104 x + 3.657	0.493	0.137	5.0
	Chinook salmon				Coho salmon
	Linear equation	r-value	p-value	dfef	Linear equation	r-value	p-value	dfef
Eastern North Pacific
1	y = −0.0015 x + 1.2109	−0.076	0.360	23.0	y = 0.0039 x + 1.8827	0.150	0.242	22.0
2	y = 0.0044 x + 1.0695	0.286	0.261	5.3	y = 0.0073 x + 1.8809	0.283	0.150	13.4
3	y = −0.0102 x + 1.1891	−0.206	0.273	9.0	y = 0.0164 x + 2.1914	0.406	0.085	11.0
4	y = −0.0563 x + 0.7562	−0.839	0.017	4.2	y = −0.1154 x + 2.6916	−0.821	0.037	3.5
5	y = −0.1352 x + 1.0437	−0.676	0.093	3.5	y = −0.0233 x + 1.7976	−0.495	0.119	5.5
All	y = −0.01 x + 1.4168	−0.640	0.077	4.4	y = 0.0003 x + 1.978	0.022	0.472	10.6

Linear equations are given along with the correlation coefficient (r), the probability (p) that the correlation was greater (or less) than 0, and the effective degrees of freedom (df_ef).

Table 3.

Statistical significance of correlations (r) and mean abundances (⁠⁠) for each species and reported regime.

	Sockeye		Chum		Pink
Regime	r		r		r
North Pacific
1	0	b	0	a	0	b
2	0	a	0	a	−	a
3	+	c	+	b	0	b
4	−	c	0	c	+	c
5	+	c	0	c	+	c
Western North Pacific
1	−	d	0	a	0	b
2	0	cd	0	a	0	a
3	0	a	+	b	−	a
4	0	bc	0	c	+	b
5	+	d	0	c	+	b
Eastern North Pacific
1	0	b	0	b	0	b
2	0	a	0	a	0	a
3	+	c	0	b	+	c
4	−	c	+	c	0	d
5	+	c	0	c	0	d
	Chinook		Coho
	r		r
Eastern North Pacific
1	0	b	0	b
2	0	b	0	b
3	0	b	+	c
4	0	a	−	abc
5	−	a	0	a

Positive and negative correlations greater than and less than 0 (p< 0.10), respectively, are indicated by + or −; correlations not different from 0 are indicated by zero. Mean species abundances followed by the same letter(s) did not differ significantly from each other. Letters indicate the magnitude of abundance, with the letter “a” indicating regimes with the lowest abundance for that species and region, the letter “b” indicating regimes with the next to lowest abundance for that species and region, etc.

Aggregate North Pacific commercial catches in numbers and weights illustrate a period of increasing abundance from the 1920s until the early 1940s (Figure 1). Catches declined near the end of World War II and remained modest until the late 1970s, whereafter they generally increased. Significant differences within and between regimes 1 and 2 were uncommon (Table 3). During regime 1, correlations were never different from zero, and during regime 2, the only significant correlation was decreasing pink salmon abundance. Mean abundances of pink and sockeye salmon declined between regimes 1 and 2. Significant positive trends were observed for sockeye and chum salmon during regime 3, but not for pink salmon; all three species were more abundant in regime 3 than regime 2. Significant positive trends in abundance were observed during regimes 4 and 5 for pink salmon and regime 5 for sockeye salmon. A lack of differences between regimes 4 and 5 for pink and chum salmon abundances does not support the existence of the regime shift reported in 1999 (Table 3). The highest catches on record in the North Pacific Ocean happened during 2009, when ∼609 million salmon were caught (∼1.1 million metric tonnes). Catches in odd-numbered years generally exceeded those of even-numbered years, because of differences in abundance between odd and even years of the most abundant species in the catch, pink salmon (Figure 1).

Figure 1.

North Pacific commercial fishery catch of sockeye, pink, and chum salmon, 1925–2009, in millions of fish (upper panel) and thousands of tonnes (lower panel). Vertical lines correspond to ocean catch years for pink salmon associated with reported regime shifts, demarcating regimes 1–5 as indicated.

Interestingly, the ranking of the species in the overall North Pacific catch was the same during each regime. Pink salmon catches were always highest, followed by chum, sockeye, coho, and Chinook salmon.

