Keywords

1 Introduction

Fisheries production depends on the interaction between human inputs (fishing effort) and the biomass of fish. While fishing effort may change over time due to socioeconomic factors, such as market interests, changing costs of effort or other reasons, fish biomass fluctuates over time due to environmental factors, such as biological productivity of its prey resources, or abiotic factors such as changes in temperature or salinity (Bakun, 1996; Cushing, 1996; Ottersen et al., 2004).

The abundance of marine fish is difficult to determine and many different methods have been developed over time (Hilborn and Walters, 1992). Briefly, fish abundance can be estimated from two basic sources: fisheries dependent data (commercial catches) or fisheries-independent data. In the first case, data series of catches are relatively easy to obtain because most countries maintain relatively long data bases on fish production by species (Evans and Grainger, 2001). However, to interpret catch data it is often necessary to know the evolution of fishing effort and catchability, which are often not available or difficult to estimate (Hilborn and Walters, 1992). Fisheries-independent programmes are more reliable but expensive to maintain and rarely have the long time spans required to assess the relationship between environmental variations and fish abundance (Sparre and Venema, 1998). In the European North Atlantic, there exist data series derived from scientific trawl surveys longer than 40 years (International Bottom Trawl Survey: http://datras.ices.dk), but in the Mediterranean, the only continuous fisheries monitoring programme (MEDITS: Mediterranean International Trawl Surveys: http://www.ifremer.fr/Medits_indices) to cover all EU member states started in 1994 and the data series is thus considerably shorter than in the Atlantic (Bertrand et al., 2002). In any case both types of data sources, commercial and trawl surveys, have advantages and disadvantages and it is convenient to use the two types of data concurrently when possible.

It is known that the changes in total biomass of a fishery target species can be due to individual changes in (one or more) individual demographic parameters. For instance, the abundance of a given year class may fluctuate from year to year due to environmental factors that cause enhanced survival (or conversely, increased mortality) of the individuals of that particular year class (Cushing, 1996; Myers, 2002; Watanabe et al., 2003). This is specially important in the first age class recruited to the fishery (known as recruitment), and several studies have shown the relationship between interannual fluctuations in recruitment and environmental variability (e.g. Mendelsohn and Mendo, 1987; Watanabe et al., 1996; Stige et al., 2006). Other key life history parameters affected by environmental variability are the individual growth rate, the natural mortality or the reproductive success (Hänninen et al., 1999; Jobling, 2002).

It is often difficult to determine the causes of the relationship between large scale climatic indices and fisheries due to the paucity of data, but also due to the complexity of the mechanisms involved (Cushing, 1996). In recent years, the usefulness of employing large scale climatic indices such as the NAO index as a proxy for complex space-time variations in ecological variables has been suggested (Hallett et al., 2004; Stenseth and Mysterud, 2005) and indeed the relationship between climatic indices and fisheries is well documented in non-Mediterranean settings, particularly in upwelling ecosystems (e.g. anchovy/sardine substitutions in the Pacific Kuroshio and Humboldt Current ecosystems and teleconnections among fish populations, reviewed in Alheit and Bakun, 2010).

The effects of NAO on North Atlantic fisheries has received attention in recent decades, especially in the last 10 years (Reid et al., 2001; Beaugrand, 2004), with studies focused particularly on cod (Gadus morhua, Mann and Drinkwater, 1994; Beaugrand et al., 2003; Stige et al., 2006). The results of these studies show that the NAO affects cod fisheries mainly by modulating the strength of cod recruitment with a time lag of 2 years (Brodziak and O’Brien, 2005). It has been shown also that the effect is stronger when spawning stock biomass (SSB) is low (Brander, 2005). Most significantly, some of these studies showed that the effect of the NAO varies spatially across North Atlantic cod stocks (synthesized in the model by Stige et al., 2006): while cod recruitment is enhanced in Newfoundland, Greenland, Iceland, northern Norway and the Faroes during positive NAO years, it is decreased in the eastern US, around the British Isles, in the North sea and in the Baltic sea (figure 4 in Stige et al., 2006).

