SEARCH

SEARCH BY CITATION

Keywords:

  • climate;
  • cod;
  • Irish Sea;
  • recruitment success;
  • spawning stock biomass;
  • threshold dynamics

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

How climatic variability and anthropogenic pressures interact to influence recruitment is a key factor in achieving sustainable resource management. However, the combined effects of these pressures can make it difficult to detect non-stationary interactions or shifts in the relationships with recruitment. Here we examine the links between climate and Irish Sea cod recruitment during a period of declining spawning stock biomass (SSB). Specifically, we test for a shift in the relationship between recruitment, SSB and climate by comparing an additive (generalized additive model, GAM) and non-additive threshold model (TGAM). The relationship between recruitment success, SSB and the climatic driver, sea surface temperature, was best described by the TGAM, with a threshold identified between recruitment and SSB at approximately 7900 t. The analysis suggests a threshold shift in the relationship between recruitment and SSB in Irish Sea cod, with cod recruitment being more sensitive to climatic variability during the recent low SSB regime.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Atlantic cod (Gadus morhua) has declined over much of its geographical range, owing to a combination of overfishing and declining recruitment (Brander, 2005; Lilly et al., 2008; ICES, 2011). The Irish Sea cod stock has undergone a 10-fold decline in spawning stock biomass (SSB) during the last 30 yr, with below-average recruitment estimated for nearly two decades (ICES, 2011). As a result of this overexploitation the stock has exhibited a decline in age structure and diversity since the 1960s (Marteinsdottir et al., 2005; ICES, 2011). Meanwhile the Irish Sea has exhibited a warming trend, most extensively in the last decade (Cannaby and Hüsrevoğlu, 2009). Against this backdrop of declining SSB and climatic variability, recruitment in Irish Sea cod was shown to be significantly negatively correlated with the North Atlantic Oscillation (NAO) index and sea surface temperature (SST) (Planque and Fox, 1998; Planque and Fredou, 1999; Brander and Mohn, 2004). More recent analysis has shown that these relationships between climatic pressures and recruitment may be non-stationary and dependent on SSB, demographic changes in SSB or environmental change (Brander, 2005; Ottersen et al., 2006; Stige et al., 2006). To further understand the relationship between recruitment, SSB and climate it is therefore necessary to account for non-stationary effects, including abrupt threshold-like shifts in these relationships (Ciannelli et al., 2004).

Many studies have demonstrated links between climate variability and recruitment of cod stocks (Planque and Fredou, 1999; Köster et al., 2005; Stige et al., 2006). Possible mechanisms include a match-mismatch between the timing of reproduction and predator/prey production and/or the connectivity between spawning and nursery areas via variations in the retention and transport processes (Beaugrand et al., 2003; Rijnsdorp et al., 2009). A reduction in cod SSB may act to increase the sensitivity of the stocks to climatic variability by reducing the spatial and temporal ‘window’ of spawning (Brander, 2005). At the ecosystem level, a reduction in biomass of top predator species (e.g., cod) can induce trophic restructuring, leading to threshold-like shifts between alternative stable states (Casini et al., 2008, 2009; Llope et al., 2011). These new alternative states can lead to conditions that are then unfavourable for the recovery of the top predator (Bundy and Fanning, 2005; Casini et al., 2009).

The Irish Sea is a semi-enclosed body of water connected to the northwest European continental shelf through a north and south channel (Fig. 1). Cod spawning grounds are located at two main sites in the western and eastern Irish Sea (Brander, 1994; Fox et al., 2000). Historical variability in the relative importance of these sites is unknown, with recent egg production estimates suggesting a roughly equal distribution of SSB between them in some years (Armstrong et al., 2011). Spawning occurs between late January and May, with peak spawning in the western Irish Sea observed between late February and early April (Fox et al., 1997, 2000; ICES, 2011). The western Irish Sea spawning sites are located inshore of a seasonal near surface gyre that develops during the late spring and is associated with the retention of crustacean and fish larvae (Dickey-Collas et al., 1996, 1997; Hill et al., 1996) (Fig. 1). Thermal stratification and freshwater input have a strong influence on the regional circulation and vertical density of the water column (Gowen et al., 1995; Horsburgh et al., 2000).

image

Figure 1. Irish Sea study area. Contours mark seasonally stratified zones, with darker shading representing relative strength of thermal stratification. Location of Methot-Isaacs-Kidd Survey (MIK) stations (circles) and AFBI groundfish survey stations (crosses).

Download figure to PowerPoint

In this paper we examine the relationship between climate and Irish Sea cod recruitment variability in concomitance with a declining SSB, and test whether this relationship follows a threshold dynamic. We compare the results from methods that assume a continuous relationship between recruitment and the predictors (generalized additive model, GAM) with those that allow a shift in the relationship between the dependent and independent variables (non-additive threshold model, TGAM).

Materials and Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Irish cod stock spawning biomass and recruitment

Estimates of recruitment (age 0 in number of individuals) and SSB (tonnes) were available from a virtual population analysis (VPA) fitted on catch and survey data for ICES area VIIaN (Irish Sea) from 1968 to 2009 (ICES, 2010) (Table 1).