In the western North Pacific, chum and pink salmon dominated catches (Figure 2). The only significant trend during regimes 1 and 2 was a decline in sockeye salmon abundance during regime 1. The mean catch of pink salmon in regime 2 was less than in regime 1; other differences between regimes 1 and 2 were not significant. Chum salmon catches increased during regime 3 and mean abundances exceeded those of regime 2. Pink salmon catches declined in regime 3 and, although sockeye catch trends did not differ from zero, mean catches in regime 3 were lower than in regime 2. Similar to the patterns in the North Pacific, significant positive abundance trends were observed during regimes 4 and 5 for pink salmon and regime 5 for sockeye salmon. Regimes 4 and 5 abundances exceeded those of regime 3 for pink, chum, and sockeye salmon (Table 3). There were no differences between regimes 4 and 5 for chum and pink salmon, implying that the reported regime shift in 1999 was not real. The record pink salmon catches in 2009 (Figure 1) were the result of exceptional catches in Asia (Figure 2), primarily Russia.

Figure 2.

Western North Pacific commercial fishery catch of sockeye, pink, and chum salmon, 1925–2009, in millions of fish (upper panel) and thousands of tonnes (lower panel). Vertical lines correspond to ocean catch years for pink salmon associated with reported regime shifts, demarcating regimes 1–5 as indicated.

We could evaluate all five salmon species in the eastern North Pacific where pink and sockeye salmon dominate the catch (Figures 3 and 4). Correlations during regimes 1 and 2 were not different from zero for any species, although mean catches declined between regimes 1 and 2 for pink, chum, and sockeye salmon (Table 3). Sockeye, pink, and coho salmon abundances increased during regime 3; mean abundances for these species, as well as chum salmon, were higher in regime 3 than regime 2. Abundance patterns varied among species for regimes 4 and 5. Chum and pink salmon were more abundant in regime 4 than in regime 3. Pink salmon abundance trends were not different from zero in regime 4 or 5, chum salmon abundances increased during regime 4, whereas sockeye and coho decreased. Chinook salmon were less abundant during regime 5 than in earlier years and abundances declined during regime 5. In general, differences during regime 5 were inconsistent.

Figure 3.

Eastern North Pacific commercial fishery catch of sockeye, pink, and chum salmon, 1925–2009, in millions of fish (upper panel) and thousands of tonnes (lower panel). Vertical lines correspond to ocean catch years for pink salmon associated with reported regime shifts, demarcating regimes 1–5 as indicated.

Figure 4.

Eastern North Pacific commercial fishery catch of Chinook and coho salmon, 1925–2009, in millions of fish (upper panel) and thousands of tonnes (lower panel). Vertical lines correspond to ocean catch years for coho salmon associated with reported regime shifts, demarcating regimes 1–5 as indicated.

Discussion

Not surprisingly, temporal salmon abundance patterns varied among species and regions. High catches in the western North Pacific during the 1930s were apparently real (Klyashtorin and Smirnov, 1995), despite concerns of the validity of some early salmon catch estimates (Kasahara, 1963). Reduced catches in the western North Pacific near the end of World War II likely resulted from, at least in part, reduced effort and reporting. Statistical differences between regimes 1 and 2 would have been greater if we had ignored catches during 1945–1949 in our analyses.

Our general finding of lesser abundances in regime 2 than regimes 1 and 3 corroborates earlier findings of Francis and Hare (1994), Beamish and Bouillon (1993), Hare and Mantua (2000), and others. Regime shifts approximately in 1949, 1977, and 1989 were apparently climate-related. Understanding mechanisms for subsequent abundance changes is complicated by the generally increasing numbers of salmon released from hatcheries (Beamish et al., 1997).

In the western North Pacific, additional releases of hatchery fish have contributed to recent increased catches of pink and chum salmon, but so also have improvements to hatchery technology resulting in the release of fish in better condition with a greater likelihood of survival. For instance, an early emphasis in Russia on releasing large numbers of pink and chum salmon fry changed in the mid-1980s to releases of smaller numbers of healthier fry (Dushkina, 1994). During the past 15 years, releases of chum and pink salmon from Russian hatcheries gradually increased (Irvine et al., 2009), again with an emphasis on improved fish quality (Kaev and Ignatiev, 2007). Japanese hatcheries focus on chum salmon; numbers of chum salmon fry released increased rapidly in the late 1970s, but have been relatively stable over the past 25 years. Korea also produces chiefly chum salmon, but releases relatively few fish (Irvine et al., 2009).