In many cases the relationship between the NAO and fisheries production has been shown to be mediated by changes in the composition and abundance of zooplankton (Fromentin and Planque, 1996; Ottersen and Stenseth, 2001; Reid et al., 2003; Molinero et al., 2008).

In the Mediterranean, the relationship between the NAO and fisheries has received less attention, probably because data series are generally shorter than in Atlantic fisheries. Lloret et al. (2001) assessed the relationship between river runoff in 13 coastal fishery species of the Catalan coasts. They found that catches per unit of effort were positively correlated with the wind mixing index in the Gulf of Lions and the runoff of local rivers, which varied synchronously with the NAO index: negative NAO indices were associated with higher river runoff (probably because during negative NAO years precipitation increases in the NW Mediterranean, Mariotti et al., 2002). Maynou (2008) showed that the detrended landings of the deepwater red shrimp Aristeus antennatus were positively correlated with the NAO index with time lags of 2 and 3 years. An ecological mechanism to explain this correlation was produced, based on current knowledge of the biology of this species and the water mass dynamics in the study area (NW Mediterranean): positive NAO years are correlated with decreased rainfall in the NW Mediterranean sea, especially in the Gulf of Lions, the Ligurian sea and adjacent hinterland (Mariotti et al., 2002). Decreased rainfall is linked to enhanced vertical mixing of the water masses in the Gulf of Lions (Demirov and Pinardi, 2002; Jordi and Hameed, 2009), due to the saltier (hence, more dense) surface water mass in low rainfall years. Knowing that food supply in winter is key to the reproductive success of A. antennatus, because during this period females switch from their normal generalist diet to a high-energy content diet based on zooplankton (Cartes et al., 2008; Maynou, 2008) postulated that positive NAO years would enhance egg production and higher abundance of recruits to the fishery.

The objective of this contribution is to assess the effects of NAO in another important Mediterranean demersal fishery resource, the hake Merluccius merluccius at the scale of the entire Mediterranean basin, in order to investigate whether ecological mechanisms linking the NAO with fisheries productivity demonstrated in local areas (Catalan sea, Ligurian sea) can be extrapolated to wider areas.

2 Results

European hake (Merluccius merluccius) is an important resource of the Mediterranean demersal fishery, with catches fluctuating between 20 and 50,000 t in recent decades (FAO Fisheries Department, 2000). The biology of this species is well-known (e.g. the review by Oliver and Massutí, 1995; Recasens et al., 1998). Hake is distributed mainly over the continental shelf between 50 and 200 m depth, although occasional large individuals are caught down to 800 m. It is a heavily exploited species and catches are at present mainly composed of juvenile individuals of age classes 0–2. The recruitment period of hake varies across Mediterranean subareas and is relatively protracted, but it occurs mainly in early summer (June–July) in the NW Mediterranean (Recasens et al., 1998).

I analyzed the annual data series of hake catches for the period 1970–2005 in the seven geographical areas of the Mediterranean used by the FAO for fisheries statistics purposes (Fig. 1) based on the FISHSTAT database (FAO Fisheries Department, 2000).

Fig. 1
figure 7_1_217618_1_En

FAO fisheries divisions in the Mediterranean and geographic names mentioned in the text (italics)

Full size image

As shown in Fig. 2, the data series of hake catches are very noisy, although some general patterns emerge. First, hake catches increased from 1970 until the early 1990s in most areas (Gulf of Lions, Adriatic, Ionian and Aegean). In these areas, and also in the Levant, hake production decreased since the early 1990s to the present due probably to overfishing (Lleonart, 2008). In the Balearic area no overall trend can be distinguished, but periods of high catches and low catches can be discerned, while in the Sardinia area hake catches have been decreasing since the beginning of the time series (Fig. 2).