Table 1. Data series used in this study: spawning stock biomass in tonnes (SSB), recruitment at age 0 (in thousands), recruitment success (Rs), annual mean sea surface temperature anomaly (SSTAnnual), spawning time SST anomaly (SSTSpawning), annual rainfall index (RainAnnual) and spawning time rainfall index (RainSpawning). The station-based winter (December–March) North Atlantic Oscillation Index (NAO)
YEARSSBRecruits R s SSTAnnualSSTSpawningRainAnnualRainSpawningNAO
196813 4446512−0.720.070.02−9.11−16.39−1.04
196912 2418506−0.36−0.02−0.66−86.81−16.54−4.89
1970978515 1310.44−0.14−0.2381.0923.71−1.89
197111 2715239−0.770.100.27−156.61−11.19−0.96
197215 87313 883−0.13−0.200.39−92.316.860.34
197320 2273107−1.870.110.78−124.11−24.592.52
197418 12111 055−0.49−0.040.7922.99−13.941.23
197517 8863533−1.620.260.03−167.81−34.741.63
197613 6475103−0.980.28−0.09−26.41−13.391.37
197712 6735529−0.83−0.020.19−3.2135.36−2.14
1978866212 0820.33−0.20−0.2927.9911.810.17
197910 42614 1960.31−0.58−0.8137.99−18.09−2.25
198012 3107923−0.440.050.3591.6921.810.56
198118 3173461−1.67−0.080.21170.5914.412.05
198220 2495264−1.350.040.1669.1916.360.8
198315 2607879−0.66−0.25−0.68−100.114.113.42
198411 2497922−0.35−0.02−0.40−34.3111.861.6
198512 0556350−0.64−0.61−0.6149.69−17.64−0.63
198612 02618 4420.43−0.78−1.38123.29−20.390.5
198712 9958743−0.40−0.49−1.27−90.514.11−0.75
198813 4923803−1.27−0.04−0.12100.8936.810.72
198914 3004904−1.070.891.20−143.3118.665.08
199087255648−0.430.671.09138.2934.563.96
1991653187510.290.07−0.44−60.311.211.03
199272311709−1.440.160.75108.7939.163.28
199362955110−0.21−0.34−0.3172.99−35.292.67
199459953699−0.48−0.17−0.2080.4946.813.03
199545753121−0.380.780.41−9.0130.863.96
1996574757930.010.13−0.22−40.51−12.79−3.78
199756142105−0.980.780.64−77.415.96−0.17
19984810881−1.700.531.3374.39−16.240.72
1999491856560.140.630.6096.79−12.641.7
2000204039850.670.650.7447.69−6.342.8
2001324246520.360.280.02−227.91−19.59−1.9
200261971234−1.610.780.88174.8922.810.76
200344052074−0.750.770.42−189.51−27.290.2
200441401269−1.180.640.59−117.01−19.44−0.07
200526901491−0.590.740.29−96.41−21.290.12
200627571236−0.800.970.4235.195.31−1.09
20071658384−1.461.281.4111.29−3.892.79
20081801574−1.140.520.84140.6911.062.1
2009119237421.140.490.2766.59−36.14−0.41

Indices of abundance for pelagic juvenile cod were also available from the Methot-Isaacs-Kidd (MIK) Net Survey from 1994 to 2009. The survey uses an MIK frame trawl with 5-mm mesh (Methot, 1986) to target pelagic stages of juvenile gadoids at fixed stations in the western Irish Sea (Fig. 1). The survey is stratified and takes place during late May and early June, coincident with the period prior to demersal settlement of pelagic juvenile gadoids.

Indices of 0-group cod were available from October groundfish survey from 1994 to 2009. The survey is carried out using a rockhopper otter trawl at fixed station positions (Fig. 1). Survey designs are stratified by depth and seabed type, with effort concentrated in the western Irish Sea.

Environmental variables

Monthly sea surface temperature (SST) values were downloaded from the U.K. Met Office Hadley Centre HadSST2 data set (http://hadobs.metoffice.com/hadsst2/) . The data are based on quality-controlled, in situ measurements of SST from ships and buoys: data for the period 1968–97 are taken from the International Comprehensive Ocean-Atmosphere Data Set, (ICOADS), and those from 1998 onwards are from the US National Oceanic and Atmospheric Administration (NOAA) National Centers for Environmental Prediction–Global Telecommunications System (NCEP–GTS). SST anomalies (hereafter referred to as SST) were determined by subtraction of the calculated climatology for 1970–2000 before spatially averaging over a grid of 1°× 1° or 5°× 5° (Rayner et al., 2006).

Monthly Northern Ireland precipitation volumes (mm) were downloaded from the U.K. Met Office Hadley Centre HadUKP data set for Northern Ireland (http://www.metoffice.gov.uk/hadobs/hadukp/). This data set incorporates a selection of long-running rainfall stations to provide a series of area averaged precipitation (Alexander and Jones, 2000).

The station-based winter (December–March) NAO index was downloaded from the Climate Analysis Section, NCAR (Boulder, CO, U.S.A.) (Hurrell, 1995). The index is based on the difference of normalized sea level pressure between stations based in Portugal and Iceland. The NAO can be used as a proxy for temperature, wind and precipitation over the North Atlantic and northwest European shelf seas (Brander and Mohn, 2004).

