- Split View
-
Views
-
Cite
Cite
Teresa Silva, Astthor Gislason, Priscilla Licandro, Gudrún Marteinsdóttir, Ana Sofia A. Ferreira, Kristinn Gudmundsson, Olafur S. Astthorsson, Long-term changes of euphausiids in shelf and oceanic habitats southwest, south and southeast of Iceland, Journal of Plankton Research, Volume 36, Issue 5, September/October 2014, Pages 1262–1278, https://doi.org/10.1093/plankt/fbu050
- Share Icon Share
Abstract
Generalized additive models (GAMs) were used to test the hypothesis that changes in physical and biological environmental conditions affected by current climatic warming would negatively impact the euphausiid populations in the North Atlantic. Two zooplankton time series were used, one collected by the Marine Research Institute (MRI) on a transect south of Iceland during spring (1990–2011) and the other by the Continuous Plankton Recorder (CPR) survey (1958–2007) in the oceanic waters south of Iceland covering all months. Due to limitations of the sampling gears used, the results mainly reflect the variations of the early stages of euphausiids. On a spatial scale, results reveal a general decline of euphausiid abundance from the east coast of Greenland to the Faroe Islands. On a temporal scale, euphausiid numbers decreased in most CPR areas from 1958 to 2007. Conversely, an increase was observed in numbers of larvae during spring 1990–2011 for the shelf south of Iceland. Single variable-based GAMs indicated that phytoplankton biomass was generally the main environmental factor regulating euphausiid abundance. Multiple variable-based GAMs showed that phytoplankton biomass was the strongest predictor of euphausiid abundance in the west, whereas in the east temperature appears to be most important. In addition, the onset of the spring bloom also affected the long-term changes in euphausiid abundance. For the oceanic areas, it is concluded that a weakened temporal synchrony between the development of young euphausiids and the phytoplankton bloom influenced by recent climate warming may have led to the observed decrease in euphausiid populations.
INTRODUCTION
Euphausiids are an important zooplankton group in the Icelandic marine ecosystem as they constitute the second largest component of the zooplankton biomass after copepods (Astthorsson and Gislason, 1995; Astthorsson et al., 2007), and at least in some regions (e.g. north of Iceland) dominate the zooplankton community in winter months (Astthorsson and Gislason, 1992). Euphausiids prey on both phytoplankton and small zooplankton, while being an important food for top predators such as cod, saithe, capelin, herring, blue whiting, whales and sea birds (Pálsson, 1983; Astthorsson and Pálsson, 1987; Vilhjálmsson, 1994; Astthorsson and Gislason, 1997b; Sigurjónsson and Víkingsson, 1997; Jaworski and Ragnarsson, 2006; Pálsson and Björnsson, 2011).
Einarsson (Einarsson, 1945) reviewed the distribution patterns and the general biology of euphausiids in the Northeastern Atlantic and around Iceland. Later studies used material from the Continuous Plankton Recorder (CPR) survey of the Sir Alistair Hardy Foundation for Ocean Science (SAHFOS) further to describe the distribution and population dynamics of euphausiids in the North Atlantic Ocean (Lindley, 1977, 1978, 1980, 1982a, b; Lindley and Williams, 1980; Williams and Lindley, 1982). More recent studies in Icelandic waters have focused on the distribution and life cycles of euphausiids in a fjord on the north-western peninsula (Astthorsson, 1990; Astthorsson and Gislason, 1992), over the shelf north of Iceland (Astthorsson and Gislason, 1997a), and in the subarctic waters of the Iceland Sea, to the north of Iceland (Gislason and Silva, 2012). The most recent studies investigated the distribution and population patterns of euphausiids in the Irminger Sea and over the northern Mid-Atlantic ridge (Saunders et al., 2007; Letessier et al., 2009, 2011). These studies showed that Thysanoessa raschii, T. inermis, T. longicaudata and Meganyctiphanes norvegica were the most common euphausiid species in these regions.
Research has shown that bottom-up rather than top-down effects regulate the marine ecosystem north and east of Iceland (Astthorsson and Vilhjálmsson, 2002; Astthorsson et al., 2007). Changes in hydrographic conditions, mainly due to variable inflow of Atlantic water, influence stratification and consequently the magnitude and timing of the phytoplankton spring bloom, which in turn affects zooplankton abundance and composition (Astthorsson and Gislason, 1995, 1998; Gislason et al., 2009). These variations have further been shown to affect growth of fish in the area (Astthorsson and Vilhjálmsson, 2002; Astthorsson et al., 2007). In the more biologically complex southern and western areas, causal events are harder to identify (Astthorsson et al., 2007).
The rise in seawater temperature driven by climate change appears to have altered the phenology, abundance and diversity of plankton in the Northeastern Atlantic Ocean (Reid, 2001; Beaugrand and Reid, 2003; Beaugrand et al., 2003; Edwards and Richardson, 2004; Reid and Valdés, 2011). Thus, climate-induced changes in the plankton community have affected higher trophic levels by asynchrony between production at the base of the food web and at higher levels (Beaugrand and Reid, 2003; Edwards and Richardson, 2004). Several studies have related a decline in the stocks of key Antarctic euphausiid species (Euphausia superba) in the Southern Ocean to climate warming and its implications for the abiotic and biotic environment (e.g. Atkinson et al., 2004; Flores et al., 2012).
We hypothesize that changes in physical and biological environmental conditions affected by the rise in sea water temperatures would negatively impact the euphausiid populations. Thus, it is the main purpose of this study is to describe the long-term and seasonal changes of euphausiids in Icelandic waters and adjacent sea areas. In particular, we aim to evaluate how environmental variables and the onset of the phytoplankton spring bloom and biomass affect the multidecadal variability of euphausiids.
METHOD
For this study, two time series of euphausiid abundance were used, one collected by the Marine Research Institute (MRI) on a transect south of Iceland in May–June from 1990 to 2011 (the Icelandic spring survey) and the other by the CPR survey covering a wider area of the Northeastern Atlantic Ocean and the whole year (1958–2007) (Fig. 1).
The Icelandic spring survey
From 1990 to 1991, zooplankton samples were collected with Hensen nets (0.42 m2 mouth area and 200 µm mesh size), whereas from 1992 to 2011 with WP2 nets (0.25 m2 mouth area and 200 µm mesh size). The plankton nets were towed from 50 m depth (or from ∼2 m above the bottom where depth was <50 m) to the surface at a velocity of ∼0.75 m s−1. HydroBios flowmeters were fitted in the net mouth to measure the volume of water filtered.