Fig. 2
figure 7_2_217618_1_En

Data series of hake (Merluccius merluccius) landings in the seven Mediterranean FAO fisheries divisions for the period 1970–2005 (data extracted from FISHSTAT, FAO Fisheries Department, 2000)

Full size image

To analyze these data sets, a Generalized Additive Model (GAM) was set up for each statistical area, using year as the main effect (because it is expected that long-term trends in interannual variations in catches are mainly due to the observed increase in fishing effort, Lleonart, 2008) and the NAO index as additional effect with time lags of 1 and 2 years (nao1 and nao2). GAMs are a flexible class of statistical regression models which allow to assess linear or nonlinear relationships between a set of predictors (in our case, year, nao1 and nao2) and a dependent variable (in our case hake catch) (Wood, 2006):

$$\begin{aligned}\log (catch) &{}= {\textrm{s}}(year) + {\textrm{s}}(nao1) + {\textrm{s}}(nao2) + \varepsilon ,\\\\ &{}\ \quad{\textrm{where }}\varepsilon \,{\textrm{is a normally distributed error term}}{\textrm{.}}\end{aligned}$$

I used smoothing splines to represent the possibly nonlinear effect of predictors and Generalized Cross Validation (GCV) to establish the optimal degree of the smooth terms. Model selection was achieved by means of Akaike’s Information Criteria to assess the goodness of fit (Wood, 2006).

The GAM models fitted for hake catch in the seven statistical subareas for the period 1970–2005 helped explain between 62% (Gulf of Lions) and 96% (Aegean sea) of the deviance in the data series (Table 1). These results show that most of the variability in the catch data series can be attributed to the year effect, which mostly reflects changes in fishing effort over time. More interestingly, the results of the GAM analysis show that the lagged NAO index explains between 2 and 8% of the deviance of the models and that some of the functional non-parametric terms are statistically significant (Table 1). The shape of these nonparametric terms can be examined in Fig. 3, which shows that hake stocks in the western and northern Mediterranean areas (Gulf of Lions, Balearic and Adriatic) are positively correlated with the NAO index at lags 1 and 2 years, while hake stocks in the eastern Mediterranean (Aegean and Levant) would be negatively correlated with these NAO indices (Fig. 3). Hake stocks in the central Mediterranean show weak correlations (Ionian) or of opposite sign (Sardinian).

Fig. 3
figure 7_3_217618_1_En

Non-parametric smooth terms determined by the GAM analysis for hake (Merluccius merluccius) landings in the seven Mediterranean FAO fisheries divisions. Left column: effect of NAO index lagged 1 year. Right column: effect of NAO index lagged 2 years

Full size image
Fig. 4
figure 7_4_217618_1_En

Autocorrelograms of significant demographic indicators (Table 2) extracted from the MEDITS database for hake (Merluccius merluccius) in the Spanish Mediterranean coast (subarea 6). Dotted lines indicate significant correlation at the p < 0.05 level

Full size image
Table 1 Results of the GAM analysis of hake (Merluccius merluccius) landings in the seven Mediterranean areas for fishery purposes defined by FAO
Full size table

To investigate the causes linking NAO with hake fisheries variability I analyzed the correlation between several life-history indicators derived from fisheries independent trawl surveys in the western Mediterranean (MEDITS surveys carried out along the continental coast of Spain, Bertrand et al., 2002) by means of autocorrelation function analysis (Venables and Ripley, 2002). The span of this data set (1994–2009) is shorter than the FAO production statistics, but includes a critical transitional period from positive values of the NAO index in the 1990s to negative or zero values later on. I extracted several demographic indicators from the MEDITS database (http://www.ifremer.fr/Medits_indices): abundance by age class, natural mortality by age class, total biomass, spawning stock biomass, and average individual length and weight (Table 2).