Analysis

Data used in the analyses are listed in Table 1. We used recruitment success (Rs) as the response, calculated as the natural logarithm of the ratio of R to SSB (Cardinale et al., 2008). This has been advocated as a robust way of investigating the effects of SSB and climate on recruitment (Stige et al., 2006; Cardinale et al., 2008). The use of Rs enables the SSB effect to be scaled out, therefore allowing us to disentangle the effects of SSB from climatic variability (Cardinale and Hjelm, 2006). The negative correlation between Irish Sea cod recruitment and SST is well documented; however, the mechanism behind the correlation remains unknown (Planque and Fox, 1998; Planque and Fredou, 1999). To investigate further the role of climate and its interaction with local oceanographic conditions we incorporated rainfall in the analysis. Rainfall was included as an indicator of freshwater runoff to the western Irish Sea coastal region associated with the density structure and a southerly current during the early part of the year (Hill et al., 1996; Horsburgh et al., 2000). For SST and rainfall, annual and peak spawning period (February–March) averages were calculated from data standardized to the time series mean.

To provide an indication of when recruitment level was generated in Irish Sea cod, a correlation matrix was constructed. The mean standardized abundance of cod, ranging from pelagic juvenile to age 0 (October), was correlated with annual recruitment estimates (age 0) (Table 2). A correlation matrix between all environmental variables Rs and SSB was constructed prior to further analysis to check for collinearity (Table 3). To avoid collinearity, we only used predictors that correlated with each other by r < 0.70 (Heinänen et al., 2008; Cardinale et al., 2009), retaining the one that explains the largest part of the variance in case of r > 0.70 between two different predictors.

Table 2. Correlation matrix of annual Irish Sea cod recruitment (age 0) (Recruits) and annual survey abundance indices. Pelagic juvenile cod estimated from western Irish Sea MIK net survey, 1994–2009 (MIK 0-gp) and October groundfish survey estimate of 0-group cod, 1994–2009 (Oct GFS 0-gp)
 MIK 0-gpOct GFS 0-gp
Oct GFS 0-gp0.77 
Recruits0.630.82
Table 3. Correlation matrix of data series used in this study: spawning stock biomass in tonnes (SSB), recruitment at age 0 (in thousands), recruitment success (Rs), annual mean sea surface temperature anomaly (SSTAnnual), spawning time SST anomaly (SSTSpawning), annual rainfall index (RainAnnual) and spawning time rainfall index (RainSpawning). The station-based winter (December–March) North Atlantic Oscillation Index (NAO)
 SSB R s SSTAnnualSSTSpawningRainAnnualRainSpawningNAO
SSB1      
R s −0.321     
SSTAnnual−0.54−0.281    
SSTSpawning−0.26−0.450.801   
RainAnnual−0.100.01−0.110.091  
RainSpawning0.07−0.180.040.150.421 
NAO0.04−0.290.280.430.200.301

First, the relationship between Rs, SSB and the environmental variables was analysed with a GAM. The GAM assumes additive and stationary relationships between the predictors and the response. Degrees of freedom of the smooth terms were estimated by minimizing the generalized cross validation (GCV) (Wood, 2001, 2004). The GCV of a model is a proxy for the out-of-sample predictive mean squared error. The GCV penalizes a large number of parameters in the model. Therefore, a model with lower GCV has more explanatory power, and hence is preferred, to a model with higher GCV. All hypothesized predictors were included in the model initially, and then a backward stepwise selection based on Akaike's information criterion (AIC) was applied to find the most parsimonious model.

Thus, the final GAM model chosen was:

  • display math(1)

To test for the existence of a threshold (Litzow and Cianneli, 2007) in the response of Rs to the predictors, we applied TGAMs (Ciannelli et al., 2004) to SSB and SST. This kind of non-additive model can be formulated as:

  • display math(2)

where two specific additive formulations are adopted for different regimes of the threshold variable Vx.

TGAMs are an extension of non-parametric regression techniques (Hastie and Tibshirani, 1990) and were chosen for their ability to represent an abrupt change in the relationships between dependent and independent variables (i.e., a threshold) at a specific value. The threshold value was selected minimizing the GCV of the whole model (Ciannelli et al., 2004). The searching algorithm runs the model for 100 possible threshold values between the lower 0.1 and the upper 0.9 quantiles.

Residuals of the TGAMs were analysed to inspect for potential deviation from the normality assumption and other anomalies in the data or in the model fit using graphical methods (quantile–quantile plots, residual distribution plots, auto-correlation functions) (Cleveland, 1993) (see Supporting Information Fig. S1). The normal distribution was used in all models.

As with the GAM, all hypothesized predictors were included in the model initially, and then a backward stepwise selection based on GCV was applied to find the most parsimonious model. This procedure was followed for TGAM models with thresholds in SSB and SST. Two final TGAM models were chosen (Table 4).