The samples were preserved in 4% buffered formaldehyde until analysis in the laboratory ashore. The procedure consisted in counting and identifying under a stereomicroscope the larger zooplankters such as adult euphausiids in the whole sample, and smaller ones in sub-samples obtained by a Motoda splitter containing at least 400 zooplankters (Motoda, 1959). Adult euphausiids were identified to the species level and the earlier developmental stages were separated into the following categories: eggs, and nauplius, metanauplius, calytopsis and furcilia stages. For the present analysis, the larval stages were treated as one group (nauplii, metanauplii, calyptopsis and furciliae).
The CPR survey
The CPR (0.013 m2 mouth area and 270 µm mesh size) collects plankton at approximately monthly intervals sampling in continuous along standard routes crossing the North Atlantic Ocean (Fig. 1). The instrument is towed by ships of opportunity at an estimated mean depth of ∼7 m and at an average velocity of ∼6.6 m s−1 (Batten et al., 2003). The distance corresponding to one CPR sample is equivalent to filtering ∼3 m3 of seawater (i.e. 10 nautical miles). At SAHFOS, the samples were processed and the plankton taxa identified and counted using standard procedures (Batten et al., 2003; Richardson et al., 2004, 2006). Monthly data from 1958 to 2007, collected in the CPR standard areas southwest, south and southeast of Iceland, were used (58–66°N, 43°W–3°E, Fig. 1). In the analysis, euphausiid furcilia, juveniles and adult counts (Lindley, 1977; Richardson et al., 2006) were combined and converted to total numbers per m3.
Explanatory variables
Concurrent with the zooplankton sampling in the Icelandic spring survey, temperature and salinity were measured at each station with a CTD (Sea Bird Electronics SBE-9) and water samples collected from five depths (0, 10, 20, 30 and 50 m) for chlorophyll a (Chl a) analysis. The water was filtered through GF/C glass fibre filters that were homogenized in 90% aqueous acetone and the Chl a extracted and measured spectrophotometrically. Temperature, salinity and Chl a values were averaged from the surface down to 50 m depth.
Data on monthly sea surface temperature (SST) from 1958 to 2011 were obtained from the UK Met Office Hadley Centre (HadISST v 1.1, available at http://badc.nerc.ac.uk) and averaged for each CPR area (Fig. 1). Monthly sea surface salinity data at ∼5 m depth were obtained from the Global Ocean Data Assimilation System (GODAS, available at http://www.esrl.noaa.gov/psd/), and averaged for each CPR area from 1980 to 2011. Monthly means of the North Atlantic Oscillation (NAO) index from 1958 to 2011 were obtained from the National Weather Service—Climate Prediction Center (available at http://www.nws.noaa.gov/). The index is a measure of the pressure difference between the Icelandic low and the Azores high. The daily NAO index is constructed by projecting the daily (00Z) 500 mb height anomalies over the Northern Hemisphere onto the loading pattern of the NAO. For the present analysis, the NAO winter index was used (average from December to April) as suggested by Fromentin and Planque (Fromentin and Planque, 1996).
Data on surface Chl a concentrations were obtained from the European Space Agency's GlobColour project (http://www.globcolour.info). Weekly Chl a concentrations from 1998 to 2011, on a 25-km grid, were averaged for the same geographic locations as the Selvogsbanki stations (an area 25 × 25 km with the station in the centre) and for the CPR areas (Fig. 1).
For all regions (Selvogsbanki transect and CPR areas), the onset of the phytoplankton spring bloom was estimated from the surface Chl a concentration data of the GlobColour project. To avoid errors in estimation due to missing data around the date of interest, a generalized additive model (GAM) fit was first applied to the surface Chl a concentrations to estimate its seasonality. The start of the spring bloom was defined as the first week of the year, in which the Chl a concentration increased by 5% above the annual median value (Siegel et al., 2002; Henson et al., 2009).
Data analysis
All the sampling gear used in the present study will inevitably under sample euphausiids (Brinton and Townsend, 1981; Sameoto et al., 1993, 2000; Nicol, 2003; Wiebe et al., 2004). To assess the efficiency of the different gears for catching euphausiids, we calculated the catching efficiency (CE) of the different plankton samplers. The calculations were done based on the methodology of Clutter and Anraku (Clutter and Anraku, 1968), CE = [(R− K/Si)/R]2, in which CE is the catching efficiency (the percentage of euphausiids that are in the path of the sampler that are captured), R is the net radius (m), K is the avoidance parameter [detection distance (m) × mean swimming speed (m s−1)] and Si is the towing speed (m s−1). The calculations were based on the different towing speeds and entrance radii of the samplers. We are not aware of any previous estimates of detection distance of euphausiids, but assume here that the euphausiids would be able to detect the sampling gear within a distance of 0.25 m. Studies on swimming speeds of euphausiids indicate that they range from 1 to 3 body lengths per second (BL s−1) with a mean of ∼1.7 BL s−1 (Price, 1989; Klevjer and Kaartvedt, 2003, 2006, 2011). We used the mean value in relative swimming speed (BL s−1) to calculate swimming speeds in units of m s−1 of euphausiids ranging in size from 1 to 5 cm. The outcome was in turn used to calculate CE for different size classes of euphausiids (1–5 cm). Using this approach, we found that the efficiency of the WP2 net for catching euphausiids ranged from 0.96 for 1 cm long euphausiids to 0.82 for 5 cm long euphausiids. Calculations for the Hensen net gave similar values (0.97–0.85). The CPR showed much lower catching efficiencies ranging from 0.87 (1 cm long euphausiids) to 0.43 (5 cm). These simple calculations demonstrate that all the sampling gears used (Hensen, WP2 and CPR) catch euphausiids up to 2 cm in size reasonably well (CE >0.70), thus including the larval and juvenile stages of the three most common euphausiid species in the North Atlantic (T. longicaudata, T. inermis and M. norvegica) as well as the adult stages of the smallest species (T. longicaudata), which also is the one most common in terms of numbers (Einarsson, 1945; Lindley, 1982b). While we recognize that the calculations presented above can only be considered approximate estimates, as both detection distance and swimming speed are estimated with great uncertainty, we can still conclude that the CE of all gears decreases with the size of the euphausiids. Probably, >70% of the larvae and juveniles are captured by all samplers. It should, however, be noted that the calculations do not take into account the effects of bow wave pressure on the avoidance behaviour of the euphausiids which is likely to be more pronounced for the CPR, being towed behind large container ships at relatively shallow depth, than for the vertically towed Hensen and WP2 nets. Due to the low efficiency of samplers for catching adult euphausiids, for the Icelandic spring survey we only used the data on eggs and larvae for the analysis of temporal and spatial variability. As to the CPR data, the euphausiids are only recorded as total euphausiids when processing the samples at SAHFOS (Batten et al., 2003) and therefore the adults could not be separated from the CPR data. Lindley and co-workers have used the CPR data in a number of studies on abundance and distribution of euphausiids in the North Atlantic (e.g. Lindley, 1977, 1978, 1980, 1982a, b; Lindley and Williams, 1980; Williams and Lindley, 1982), while realizing that the CPR samplers do not catch equally all developmental stages (A. Lindley, personal communication). In spite of the limitations posed by the relatively low CE of the devices used for sampling the euphausiids in the present study, we nevertheless feel confident in using the data as indices of temporal and spatial variability (thus assuming that the CE does not change with time or region). We realize, however, that they cannot be used in order to compare absolute abundances.