Table 2 Demographic indicators of hake (Merluccius merluccius) analyzed by autocorrelation functions. Indicators were derived from the annual MEDITS trawl surveys carried out along the Mediterranean continental coast of Spain (1994–2009)
Full size table

The results of autocorrelation function analysis (Fig. 4) show that the mean length of recruits (hake of age class 0) is positively correlated with NAO at lag 1, i.e. recruits are larger than average after positive NAO years. Also the abundance of adults (age classes 3–6) of hake is increased 1 and 2 years after positive NAO years, while the average weight of adults is positively correlated with NAO at lag 1. Despite this increase in abundance and mean weight of adults, no increase in spawning stock biomass could be shown in positive NAO years, probably because the biomass of the adult fraction of the population is much lower than the juvenile fraction (age classes 0–2) in this heavily exploited population. Natural mortality was not correlated to the NAO. These results show that in the Balearic subarea, and probably also in the Gulf of Lions and Adriatic, positive NAO years have a positive effect in at least 3 critical indicators of the hake life history (length of recruits, abundance of adults and weight of adults) suggesting that the positive effects on the catches shown earlier from the FAO production data series can be related to increase size and weight of the hake population and not, for instance, to decreased natural mortality or increased reproductive success, as in the red shrimp (Maynou, 2008).

3 Conclusions

The North Atlantic Oscillation has profound influences on fisheries productivity, mediated by complex dynamics of ocean processes and biological productivity, as shown in North Atlantic fish stocks (Ottersen and Stenseth, 2001; Reid et al., 2001, 2003; Stenseth and Mysterud, 2005; Stige et al., 2006), but also in the Mediterranean (Lloret et al., 2001; Maynou, 2008).

In fisheries research the influence of environmental variation on stock productivity may be difficult to detect due to the lack of long and consistent time series of the main oceanographic and ecological parameters at appropriate scales. Large-scale climatic indices such as the NAO are often better proxies of ecological processes than local weather variables, because they reduce the complexities of time and space variability in environmental phenomena to simple measures (Hallett et al., 2004; Stenseth and Mysterud, 2005), and are specially useful for fish abundance data where comparable environmental data sets are non existent or expensive to obtain (e.g. for deep-sea fishery resources, Maynou, 2008).

The influence of the NAO on fish stocks has been shown to vary geographically, with impacts of different sign in cod stocks of, broadly, the western Atlantic and the eastern Atlantic (Stige et al., 2006). In Mediterranean hake stocks, present results suggest also geographically varying impacts of the NAO. A N/S – W/E trend in the effects of NAO has been evidenced: in the N and W areas (Balearic sea, Gulf of Lions, Adriatic sea) the effect is close to linear and positive (enhanced hake catches at positive NAO values). In the E and S subareas (Aegean sea, Levantine sea) the effects is negative (enhanced catches at negative NAO values), while in the two central subareas (Sardinian seas and Ionian sea) the effect is weak but mainly positive. Focusing on the Balearic area, I used a shorter data set derived from fisheries independent surveys to investigate the possible biological causes of enhanced catches by NAO in this area. I showed that the positive relationship between NAO and catches of hake is related to the positive effect of NAO (with temporal lags of 1 or 2 years) on at least 3 critical indicators of the hake life history: length of recruits, abundance of adults and weight of adults. These results suggest that the positive effects of NAO on Mediterranean hake catches can be related to increased size and weight of the hake population and not, for instance, to decreased natural mortality nor increased reproductive success, as in the red shrimp (Maynou, 2008).

Although the effect of the NAO on fisheries of Mediterranean stocks has been shown in a few studies (Lloret et al., 2001; Maynou, 2008; present work for hake), further research is needed to investigate the relationships linking the North Atlantic Oscillation, Mediterranean oceanography and biological productivity. Although it is likely that the effect of the NAO on higher trophic levels (which constitute the fisheries target species) is mediated by the effects of NAO on zooplankton (Fromentin and Planque, 1996; Maynou, 2008; Molinero et al., 2008) the details of these relationships are often mere working hypotheses that need to be refined and tested.