Table 4. Results of generalized additive model (GAM) and non-additive threshold model (TGAM) analysis. For each model the following are given: deviance explained (Dev. Expl.), the proportion of variance explained (r2), genuine cross-validation squared prediction error (gCV), the AIC value and number of observations (n). For each predictor, stock spawning biomass (SSB) and spawning time SST anomaly (SSTSpawning), the effective degrees of freedom (ed.f.) and significance value (P) are provided. For TGAMs, threshold value (T) and number of observations above and under the threshold are also given [n(T)]. The best model is indicated in bold
ModelThreshold T Dev. Expl. r 2 gCVAICPredictorsFed.f.Pn (T)
GAM  41.10.380.355 SSB13.61.0<0.001 
SSTSpawning20.41.0<0.001 
TGAM 1 SSB 7926 t 61.0 0.56 0.348 65.61 SSB > T 15.2 1.4 <0.001 23
SSB ≤ T 11.4 1.4 <0.001 19
SST Spawning  > T 5.0 1.4 <0.02 23
SST Spawning  ≤ T 23.6 1.4 <0.001 19
TGAM2SST+0.2853.30.49 69.22SSB ≤ T13.01.7<0.00121
SSTSpawning > T11.71.7<0.00121

TGAM1 tested the effect of an SSB threshold on the relationship, whereas TGAM2 tested the effect of a SST threshold on the relationship. The maximum number of knots (k = 4) was limited for the smoothing term of SSB and SSTSpawning to simplify the interpretation of the results.

Finally, to compare the TGAM and GAM, the genuine cross-validatory squared prediction error (gCV) was calculated. The data cases are deleted one at a time with the response of the deleted case predicted by the examined model (GAM or TGAM) fitted to all the remaining cases. A square prediction error is then calculated, and the same routine is repeated across all data cases. The mean of all squared prediction error is then the gCV for the examined model (Ciannelli et al., 2004). GCV and AIC cannot be applied to the comparison between GAM and TGAM as there is no obvious way how the threshold extra parameter should be penalized. Thus, the model selection strategy that we adopted was that of using the GCV to select first the best GAM and TGAM, and then the gCV for their comparison (Ciannelli et al., 2004). All analyses were performed using the mgcv r library package (http://www.r-project.org) (Wood, 2001).

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Trends in recruitment, SSB and Rs

SSB has shown a decline since the late 1980s, and recruitment followed a similar decline from a peak in the 1980s (Fig. 2). A significant negative regression between Rs and SSB (r2 = 0.10; P < 0.05) was observed, with increased variability in the regression observed at SSB estimates below ~8000 t (Fig. 3). Correlations between the survey abundance indices and annual recruitment estimates (1994–2009) showed the strongest correlation (0.82) between age 0 (October) and recruits (Table 2).

image

Figure 2. Irish Sea cod SSB (tonnes) and recruitment at age-0 (in thousands) from 1968 to 2009 as estimated from virtual population analysis (ICES, 2010). Annual SST anomaly calculated from U.K. Met Office Hadley Centre HadSST2 dataset 1968-2009.

Download figure to PowerPoint

image

Figure 3. Regression (r2 = 0.10; P < 0.05) between Irish Sea cod SSB (tonnes) and recruitment success (Rs) during 1968–2009.

Download figure to PowerPoint

Trends in environmental variables

A correlation matrix between all the environmental variables Rs and SSB was constructed prior to further analysis (Table 3). SSB and all environmental variables except RainAnnual were negatively correlated with Rs. Due to the correlation between SSTSpawning and SSTAnnual (0.80), SSTAnnual was removed from further analysis. None of the correlations between the remaining predictors exceeded 0.70 (i.e., −0.54 between SSB and SSTAnnual), so all remaining predictors were retained for the initial GAM and TGAM backward stepwise selection process.

GAM and TGAM analysis

After the backward stepwise selection process, a general GAM with SSB and SSTSpawing explained the largest part of the deviance and had the lowest AIC. The same predictors were retained in both the SSB and SST threshold TGAM selection processes (Table 4). Results of the basic GAM suggested a negative linear response between Rs, SSB and SST (Supporting Information Fig. S3). SSTSpawning had the strongest effect on Rs, with the final model explaining 41.1% of the total deviance in Rs (Table 4). Analysis of the residuals did not reveal any major departure from the main model assumptions of normality and homogeneity of variance (Supporting Information Fig. S2).

Using the TGAM, thresholds were identified in the relationship between the predictors SSB (7926 t) and SSTSpawning (SST anomaly +0.28, range −1.75 to +1.41) and the response Rs (Supporting Information Fig. S4). For SSB, the minimum was not sharply defined, reflecting the finite number of observations (42 yr) into which the data set could be split and limited SSB observations between 7350 and 8500 t. We therefore selected the midpoint (7926 t) of this minimum range to represent the threshold value (Fig. 4). TGAM1 with a threshold in SSB performed better than TGAM2 with SSTSpawning thresholds (in terms of AIC), suggesting the threshold in SSB had more influence on the response than a threshold in SST (Table 4). Within the TGAM1 model, SSB above the threshold and SST below the threshold had the strongest negative effects on recruitment (Fig. 5). Comparison between the GAM and best TGAM using gCV suggested that TGAM outperformed GAM, explaining 20% more deviance (Table 4).

image

Figure 4. Threshold estimation for SSB. Dotted line represents generalized cross validation (GCV) estimated threshold. Threshold in SSB taken as midway point of GCV minima (dashed line).

Download figure to PowerPoint

image

Figure 5. TGAM1 smooths of the conditioned effects of SSB and SST on the level of Rs at high and low SSB. The y-axis is scaled to zero and reflects the relative importance of the covariate. Rugplot on the x-axis represents observations, dash lines are 95% confidence intervals on the smooths, generated from bootstrapping. 1.4 degrees of freedom (ed.f.) on all axes.