Although the basic design of the CPR has remained relatively unchanged since the late 1940s, there has been a steady increase in the speed at which it is towed (Batten et al., 2003). This has led to significant decreases in sampled volume due to higher incidence of net clogging, which is not accounted for by the standardized plankton analysis at SAHFOS. It is conceivable that the significant decrease in sampling volume would result in decreases in calculated densities of plankton. The issue is clearly relevant for the consistency and interpretation of the CPR data, and it has therefore received considerable attention. John et al. (John et al., 2002) quantified the relationship between the volume of water filtered per sample and the extent of clogging using flow metered tows, and concluded that the effect of clogging on filtering rates was not great. Similarly, Batten et al. (Batten et al., 2003) reported that although there was some evidence that flow would be reduced with time, estimates of plankton abundance for large areas remained unaffected. The most recent study on the long-term changes in volume filtered by the CPR and their relevance to CPR data interpretation (Jonas et al., 2004) similarly found no significant correlation between long-term changes in ship speed and two commonly used indicators of plankton variability, the phytoplankton colour and total copepods indices. Based on these studies, we feel confident that the effect of long-term changes in filtered volume on euphausiid abundance is small compared with influences of other factors.
For both data sets (Icelandic spring survey and CPR), the distribution of the euphausiid abundance data was strongly skewed to the right and with several zero observations; therefore, a logarithmic transformation [ln(x + 1)] was applied. To avoid estimation errors because of missing values in the CPR data, an interpolation of the data was made using data-interpolating empirical orthogonal functions (Beckers and Rixen, 2003).
For both data sets (Icelandic spring survey and CPR), two-way analysis of variance (ANOVA) was used to test statistically if the long-term variability was similar during day and night. Day samples were defined as samples taken between sunrise and sunset and night samples between sunset and sunrise. In a test like this, a significant interaction term [year versus daytime (day or night)] means that the long-term variability is different during day and night. The interaction term (year versus day and night) was, however, only significant for one area (B7, ANOVA, P < 0.05). As it may be assumed that euphausiids stay shallower in the water column during night than during day (e.g. Eriksen and Dalpadado, 2011), and also as euphausiids probably better avoid sampling gear during day time (Wiebe et al., 2004), night samples were considered to better represent abundance. Therefore for area B7, the night data were used to infer the long-term changes. For the remaining regions, day and night samples were combined.
To visually compare euphausiid abundance values collected by the CPR survey and the Icelandic spring survey, the data were standardized to a mean of 0 and SD of 1.
GAMs were used to analyse long-term changes in abundance as function of hydrographic and biological variables (Wood, 2006). The data were characterized by a large number of zero observations that would make the use of standard error distribution for GAM analysis (Gaussian) inappropriate. Therefore, an approach combing two models, presence/absence and abundance larger than zero, was used (Stefánsson and Pálsson, 1997; Barry and Welsh, 2002). The first model uses the presence/absence of the euphausiids as the response variable assuming a binomial distribution (logit link function). In the second model, euphausiid numbers larger than zero are used as the response variable which is assumed to be Gaussian distributed. In all models, thin plate regression splines were used with a maximum of three effective degrees of freedom as a smoothing function for each predictor. In the GAMs, temperature, salinity, Chl a (surface values or average from 0 to 50 m), onset of the phytoplankton spring bloom and NAO winter index were used as predictors.
Both single and multiple variable-based GAMs were used to study the influence of the predictors on euphausiid abundance. However, as the different explanatory variables did not have the same temporal coverage, the GAMs were limited to the period 1998–2007 for the CPR data and 1998–2011 for the Icelandic spring survey data, when records for all the environmental predictors were available.
The first step in the GAM analysis was to use single variable-based GAMs to identify the relationships between individual predictors and the euphausiid abundance in the different regions. The fitted models were evaluated based on (i) the percentage of deviance explained, (ii) the un-biased risk estimator (UBRE) or generalized cross-validation (GCV) scores and (iii) the smooth confidence region (Planque et al., 2007).
The second step was to use multiple variable-based GAMs to estimate the combined effect of more than one predictor on euphausiid abundance. High collinearity between predictors was assessed by pairwise scatterplots; Pearson's correlation coefficients and variance inflation factors with the cut-off value of 5 were used to remove collinear variables as recommended by Zuur et al. (Zuur et al., 2009). In the multiple variable-based GAMs, the predictors were chosen by a backward-elimination process for the least significant predictor based on χ2-statistic for the presence/absence models and the F-statistic for the abundance larger than zero models (Stefánsson and Pálsson, 1997). The multiple variable-based GAMs with the lowest UBRE or GCV scores were selected as the best-fitted models (Wood, 2006). The data analysis was conducted with the mgcv package version 1.7–22 in the statistical software R (Wood, 2011).
RESULTS
Sea surface temperature
Interannual variations in SSTs for all regions from 1958 to 2011 are shown in Fig. 2. All regions showed a general cooling trend during the 1960s. From the early 1970s until mid-1990s, temperatures fluctuated while being generally low. Thereafter, a general warming trend was apparent in all regions with maxima in 2003 and 2008. After 2008, surface temperatures dropped dramatically in all areas (Fig. 2).
From 1958 to 2011, the overall mean SSTs varied from ∼7.0 to 8.5°C. Area B7 was on average the coldest region (long-term average ∼5.7°C) and B5 the warmest (long-term average ∼9.5°C).
Chlorophyll a
In the areas south of Iceland, the phytoplankton spring bloom began between late April and mid-May (i.e. in Weeks ∼16–20) from 1998 to 2009, whereas after 2009 it started between mid-May and early June (i.e. in Weeks ∼19–22) (Fig. 3A).
In most areas, there was a more or less gradual increase in the yearly average Chl a concentrations from the beginning of the time series (Fig. 3B). In areas B7 and B6, Chl a peaked around 2003 and 2004, respectively, while for the other areas concentrations were more stable (Fig. 3B). Since 2007, a general increase in Chl a concentrations was apparent in all areas.
In the CPR areas, the long-term average of Chl a for the whole series (1998–2011) fluctuated between 1.62 mg Chl a m−3 in area B5 to 2.01 mg Chl a m−3 in area A6. On average, in spring 2.12 mg Chl a m−3 were measured along the Selvogsbanki transect.
Due to cloud cover, no data on the surface Chl a concentrations could be obtained for area B4 and, therefore, neither the onset of the phytoplankton spring bloom nor the long-term variability of Chl a can be illustrated.