Hiroi (1998) and Kaeriyama (1998) reported increasing return rates (adult catches/juveniles released) for Japanese hatchery chum salmon from Hokkaido and Honshu, respectively, from the late 1970s through the early 1990s. Saito and Nagasawa (2009) found that increased return rates for chum salmon from two regions in Hokkaido were related, at least in part, to increases in the size of fry released. Morita et al. (2006a) attributed increases in coastal catches of chum and pink salmon to various factors, including improvements in hatchery technologies and climate change. However, Morita et al. (2006b) concluded that recent increases in Japanese pink salmon catch were largely explained by climate change. Regardless of the precise mechanism responsible for improved survival, the benefits from enhancement for the salmon catch in Asia have been significant.

In the eastern North Pacific, all five species are enhanced, with pink and chum salmon the major species in Alaska, sockeye and chum in Canada, and Chinook to the south. Hatchery releases in Alaska increased rapidly in the 1970s and the 1980s and since then have been relatively constant. In Canada, after peaking in the early 1990s, release numbers declined over the next decade and since then have been reasonably constant. In the southern United States, hatchery production has declined since the mid-1990s (Irvine et al., 2009).

It is clear that many of the salmon in the North Pacific are of hatchery origin (Hilborn and Eggers, 2000; Kaeriyama, 2008). Pink and chum salmon, the most numerous of the species and of which many enhanced, are released to the ocean early in their life when they are small. Therefore, almost all their biomass is accumulated in the marine environment. Consequently, regardless of whether wild salmon are being replaced by hatchery salmon, our results indicate that the ocean is producing more salmon biomass than previously. A better understanding of density-dependent effects and interactions among species and wild and hatchery salmon are needed to be confident about future climate-related effects on wild Pacific salmon.

We did not attempt to convert our ocean abundance indices to run-size biomass as Eggers (2009), Kaeriyama et al. (2009), and Fukuwaka et al. (2010) have done recently. To the best of our knowledge, biomass time-series have not been developed for coho and Chinook salmon at the scale of the eastern North Pacific, and developing appropriate catch models to do so was beyond the scope of our study. Run-size biomass estimates include both fish that contribute to the next generation (i.e. spawners) and fish that do not (i.e. catch). If exploitation rates vary, so will the portion of the total biomass that contributes to the next generation, confounding the interpretation of total biomass time-series. Our approach assumed that commercial catch was proportional to overall abundance. We felt this was a reasonable assumption at the scale with which we are working. Because almost all the biomass of adult salmon is accumulated in the ocean, our estimates of catches in metric tonnes should be reasonable indices of biomass.

Climate effects on salmon can be assessed at various scales, including species, regional stock groupings, and biological populations within stock groupings. In this study, species of Pacific salmon were assessed separately within the eastern and the western North Pacific Ocean, using catch trends as proxies of abundance. Aggregate catch statistics appeared to be reasonable abundance proxies at this scale, but would have masked differences with smaller units. When one wishes to assess smaller units of salmon, parameters other than catch should be examined, such as spawner numbers and estimates of survival or recruits-per-spawner. Appropriate salmon status indicators are scale-dependent.

In summary, our analyses confirm major climate-related changes in abundance associated with reported ecosystem regime shifts approximately in 1947, 1977, and 1989. There is little evidence to support a shift after 1989, although species abundance patterns varied during this recent period. The production of pink and chum salmon has continued to increase overall since 1990; coho and Chinook salmon in the eastern North Pacific have done relatively poorly, whereas sockeye salmon abundance patterns have been variable. Recent increases in pink and chum salmon are only partly the result of enhancement, because only in Russia have there been recent increases in hatchery releases, whereas catches in Russia and elsewhere have increased. Coho and Chinook declines after 1989 are partly explained by reduced effort in response to conservation concerns, but evidence of declining marine survivals for wild coho (Hobday and Boehlert, 2001; Beamish et al., 2008) indicate a likely climate effect as well.

Acknowledgements

We thank other members of the NPAFC Stock Assessment Working Group for gathering and providing national datasets, Lana Fitzpatrick for assistance with data assembly, and two reviewers for constructive comments.