Download figure to PowerPoint

TGAM with a threshold in SST suggested that SSB only had an effect during the low SST regime, and SST was only significant during high SST (Table 4).

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Here we show that climatic pressures influencing Irish Sea cod recruitment are non-stationary, dependent on the level of SSB. Specifically, the threshold analysis suggests a shift in the relationship between recruitment and SSB at a threshold of ~7900 t, with climatic pressures more influential below this threshold. Stock recruitment relationships are traditionally assumed to be continuous, characterized by one specific stock recruitment curve (i.e., Ricker, Beverton-Holt). However, threshold changes between high and low regime states in stock recruitment relationships may occur, due to changes in population, community or ecosystem scale processes (Garvey et al., 2009). We have shown that Irish Sea cod SSB has a significant negative relationship with recruitment success in both regimes (high and low) identified by the threshold. The difference in slope between these regimes suggests a difference in the underlying recruitment dynamics.

It is unclear whether the SSB threshold is evidence of changes in the biological and demographic properties of the Irish Sea cod stock, or related to a shift in the functioning of the Irish Sea ecosystem as a whole. A decline in SSB has been shown to increase the sensitivity of recruitment to climatic variability in cod stocks (Brander, 2005; Ottersen et al., 2006). Older, larger fish have a higher reproductive capacity (Marshall et al., 1998; Marteinsdottir and Thorarinsson, 1998) and spawn over greater spatial (vertical and horizontal) and temporal scales, increasing the dispersal potential of eggs (Kjesbu et al., 1996; Marteinsdottir and Begg, 2002; Secor, 2007). This decreases the dependence of egg and larval survival on a restricted range of biotic and abiotic conditions (Planque et al., 2010), thereby decreasing the sensitivity of recruitment to climatic variability (Begg and Marteinsdottir, 2002; Anderson et al., 2008). Additionally, recruitment variability generally increases in populations reduced to low SSB levels, as the proportional influence of density-independent mortality operating during the early life history stages increases (Myers and Cadigan, 1993; Myers, 2001). The threshold identified in Irish Sea cod SSB may represent a limit at which the proportional influence of density-dependent and density-independent mortality switches. The correlation observed in this study between pelagic juvenile cod indices and annual recruitment estimates suggests that at the current low level of SSB, mortality processes during these early life history stages are important in setting recruitment levels. This is supported by observations from other European cod stocks (Brander, 2005). Strong density-dependent mortality has been detected at the juvenile stage (benthic) in Irish Sea cod using historical time-series (Myers and Cadigan, 1993), during the higher SSB regime. The switch between density-dependent and independent processes may be related to habitat use and availability, whereby during years of low population size, habitat occupancy is affected by abiotic variables such as SST, whereas at large population size density-dependence is the dominating mechanism (Ciannelli et al., 2012).

Alternatively the threshold identified in SSB may represent an ecological threshold for cod in the Irish Sea or the presence of a regime shift. Top down and bottom up processes can generate abrupt threshold-like shifts in ecosystems from one stable state to another (Casini et al., 2008, 2009; Möllmann et al., 2008; Llope et al., 2011). Removal of large predators (e.g., cod) from an ecosystem can have a cascade effect throughout the food web, generating a threshold-like shift between alternative states (Frank et al., 2005; Casini et al., 2009; Llope et al., 2011). Many northern-shelf ecosystems have seen an increase in the abundance of small forage species and a drop in abundance of larger bodied predator fish species (e.g., cod) (Frank et al., 2005; Möllmann et al., 2008; Casini et al., 2009). This predator/prey reversal can led to dispensatory effects in the top predator species due to the increased predation and competition from the forage species (Walters and Kitchell, 2001). The pelagic forage species sprat (Sprattus sprattus) in the Irish Sea has been shown to be an important predator of fish eggs and may be a potential competitor for zooplankton prey with fish larvae (Lynam et al., 2011; Plirú et al., 2012). An increase in forage species abundance may ultimately hinder the recovery of the predator species and the return of the ecosystem to the prior state (Bundy and Fanning, 2005; Casini et al., 2009).

Previous studies of Irish Sea cod have assumed a constant and stationary relationship between recruitment, SSB and SST (Planque and Fox, 1998; Planque and Fredou, 1999). Using a GAM in this study, which assumed a stationary relationship, SST was found to be the major driver of recruitment variability throughout the time-series. However, the GAM was shown to be a less optimal model than the TGAM1. Although this confirms the negative effect of SST on Irish Sea cod recruitment as demonstrated in earlier studies (Planque and Fox, 1998; Planque and Fredou, 1999), the current study highlights the importance of accounting for non-stationary effects in this relationship. Ultimately the Irish Sea cod stock appears more resilient to SST variability at SSB above the threshold. The TGAM2 with SST in the threshold also suggested that SST was the main predictor of recruitment, whereas SSB was only significant during the low SST regime. However, this model was rejected in favour of TGAM1.