Euphausiids: seasonal changes
The seasonal variations in the abundance of euphausiids were generally characterized by low abundance during winter and high abundance during summer months (Fig. 4, right panels). In area A6, numbers started to increase in May–June with the highest values being observed in June and August. Numbers remained high until September when they started to decrease and had reached low winter values by October. In areas B7 and B6, a significant increase had taken place already in March. In these regions, the highest values were observed in June and July. The main decrease in numbers occurred after August and by November abundance had returned to low winter values. In area B5, the main increase took place from April to June. Numbers peaked in June and remained high until August–September, when they started to decrease and had reached low winter values again in November–December. In area B4, the main increase in numbers occurred in May and June. The maximum was reached in June. Numbers remained relatively high until September, but by November they had returned to low winter values (Fig. 4, right panel).
The description above is based on the long-term averages (Fig. 4, right panels). However, there is considerable variability in the seasonal pattern among years (Fig. 4, left panels). Thus, in several years, there is clearly only one maximum in numbers during the summer (e.g. in the mid-1990s in area B7), whereas in other years two maxima are evident (e.g. in the early 1980s in area B7).
The mean annual number of euphausiids was highest in CPR areas B7, B6 and B5 (∼4–6 euphausiids m−3) and lowest in area B4 (∼3 euphausiids m−3). A gradual decrease in mean annual numbers is evident from west to east, i.e. from east of Greenland (B7) to the west of the Faroe Islands (B4) (Figs 1 and 4).
Euphausiids: long-term changes
The abundance of euphausiids during 50 years of sampling with the CPR is shown in Fig. 5. The values shown are standardized annual averages. In all areas, the numbers fluctuated considerably, while generally showing a decreasing trend in all the CPR areas apart from B6 (simple linear regressions: B7 r2 = 0.11, P < 0.05; A6 r2 = 0.25, P < 0.001; B5 r2 = 0.22, P < 0.001 and B4 r2 = 0.22, P < 0.001). It is noteworthy that, in all areas, numbers were relatively low at the end of the series and in three areas (B6, B5 and B4), the lowest values were in fact observed during the last year of the series (2007).
On the Selvogsbanki transect, the long-term changes of larvae in late May were rather similar to those of the eggs (see smoothed curved in Fig. 6A). Numbers of both eggs and larvae were high at the start of the series (1990), and maxima of both eggs and larvae were observed ∼1993–1994, 2000 and 2005. Around 2008–2009 the numbers of larvae peaked again, while the number of eggs was at a low.
Linear regression analysis showed no long-term trend for the number of eggs between 1990 and 2011 (P > 0.05) (Fig. 6A), while the number of larvae increased significantly during the same period (P < 0.05). Thus, at least for the euphausiid larvae, the Icelandic time series (Fig. 5) does not reflect the same overall decreasing trend as the CPR series (Fig. 6), probably mainly because of its much shorter time span.
On average, the number of eggs along the Selvogsbanki transect tended to increase from inshore to offshore from 1990 to 2011 (ANOVA, P < 0.001), with highest numbers observed at Stations 4 and 5 (Tukey's HSD, P < 0.001; Fig. 6B). In contrast, the abundance of larvae was similar along the transect (ANOVA, P > 0.05).
Single variable-based GAMs
For all the CPR areas combined, both the presence/absence and abundance larger than zero models showed similar results (Fig. 7 and Table I).
All CPR areas . | Temperature . | Salinity . | Chlorophyll a . | OPB . | NAO . |
---|---|---|---|---|---|
Presence/absence | |||||
% Deviance | 1.5 | 3.2 | 5.4 | 0.2 | 0.3 |
UBRE | −0.02 | −0.04 | −0.26 | −0.14 | −0.01 |
P (χ2) | 0.05 | <0.001 | <0.01 | 0.82 | 0.61 |
Abundance >0 | |||||
% Deviance | 2.1 | 8.7 | 14.4 | 3.4 | 3.3 |
GCV | 2.92 | 2.74 | 2.54 | 3.12 | 2.90 |
P (F) | <0.01** | <0.001 | <0.001 | <0.01 | <0.01 |
All CPR areas . | Temperature . | Salinity . | Chlorophyll a . | OPB . | NAO . |
---|---|---|---|---|---|
Presence/absence | |||||
% Deviance | 1.5 | 3.2 | 5.4 | 0.2 | 0.3 |
UBRE | −0.02 | −0.04 | −0.26 | −0.14 | −0.01 |
P (χ2) | 0.05 | <0.001 | <0.01 | 0.82 | 0.61 |
Abundance >0 | |||||
% Deviance | 2.1 | 8.7 | 14.4 | 3.4 | 3.3 |
GCV | 2.92 | 2.74 | 2.54 | 3.12 | 2.90 |
P (F) | <0.01** | <0.001 | <0.001 | <0.01 | <0.01 |
Predictors used in the study were SST, salinity, surface Chl a concentration, onset of the phytoplankton spring bloom (OPB) and NAO winter index. For each predictor, percentage of deviance explained (% Deviance), UBRE or GCV scores, chi-squared (χ2) or F-test (F) significances are given. Unless otherwise stated with asterisks, 3 effective degrees of freedom (edf) were used.
**2 edf.
All CPR areas . | Temperature . | Salinity . | Chlorophyll a . | OPB . | NAO . |
---|---|---|---|---|---|
Presence/absence | |||||
% Deviance | 1.5 | 3.2 | 5.4 | 0.2 | 0.3 |
UBRE | −0.02 | −0.04 | −0.26 | −0.14 | −0.01 |
P (χ2) | 0.05 | <0.001 | <0.01 | 0.82 | 0.61 |
Abundance >0 | |||||
% Deviance | 2.1 | 8.7 | 14.4 | 3.4 | 3.3 |
GCV | 2.92 | 2.74 | 2.54 | 3.12 | 2.90 |
P (F) | <0.01** | <0.001 | <0.001 | <0.01 | <0.01 |
All CPR areas . | Temperature . | Salinity . | Chlorophyll a . | OPB . | NAO . |
---|---|---|---|---|---|
Presence/absence | |||||
% Deviance | 1.5 | 3.2 | 5.4 | 0.2 | 0.3 |
UBRE | −0.02 | −0.04 | −0.26 | −0.14 | −0.01 |
P (χ2) | 0.05 | <0.001 | <0.01 | 0.82 | 0.61 |
Abundance >0 | |||||
% Deviance | 2.1 | 8.7 | 14.4 | 3.4 | 3.3 |
GCV | 2.92 | 2.74 | 2.54 | 3.12 | 2.90 |
P (F) | <0.01** | <0.001 | <0.001 | <0.01 | <0.01 |
Predictors used in the study were SST, salinity, surface Chl a concentration, onset of the phytoplankton spring bloom (OPB) and NAO winter index. For each predictor, percentage of deviance explained (% Deviance), UBRE or GCV scores, chi-squared (χ2) or F-test (F) significances are given. Unless otherwise stated with asterisks, 3 effective degrees of freedom (edf) were used.