The link between SST and recruitment success may result from a range of processes operating at the individual to ecosystem level (Rijnsdorp et al., 2009). Temperature has direct effects on the physiology of larval cod, affecting growth rates and mortality (Campana, 1996). At the population level, variations in SST result in variable spawning times (Kjesbu, 1994; Kjesbu et al., 2010), and possible match/mismatch between prey/predators and/or suitable hydrodynamic conditions required for successful transport to nursery areas (Brander, 2005; Rijnsdorp et al., 2009). For example, SST has been shown to influence larval cod survival and recruitment in the North Sea through the interaction between growth and planktonic prey availability (qualitative and quantitative) (Beaugrand et al., 2003; Olsen et al., 2011).

SST also influences the abiotic environment which is closely coupled to fish recruitment and distribution. The physical oceanography of the western Irish Sea, in particular the seasonal development of the western Irish Sea gyre, is influenced by variability in SST (Olbert et al., 2011). Increases in SST have been shown to reduce the retention potential of the gyre, leading to the possible dispersal and increased mortality of eggs and larvae (Emsley et al., 2005; Olbert et al., 2011).

We have demonstrated the presence of a threshold shift in the relationship between recruitment and SSB in Irish Sea cod, with recruitment more sensitive to climatic variability during the more recent low SSB regime. Further study is required to ascertain whether the SSB threshold represents a shift in the recruitment dynamics of Irish Sea cod at the population or ecosystem level. In terms of management, assumptions of constant relationships between SSB, recruitment and climatic drivers will lead to inaccurate predictions of stock resilience in the face of climatic variability.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

The authors acknowledge the dedicated work of the many scientists who have contributed to the development and maintenance of the data sets used in this study. We also acknowledge the three reviewers whose thoughtful and valuable comments greatly improved the manuscript. The work reported in this paper was partly funded by the Department of Agriculture and Rural Development, Northern Ireland, research contract 0626 ‘Recruitment process and stock dynamics of Irish Sea fin-fish populations’.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information
  • Alexander, L.V. and Jones, P.D. (2000) Updated precipitation series for the U.K. and discussion of recent extremes. Atmos. Sci. Lett. 1:142150.
  • Anderson, C.N., Hsieh, C., Sandin, S.A. et al. (2008) Why fishing magnifies fluctuations in fish abundance. Nature 452:835839.
  • Armstrong, M., Beggs, S., Goodsir, F. et al. (2011) Egg production survey estimates of spawning stock biomass of cod, haddock and plaice in the Irish Sea: 1995–2010. ICES Working Group on Celtic Seas Ecoregion 2011, WD 9, pp. 121.
  • Beaugrand, G., Brander, K.M., Lindley, J.A., Souissi, S. and Reid, P.C. (2003) Plankton effect on cod recruitment in the North Sea. Nature 426:661664.
  • Begg, G.A. and Marteinsdottir, G. (2002) Environmental and stock effects on spawning origins and recruitment of cod Gadus morhua. Mar. Ecol. Prog. Ser. 229:263277.
  • Brander, K.M. (1994) The location and timing of cod spawning around the British Isles. ICES J. Mar. Sci. 51:7189.
  • Brander, K.M. (2005) Cod recruitment is strongly affected by climate when stock biomass is low. ICES J. Mar. Sci. 62:339343.
  • Brander, K. and Mohn, R. (2004) Effect of the North Atlantic Oscillation on recruitment of Atlantic cod (Gadus morhua). Can. J. Fish. Aquat. Sci. 61:15581564.
  • Bundy, A. and Fanning, L.P. (2005) Can Atlantic cod (Gadus morhua) recover? Exploring trophic explanations for the non-recovery of the cod stock on the eastern Scotian Shelf, Canada. Can. J. Fish. Aquat. Sci. 62:14741489.
  • Campana, S. (1996) Year-class strength and growth rate in young Atlantic cod Gadus morhua. Mar. Ecol. Prog. Ser. 135:2126.
  • Cannaby, H. and Hüsrevoğlu, Y.S. (2009) The influence of low-frequency variability and long-term trends in North Atlantic sea surface temperature on Irish waters. ICES J. Mar. Sci. 66:14801489.
  • Cardinale, M. and Hjelm, J. (2006) Marine fish recruitment variability and climate indices. Mar. Ecol. Prog. Ser. 309:307.
  • Cardinale, M., Hjelm, J. and Casini, M. (2008) Disentangling the Effect of Adult Biomass and Temperature on the Recruitment Dynamics of Fishes. Resiliency of Gadid Stocks to Fishing and Climate Change. Alaska Sea Grant College Program, pp. 221237.
  • Cardinale, M., Mollmann, C., Bartolino, V. et al. (2009) Effect of environmental variability and spawner characteristics on the recruitment of Baltic herring Clupea harengus populations. Mar. Ecol. Prog. Ser. 388:221234.
  • Casini, M., Lövgren, J., Hjelm, J., Cardinale, M., Molinero, J.C. and Kornilovs, G. (2008) Multi-level trophic cascades in a heavily exploited open marine ecosystem. Proc. Biol. Sci. 275:17931801.
  • Casini, M., Hjelm, J., Molinero, J.-C. et al. (2009) Trophic cascades promote threshold-like shifts in pelagic marine ecosystems. Proc. Natl Acad. Sci. U S A 106:197202.
  • Ciannelli, L., Chan, K.-S., Bailey, K.M. and Stenseth, N.C. (2004) Nonadditive effects of the environment on the survival of a large marine fish population. Ecology 85:34183427.
  • Ciannelli, L., Bartolino, V. and Chan, K.-S. (2012) Non-additive and non-stationary properties in the spatial distribution of a large marine fish population. Proc. Biol. Sci. 279:36353642.
  • Cleveland, W.S. (1993) Visualizing Data. Summit, NJ: Hobart Press.
  • Dickey-Collas, M., Gowen, R.J. and Fox, C.J. (1996) Distribution of larval and juvenile fish in the western Irish Sea: relationship to phytoplankton, zooplankton biomass and recurrent physical features. Mar. Freshw. Res. 47:169181.
  • Dickey-Collas, M., Brown, J., Fernand, L., Hill, A.E., Horsburgh, K.J. and Garvine, R.W. (1997) Does the western Irish Sea gyre influence the distribution of pelagic juvenile fish? J. Fish Biol. 51:206229.
  • Emsley, S.M., Tarling, G.A. and Burrows, M.T. (2005) The effect of vertical migration strategy on retention and dispersion in the Irish Sea during spring-summer. Fish. Oceanogr. 14:161174.
  • Fox, C.J., Dickey-Collas, M. and Winpenny, A.J. (1997) Spring plankton surveys of the Irish Sea in 1995: the distribution of fish eggs and larvae. Sci. Ser. Tech. Rep. CEFAS, Lowestoft, 104, pp. 106.
  • Fox, C.J., O'Brien, C.M., Dickey-Collas, M. and Nash, R.D.M. (2000) Patterns in the spawning of cod (Gadus morhua L.), sole (Solea solea L.) and plaice (Pleuronectes platessa L.) in the Irish Sea as determined by generalized additive modelling. Fish. Oceanogr. 9:3349.
  • Frank, K.T., Petrie, B., Choi, J.S. and Leggett, W.C. (2005) Trophic cascades in a formerly cod-dominated ecosystem. Science 308:16211623.
  • Garvey, J.E., Wright, R.A. and Marschall, E.A. (2009) Searching for threshold shifts in spawner-recruit data. Can. J. Fish. Aquat. Sci. 66:312320.
  • Gowen, R.J., Stewart, B.M., Mills, D.K. and Elliott, P. (1995) Regional differences in stratification and its effect on phytoplankton production and biomass in the northwestern Irish Sea. J. Plankton Res. 17:753769.
  • Hastie, T. and Tibshirani, R. (1990) Generalized Additive Models. New York: Chapman & Hall.
  • Heinänen, S., Rönkä, M. and Von Numers, M. (2008) Modelling the occurrence and abundance of a colonial species, the arctic tern Sterna paradisaea in the archipelago of SW Finland. Ecography 31:601611.
  • Hill, A.E., Brown, J. and Fernand, L. (1996) The western Irish Sea gyre: a retention system for Norway lobster (Nephrops norvegicus)?. Oceanol. Acta 19:357368.
  • Horsburgh, K.J., Hill, A.E., Brown, J., Fernand, L., Garvine, R.W. and Angelico, M.M.P. (2000) Seasonal evolution of the cold pool gyre in the western Irish Sea. Prog. Oceanogr. 46:158.
  • Hurrell, J.W. (1995) Decadal trends in the North Atlantic Oscillation: regional temperatures and precipitation. Science 269:676679.