**2 edf.
For the presence/absence models, Chl a was the strongest associated predictor, explaining overall 5.4% of the presence/absence recorded in recent years, followed by salinity (3.2%) and temperature (1.5%) (Table I). The probability of euphausiids being present generally increased with increasing Chl a concentrations, while the opposite was true for salinity (Fig. 7A and B). The relationship between euphausiid occurrence and temperature was more complex, the probability of presence showing a decreasing trend from ∼4 to ∼8°C and an increasing trend after that (Fig. 7C). It should be noted that the confidence limits were wide at the highest Chl a concentrations and lowest salinities due to limited data.
For the abundance larger than zero models, Chl a was the strongest associated explanatory variable (14.4%, Table I), with salinity (8.7%) and temperature (2.1%) being less important. As for the presence–absence models, the abundance of euphausiids where it exceeded zero was generally positively related to Chl a up to ∼1.5 mg Chl am−3, and negatively related to salinity (Fig. 7D and E). For temperature, the model indicated a decreasing trend from ∼4 to ∼8°C, and an increasing trend after that (Fig. 7F).
On the Selvogsbanki transect, the probability of finding euphausiid eggs in the samples was mainly explained by salinity (12.6%, Table II), with lowest occurrence of eggs being recorded at salinities ∼34.6 (Fig. 8A). Confidence limits are, however, high at the lower salinity values due to limited data. The abundance of eggs when it was non-zero along the Selvogsbanki transect showed no significant relationship with any of the predictors (Table II).
Selvogsbanki . | Temperature . | Salinity . | Chlorophyll a . | OPB . | NAO . |
---|---|---|---|---|---|
Presence/absence | |||||
Eggs | |||||
% Deviance | 2.3 | 12.6 | 7.0 | 6.5 | 0.5 |
UBRE | 0.51 | 0.37 | 0.45 | 0.45 | 0.54 |
P (χ2) | 0.62 | 0.05 | 0.14 | 0.35 | 0.92 |
Larvae | |||||
% Deviance | 1.7 | 33.8 | 6.4 | 35.7 | 58.2 |
UBRE | −0.48 | −0.54 | −0.50 | −0.60 | −0.71 |
P (χ2) | 0.49 | 0.50 | 0.30* | 0.33* | 0.42 |
Abundance >0 | |||||
Eggs | |||||
% Deviance | 2.3 | 11.0 | 8.6 | 9.3 | 11.6 |
GCV | 2.70 | 2.46 | 2.53 | 2.51 | 2.44 |
P (F) | 0.88 | 0.34 | 0.47 | 0.43 | 0.32 |
Larvae | |||||
% Deviance | 14.8 | 8.2 | 1.0 | 5.3 | 10.1 |
GCV | 1.54 | 1.66 | 1.79 | 1.71 | 1.62 |
P (F) | 0.02 | 0.16 | 0.89 | 0.34 | 0.09 |
Selvogsbanki . | Temperature . | Salinity . | Chlorophyll a . | OPB . | NAO . |
---|---|---|---|---|---|
Presence/absence | |||||
Eggs | |||||
% Deviance | 2.3 | 12.6 | 7.0 | 6.5 | 0.5 |
UBRE | 0.51 | 0.37 | 0.45 | 0.45 | 0.54 |
P (χ2) | 0.62 | 0.05 | 0.14 | 0.35 | 0.92 |
Larvae | |||||
% Deviance | 1.7 | 33.8 | 6.4 | 35.7 | 58.2 |
UBRE | −0.48 | −0.54 | −0.50 | −0.60 | −0.71 |
P (χ2) | 0.49 | 0.50 | 0.30* | 0.33* | 0.42 |
Abundance >0 | |||||
Eggs | |||||
% Deviance | 2.3 | 11.0 | 8.6 | 9.3 | 11.6 |
GCV | 2.70 | 2.46 | 2.53 | 2.51 | 2.44 |
P (F) | 0.88 | 0.34 | 0.47 | 0.43 | 0.32 |
Larvae | |||||
% Deviance | 14.8 | 8.2 | 1.0 | 5.3 | 10.1 |
GCV | 1.54 | 1.66 | 1.79 | 1.71 | 1.62 |
P (F) | 0.02 | 0.16 | 0.89 | 0.34 | 0.09 |
Predictors used in the study were temperature, salinity, Chl a concentration averaged from 0 to 50 m, onset of the phytoplankton spring bloom (OPB) and NAO winter index. For each predictor, percentage of deviance explained (% Deviance), UBRE or GCV scores, chi-squared (χ2) or F-test (F) significances are given. Unless otherwise stated with asterisks, 3 effective degrees of freedom (edf) were used.
*1 edf.
Selvogsbanki . | Temperature . | Salinity . | Chlorophyll a . | OPB . | NAO . |
---|---|---|---|---|---|
Presence/absence | |||||
Eggs | |||||
% Deviance | 2.3 | 12.6 | 7.0 | 6.5 | 0.5 |
UBRE | 0.51 | 0.37 | 0.45 | 0.45 | 0.54 |
P (χ2) | 0.62 | 0.05 | 0.14 | 0.35 | 0.92 |
Larvae | |||||
% Deviance | 1.7 | 33.8 | 6.4 | 35.7 | 58.2 |
UBRE | −0.48 | −0.54 | −0.50 | −0.60 | −0.71 |
P (χ2) | 0.49 | 0.50 | 0.30* | 0.33* | 0.42 |
Abundance >0 | |||||
Eggs | |||||
% Deviance | 2.3 | 11.0 | 8.6 | 9.3 | 11.6 |
GCV | 2.70 | 2.46 | 2.53 | 2.51 | 2.44 |
P (F) | 0.88 | 0.34 | 0.47 | 0.43 | 0.32 |
Larvae | |||||
% Deviance | 14.8 | 8.2 | 1.0 | 5.3 | 10.1 |
GCV | 1.54 | 1.66 | 1.79 | 1.71 | 1.62 |
P (F) | 0.02 | 0.16 | 0.89 | 0.34 | 0.09 |
Selvogsbanki . | Temperature . | Salinity . | Chlorophyll a . | OPB . | NAO . |
---|---|---|---|---|---|
Presence/absence | |||||
Eggs | |||||
% Deviance | 2.3 | 12.6 | 7.0 | 6.5 | 0.5 |
UBRE | 0.51 | 0.37 | 0.45 | 0.45 | 0.54 |
P (χ2) | 0.62 | 0.05 | 0.14 | 0.35 | 0.92 |
Larvae | |||||
% Deviance | 1.7 | 33.8 | 6.4 | 35.7 | 58.2 |
UBRE | −0.48 | −0.54 | −0.50 | −0.60 | −0.71 |
P (χ2) | 0.49 | 0.50 | 0.30* | 0.33* | 0.42 |
Abundance >0 | |||||
Eggs | |||||
% Deviance | 2.3 | 11.0 | 8.6 | 9.3 | 11.6 |
GCV | 2.70 | 2.46 | 2.53 | 2.51 | 2.44 |
P (F) | 0.88 | 0.34 | 0.47 | 0.43 | 0.32 |
Larvae | |||||
% Deviance | 14.8 | 8.2 | 1.0 | 5.3 | 10.1 |
GCV | 1.54 | 1.66 | 1.79 | 1.71 | 1.62 |
P (F) | 0.02 | 0.16 | 0.89 | 0.34 | 0.09 |
Predictors used in the study were temperature, salinity, Chl a concentration averaged from 0 to 50 m, onset of the phytoplankton spring bloom (OPB) and NAO winter index. For each predictor, percentage of deviance explained (% Deviance), UBRE or GCV scores, chi-squared (χ2) or F-test (F) significances are given. Unless otherwise stated with asterisks, 3 effective degrees of freedom (edf) were used.
*1 edf.
The probability of euphausiid larvae being present along the Selvogsbanki transect exhibited no significant relationship with any of the predictors (Table II). The abundance of larvae was best explained by temperature (14.8%, Table II) following a dome-shaped functional form, peaking at temperatures between ∼8.5 and 9°C (Fig. 8B).
Multiple variable-based GAMs
GAMs based on multiple explanatory variables from the CPR areas were established. For the combined data set, the multiple variable-based GAMs significantly improved the predictions of euphausiid abundance (Table III), when compared with the single variable-based GAMs (Table I). All the environmental predictors considered, except the onset of the phytoplankton spring bloom, significantly contributed to the overall models (Table III). The multiple variable-based GAMs explained, respectively, ∼17.5 and 27.6% of the interannual variability in euphausiid presence/absence and abundance, where it was non-zero (Table III).
CPR areas . | Temperature . | Salinity . | Chlorophyll a . | OPB . | NAO . | % Deviance . | UBRE/GCV . |
---|---|---|---|---|---|---|---|
Presence/absence | |||||||
All areas | 0.02 | 0.008 | 0.14 | – | 0.008 | 17.5 | −0.45 |
B7 | 0.05* | – | 0.65* | – | – | 17.0 | −0.56 |
B6 | – | – | – | – | – | – | – |
A6 | <0.001 | – | – | – | – | 4.0 | −0.17 |
B5 | 0.004 | – | – | 0.037 | – | 15.4 | −0.20 |
B4 | <0.001 | – | – | – | <0.001 | 8.0 | −0.05 |
Abundance >0 | |||||||
All areas | 0.0025 | <0.001 | <0.001 | – | <0.001 | 27.6 | 2.33 |
B7 | – | – | <0.001 | 0.209 | – | 41.4 | 1.27 |
B6 | – | <0.001 | 0.002 | 0.021 | 0.007 | 57.0 | 2.58 |
A6 | – | 0.011 | 0.033 | <0.001 | 0.002 | 65.8 | 1.51 |
B5 | 0.015 | – | <0.001 | – | – | 48.2 | 1.49 |
B4 | <0.001 | – | – | – | – | 27.2 | 1.36 |
CPR areas . | Temperature . | Salinity . | Chlorophyll a . | OPB . | NAO . | % Deviance . | UBRE/GCV . |
---|---|---|---|---|---|---|---|
Presence/absence | |||||||
All areas | 0.02 | 0.008 | 0.14 | – | 0.008 | 17.5 | −0.45 |
B7 | 0.05* | – | 0.65* | – | – | 17.0 | −0.56 |
B6 | – | – | – | – | – | – | – |
A6 | <0.001 | – | – | – | – | 4.0 | −0.17 |
B5 | 0.004 | – | – | 0.037 | – | 15.4 | −0.20 |
B4 | <0.001 | – | – | – | <0.001 | 8.0 | −0.05 |
Abundance >0 | |||||||
All areas | 0.0025 | <0.001 | <0.001 | – | <0.001 | 27.6 | 2.33 |
B7 | – | – | <0.001 | 0.209 | – | 41.4 | 1.27 |
B6 | – | <0.001 | 0.002 | 0.021 | 0.007 | 57.0 | 2.58 |
A6 | – | 0.011 | 0.033 | <0.001 | 0.002 | 65.8 | 1.51 |
B5 | 0.015 | – | <0.001 | – | – | 48.2 | 1.49 |
B4 | <0.001 | – | – | – | – | 27.2 | 1.36 |
Predictors used in the study were surface temperature, salinity, surface Chl a concentration, onset of the phytoplankton spring bloom (OPB) and NAO winter index. Percentage of deviance explained (% Deviance), UBRE or GCV scores are shown for the best-fitted GAMs. Significance is given for each predictor used in the models. Unless otherwise stated with asterisks, 3 effective degrees of freedom (edf) were used. Presence/absence of GAMs could not be constructed for area B6.
*1 edf.
CPR areas . | Temperature . | Salinity . | Chlorophyll a . | OPB . | NAO . | % Deviance . | UBRE/GCV . |
---|---|---|---|---|---|---|---|
Presence/absence | |||||||
All areas | 0.02 | 0.008 | 0.14 | – | 0.008 | 17.5 | −0.45 |
B7 | 0.05* | – | 0.65* | – | – | 17.0 | −0.56 |
B6 | – | – | – | – | – | – | – |
A6 | <0.001 | – | – | – | – | 4.0 | −0.17 |
B5 | 0.004 | – | – | 0.037 | – | 15.4 | −0.20 |
B4 | <0.001 | – | – | – | <0.001 | 8.0 | −0.05 |
Abundance >0 | |||||||
All areas | 0.0025 | <0.001 | <0.001 | – | <0.001 | 27.6 | 2.33 |
B7 | – | – | <0.001 | 0.209 | – | 41.4 | 1.27 |
B6 | – | <0.001 | 0.002 | 0.021 | 0.007 | 57.0 | 2.58 |
A6 | – | 0.011 | 0.033 | <0.001 | 0.002 | 65.8 | 1.51 |
B5 | 0.015 | – | <0.001 | – | – | 48.2 | 1.49 |
B4 | <0.001 | – | – | – | – | 27.2 | 1.36 |
CPR areas . | Temperature . | Salinity . | Chlorophyll a . | OPB . | NAO . | % Deviance . | UBRE/GCV . |
---|---|---|---|---|---|---|---|
Presence/absence | |||||||
All areas | 0.02 | 0.008 | 0.14 | – | 0.008 | 17.5 | −0.45 |
B7 | 0.05* | – | 0.65* | – | – | 17.0 | −0.56 |
B6 | – | – | – | – | – | – | – |
A6 | <0.001 | – | – | – | – | 4.0 | −0.17 |
B5 | 0.004 | – | – | 0.037 | – | 15.4 | −0.20 |
B4 | <0.001 | – | – | – | <0.001 | 8.0 | −0.05 |
Abundance >0 | |||||||
All areas | 0.0025 | <0.001 | <0.001 | – | <0.001 | 27.6 | 2.33 |
B7 | – | – | <0.001 | 0.209 | – | 41.4 | 1.27 |
B6 | – | <0.001 | 0.002 | 0.021 | 0.007 | 57.0 | 2.58 |
A6 | – | 0.011 | 0.033 | <0.001 | 0.002 | 65.8 | 1.51 |
B5 | 0.015 | – | <0.001 | – | – | 48.2 | 1.49 |
B4 | <0.001 | – | – | – | – | 27.2 | 1.36 |
Predictors used in the study were surface temperature, salinity, surface Chl a concentration, onset of the phytoplankton spring bloom (OPB) and NAO winter index. Percentage of deviance explained (% Deviance), UBRE or GCV scores are shown for the best-fitted GAMs. Significance is given for each predictor used in the models. Unless otherwise stated with asterisks, 3 effective degrees of freedom (edf) were used. Presence/absence of GAMs could not be constructed for area B6.