  • ICES (2010) Report of the Working Group on the Celtic Seas Ecoregion (WGCES), 12–20 May 2010, Copenhagen, pp. 1435.
  • ICES (2011) Report of the Working Group for Celtic Seas Ecoregion (WGCSE), 11–19 May 2011. Copenhagen, Denmark, pp. 1573.
  • Kjesbu, O.S. (1994) Time of start of spawning in Atlantic cod (Gadus morhua) females in relation to vitellogenic oocyte diameter, temperature, fish length and condition. J. Fish Biol. 45:719735.
  • Kjesbu, O.S., Solemdal, P., Bratland, P. and Fonn, M. (1996) Variation in annual egg production in individual captive Atlantic cod (Gadus morhua). Can. J. Fish. Aquat. Sci. 53:610620.
  • Kjesbu, O.S., Righton, D., Kruger-Johnsen, M. et al. (2010) Thermal dynamics of ovarian maturation in Atlantic cod (Gadus morhua). Can. J. Fish. Aquat. Sci. 67:605625.
  • Köster, F.W., Möllmann, C., Hinrichsen, H.H. et al. (2005) Baltic cod recruitment - the impact of climate variability on key processes. ICES J. Mar. Sci. 62:14081425.
  • Lilly, G.R., Wieland, K., Rothschild, B.J. et al. (2008) Decline and Recovery of Atlantic Cod (Gadus morhua) Stocks throughout the North Atlantic. Resiliency of Gadid Stocks to Fishing and Climate Change. Alaska Sea Grant College Program, pp. 3966.
  • Litzow, M. and Cianneli, L. (2007) Oscillating trophic control induces community reorganization in a marine ecosystem. Ecol. Lett. 10:11241134.
  • Llope, M., Daskalov, G.M., Rouyer, T.A. et al. (2011) Overfishing of top predators eroded the resilience of the Black Sea system regardless of the climate and anthropogenic conditions. Glob. Change Biol. 17:12511265.
  • Lynam, C.P., Lilley, M.K.S., Bastian, T., Doyle, T.K., Beggs, S.E. and Hays, G.C. (2011) Have jellyfish in the Irish Sea benefited from climate change and overfishing? Glob. Change Biol. 17:767782.
  • Marshall, C.T., Kjesbu, O.S., Yaragina, N.A., Solemdal, P. and Ulltang, O. (1998) Is spawner biomass a sensitive measure of the reproductive and recruitment potential of Northeast Arctic cod? Can. J. Fish. Aquat. Sci. 55:17661783.
  • Marteinsdottir, G. and Begg, G.A. (2002) Essential relationships incorporating the influence of age, size and condition on variables required for estimation of reproductive potential in Atlantic cod Gadus morhua. Mar. Ecol. Prog. Ser. 235:235256.
  • Marteinsdottir, G. and Thorarinsson, K. (1998) Improving the stock-recruitment relationship in Icelandic cod (Gadus morhua) by including age diversity of spawners. Can. J. Fish. Aquat. Sci. 55:13721377.
  • Marteinsdottir, G., Ruzzante, D.E. and Nielsen, E.E. (2005) History of the North Atlantic cod stocks. ICES CM, 2005/AA19, pp. 17.
  • Methot, R.D. (1986) Frame trawl for sampling pelagic juvenile fish. California Cooperative Oceanic Fisheries Investigations Report XXVII, 267278.
  • Möllmann, C., Müller-Karulis, B., Kornilovs, G. and St John, M.A. (2008) Effects of climate and overfishing on zooplankton dynamics and ecosystem structure: regime shifts, trophic cascade, and feedback loops in a simple ecosystem. ICES J. Mar. Sci. 65:302310.
  • Myers, R.A. (2001) Stock and recruitment: generalizations about maximum reproductive rate, density dependence, and variability using meta-analytic approaches. ICES J. Mar. Sci. 58:937951.
  • Myers, R.A. and Cadigan, N.G. (1993) Density-dependent juvenile mortality in marine demersal fish. Can. J. Fish. Aquat. Sci. 50:15761590.
  • Olbert, A.I., Hartnett, M., Dabrowski, T. and Mikolajewicz, U. (2011) Long-term inter-annual variability of a cyclonic gyre in the western Irish Sea. Cont. Shelf Res. 31:13431356.
  • Olsen, E.M., Ottersen, G., Llope, M., Chan, K.-S., Beaugrand, G. and Stenseth, N.C. (2011) Spawning stock and recruitment in North Sea cod shaped by food and climate. Proc. Biol. Sci. 278:504510.
  • Ottersen, G., Hjermann, D.O. and Stenseth, N.C. (2006) Changes in spawning stock structure strengthen the link between climate and recruitment in a heavily fished cod (Gadus morhua) stock. Fish. Oceanogr. 15:230243.
  • Planque, B. and Fox, C.J. (1998) Interannual variability in temperature and the recruitment of Irish Sea cod. Mar. Ecol. Prog. Ser. 172:101105.
  • Planque, B. and Fredou, T. (1999) Temperature and the recruitment of Atlantic cod (Gadus morhua). Can. J. Fish. Aquat. Sci. 56:20692077.
  • Planque, B., Fromentin, J.M., Cury, P. et al. (2010) How does fishing alter marine populations and ecosystems sensitivity to climate? J. Mar. Syst. 79:403417.
  • Plirú, A., van der Kooij, J., Engelhard, G.H., Fox, C.J., Milligan, S.P. and Hunter, E. (2012) Sprat feeding behaviour, selective predation, and impact on plaice egg mortality. ICES J. Mar. Sci. 69:10191029.
  • Rayner, N.A., Brohan, P., Parker, D.E. et al. (2006) Improved analyses of changes and uncertainties in sea surface temperature measured in situ since the mid-nineteenth century: the HadSST2 Dataset. J. Clim. 19:446469.
  • Rijnsdorp, A.D., Peck, M.A., Engelhard, G.H., Möllmann, C. and Pinnegar, J.K. (2009) Resolving the effect of climate change on fish populations. ICES J. Mar. Sci. 66:15701583.
  • Secor, D.H. (2007) The year-class phenomenon and the storage effect in marine fishes. J. Sea Res. 57:91103.
  • Stige, L.C., Ottersen, G., Brander, K., Chan, K.S. and Stenseth, N.C. (2006) Cod and climate: effect of the North Atlantic Oscillation on recruitment in the North Atlantic. Mar. Ecol. Prog. Ser. 325:227241.
  • Walters, C. and Kitchell, J.F. (2001) Cultivation/depensation effects on juvenile survival and recruitment: implications for the theory of fishing. Can. J. Fish. Aquat. Sci. 58:3950.
  • Wood, S.N. (2001) mgcv. GAMs and Generalized Ridge Regression for R. R News, 1/2, 2025.
  • Wood, S.N. (2004) Stable and efficient multiple smoothing parameter estimation for generalized additive models. J. Am. Stat. Assoc. 99:673686.

Supporting Information

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information
FilenameFormatSizeDescription
fog12043-sup-0001-FigureS1-S4.docxWord document614K

Figure S1. Diagnostic plots TGAM1.

Figure S2. Diagnostic plots GAM.

Figure S3. GAM output.

Figure S4. Threshold identification for SST (left panel) and SSB (right panel).

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.