*1 edf.
Multiple variable-based GAMs were also set up for the CPR areas separately. For the presence/absence models, temperature came out as a significant explanatory variable in majority of the models (B7, A6, B5 and B4) (Table III). Depending on regions, presence of euphausiids was also related to Chl a (B7), onset of the phytoplankton spring bloom (B5) or NAO (B4).
For the abundance larger than zero models, the relative role of the different explanatory variables also varied by CPR areas. East of Greenland (B7), the abundance of euphausiids was related to Chl a concentration and the onset of the phytoplankton spring bloom, whereas south and southwest of Iceland (B6 and A6) also salinity and the NAO were relevant. Farther east, abundance was either related to temperature and Chl a (B5) or temperature alone (B4).
For the more coastal areas south of Iceland (Selvogsbanki transect), the multiple variable-based GAMs established a relationship between the presence/absence of eggs and salinity and Chl a (together explaining ∼25% of the interannual variability), while no relationship could predict the presence/absence of larvae (Table IV).
Selvogsbanki . | Temperature . | Salinity . | Chlorophyll a . | OPB . | NAO . | % Deviance . | UBRE/GCV . |
---|---|---|---|---|---|---|---|
Presence/absence | |||||||
Eggs | – | 0.02 | 0.04 | – | – | 24.9 | 0.3 |
Larvae | – | – | – | – | – | – | – |
Abundance >0 | |||||||
Eggs | – | – | – | – | – | – | – |
Larvae | 0.006** | – | – | 0.03 | 0.006 | 35.6 | 1.5 |
Selvogsbanki . | Temperature . | Salinity . | Chlorophyll a . | OPB . | NAO . | % Deviance . | UBRE/GCV . |
---|---|---|---|---|---|---|---|
Presence/absence | |||||||
Eggs | – | 0.02 | 0.04 | – | – | 24.9 | 0.3 |
Larvae | – | – | – | – | – | – | – |
Abundance >0 | |||||||
Eggs | – | – | – | – | – | – | – |
Larvae | 0.006** | – | – | 0.03 | 0.006 | 35.6 | 1.5 |
Predictors used in this study were temperature, salinity, Chl a concentration averaged from 0 to 50 m, onset of the phytoplankton spring bloom (OPB) and NAO winter index. Percentage of deviance explained (% Deviance), UBRE or GCV scores are shown for the best-fitted GAMs. Significance is given for each predictor used in the models. Unless otherwise stated with asterisks, 3 effective degrees of freedom (edf) were used. Presence/absence of GAMs could not be constructed for larvae and abundance larger than zero models could not be constructed for eggs.
**2 edf.
Selvogsbanki . | Temperature . | Salinity . | Chlorophyll a . | OPB . | NAO . | % Deviance . | UBRE/GCV . |
---|---|---|---|---|---|---|---|
Presence/absence | |||||||
Eggs | – | 0.02 | 0.04 | – | – | 24.9 | 0.3 |
Larvae | – | – | – | – | – | – | – |
Abundance >0 | |||||||
Eggs | – | – | – | – | – | – | – |
Larvae | 0.006** | – | – | 0.03 | 0.006 | 35.6 | 1.5 |
Selvogsbanki . | Temperature . | Salinity . | Chlorophyll a . | OPB . | NAO . | % Deviance . | UBRE/GCV . |
---|---|---|---|---|---|---|---|
Presence/absence | |||||||
Eggs | – | 0.02 | 0.04 | – | – | 24.9 | 0.3 |
Larvae | – | – | – | – | – | – | – |
Abundance >0 | |||||||
Eggs | – | – | – | – | – | – | – |
Larvae | 0.006** | – | – | 0.03 | 0.006 | 35.6 | 1.5 |
Predictors used in this study were temperature, salinity, Chl a concentration averaged from 0 to 50 m, onset of the phytoplankton spring bloom (OPB) and NAO winter index. Percentage of deviance explained (% Deviance), UBRE or GCV scores are shown for the best-fitted GAMs. Significance is given for each predictor used in the models. Unless otherwise stated with asterisks, 3 effective degrees of freedom (edf) were used. Presence/absence of GAMs could not be constructed for larvae and abundance larger than zero models could not be constructed for eggs.
**2 edf.
The multiple variable-based GAMs explained ∼36% of the abundance of larvae where they exceeded zero, while no relationship could predict the abundance of eggs. The number of larvae was mainly related to temperature, the onset of the phytoplankton spring bloom and NAO (Table IV).
DISCUSSION
Seasonal cycle
In the open waters of the Northeastern Atlantic, euphausiids reached maximum abundance from May to August (Fig. 4). This confirms previous findings of Lindley (Lindley, 1978) for a wider area of the North Atlantic and for the shelf areas south (Gislason and Astthorsson, 1995) and north of Iceland (Astthorsson and Gislason, 1997a; Gislason and Astthorsson, 1998). Year-to-year changes in seasonal abundance of euphausiids may be due to different life history traits of different dominant species. Lindley (Lindley, 1978) found that, in the warm Atlantic waters, T. longicaudata has two annual generations, while only one generation in the colder waters. In the present material, species were not distinguished so it is impossible to evaluate if a particular species has more than one annual generation, e.g. in warm years. Nonetheless, our results show two annual peaks during the summer in some years (Fig. 4, left panels, B7 and A6).
In agreement with Letessier et al. (Letessier et al., 2009), we found that euphausiids in surface layers gradually decreased from the east of Greenland to the west of the Faroe Islands (Fig. 4). Presumably, this pattern reflects the higher production of euphausiids in the western regions of the North Atlantic as opposed to the eastern regions. As suggested by Saunders et al. (Saunders et al., 2007), the complex water mass dynamics in the Irminger Sea and over the Reykjanes ridge (Holliday et al., 2006), characterized by relatively warm temperatures and high food availability, may allow euphausiids to grow larger and to have a longer life span in these regions compared with other regions. As larger females produce more eggs (Cuzin-Roudy, 2000), this would then contribute to greater euphausiid productivity in these regions.
Physical parameters, such as local topography and water mass dynamics, which in the area of study are complex and in some regions influenced by freshwater run-off from rivers/glaciers, may influence euphausiid abundance and seasonal occurrence in different ways (Dalpadado et al., 2008a; Buchholz et al., 2010; McGinty et al., 2011). In agreement with the study of Lindley (Lindley, 1980), we found a low number of euphausiids in the surface layer in the winter months (Fig. 4). The relatively low abundance in the winter months may be related to high mortality during a season when food is scarce and also the fact that euphausiids tend to stay relatively deep in the water column during the winter (Lindley, 1980; Mackas et al., 2012).
Long-term changes
Our findings show that, in most of the open Northeastern Atlantic, euphausiid abundance decreased from 1958 to 2007 (Fig. 5). Similarly, Beaugrand and Reid (Beaugrand and Reid, 2003) and Beaugrand et al. (Beaugrand et al., 2003) reported euphausiids to have declined during recent decades in the North Sea and adjacent seas. Our study demonstrates that this declining trend is widespread in the whole Northeastern Atlantic and still ongoing. Although the decrease in area B6 was not significant (Fig. 5), the lowest numbers were nevertheless observed during the last years of the time series, thus indicating a decline there also.
While the CPR data showed that total numbers of euphausiids generally declined from 1958 to 2007 (Fig. 5), the number of larvae along the Selvogsbanki transect showed an increasing trend from 1990 to 2011 (Fig. 6A). The apparent discrepancy between the two time series is most likely related to the much shorter time span of the Selvogsbanki series. However, such an opposite trend may also be related to differences in sampling techniques (WP2 nets versus CPR) which makes a comparison difficult (Beare et al., 2000), as well as to differences in the frequency and spatial extent of the sampling. Thus, the Icelandic time series refers to the spring period (May–June) only, whereas the CPR data cover the whole year. Additionally, the Selvogsbanki data were mostly collected on the south Icelandic shelf, while the CPR data were mainly collected off shelf (Fig. 1). The different long-term trends exhibited by the CPR data and the Icelandic spring survey data could thus reflect onshore–offshore differences in the interannual variability and/or in species composition. Indeed, T. inermis tends to be most abundant on the Icelandic shelves, M. norvegica mainly over the shelve edges, while T. longicaudata is mainly found in the oceanic areas (Einarsson, 1945; Mauchline and Fisher, 1969). Finally, euphausiid patchiness and the fact that they spawn in multiple batches (Cuzin-Roudy, 2000) may have a greater influence on the data from the relatively small and infrequently sampled Selvogsbanki area compared with the larger and more frequently sampled CPR areas.
The spatial distribution of euphausiid eggs along the Selvogsbanki transect showed a distinctive pattern of increasing abundance from onshore to offshore, while the larvae were more or less evenly distributed along the transect (Fig. 6B). This probably reflects that the distribution of the larvae was affected by horizontal advection and mixing processes for a longer period of time than the eggs that stay for a much shorter time as plankton. Topographical dissimilarities and different water masses characterized the stations along the transect, Stations 1–3 being influenced by freshwater run-off from land, as opposed to Stations 4 and 5, that are more oceanic and mainly influenced by Atlantic waters (Stefánsson and Ólafsson, 1991; Gislason and Astthorsson, 2004). Such cross-shelf environmental differences could partly explain the variability illustrated in Fig. 6. The increasing number of eggs from onshore to offshore probably reflects the distribution of spawning adults with spawning mainly taking place near the shelf edge.
Drivers of euphausiid abundance
The single variable-based GAMs were designed to study the effects of individual explanatory variables on euphausiid abundance. For the combined CPR data set, these analyses showed that phytoplankton biomass was the single most important factor affecting the long-term changes of the euphausiids (Table I). The importance of phytoplankton for the growth and development of euphausiids has been demonstrated in several other studies (see Atkinson et al., 2008 for a review). In our study, the highest abundance of euphausiids was associated with Chl a concentrations of >1.5 mg m−3 (Fig. 7D). Letessier et al. (Letessier et al., 2009), on the other hand, found temperature to be the single most important factor influencing euphausiid numerical abundance, with Chl a playing only a minor role. The reason for this is probably related to the fact that the study of Letessier et al. (Letessier et al., 2009) spans the whole Atlantic Ocean from polar to tropical latitudes, so the temperature range is much larger than in the present study. As noted by Zarauz et al. (Zarauz et al., 2007), GAMs are only statistical models and unable to identify causal relationships. The effects of temperature in our study may thus be masked by the effects of temperature on phytoplankton spring bloom development.
The multiple variable-based GAMs showed that euphausiid abundance is affected by different factors in different areas, as also observed by McGinty et al. (McGinty et al., 2011) for calanoid copepods. In the west, Chl a is the major predictor of euphausiid abundance, whereas in the east temperature appears most important (Table III). The difference could possibly be related to differences in species composition between west and east, with different species having different requirements/responses to changes in climate and food environment. The limited information that is available on the distribution of euphausiids in the North Atlantic indicates T. longicaudata as the numerically dominant species in all regions, while M. norvegica appears more abundant in the eastern areas (Einarsson, 1945; Lindley, 1982b). Thus, it is possible that the different responses to environmental factors partly reflect a longitudinal gradient in euphausiid species composition.
Earlier studies have shown that zooplankton composition on the Selvogsbanki transect is mainly governed by salinity and phytoplankton biomass (Gislason et al., 2009). In this study, the probability of occurrence of euphausiid eggs at Selvogsbanki was positively related to salinity >35 (Table II and Fig. 8A), indicating that euphausiids prefer spawning in areas of relatively high salinities. In fact, Einarsson (Einarsson, 1945) reported the main spawning of M. norvegica and T. longicaudata to occur in areas with pure Atlantic water.
The present study also demonstrates that for the recent warm period (1998–2011), changes in temperature, NAO and in the timing of the phytoplankton bloom influenced the numbers of larvae on Selvogsbanki (Table IV). Lower wind stress during negative NAO may result in a stronger stratification, triggering the phytoplankton bloom (Henson et al., 2009) and in turn an increase in euphausiid numbers. According to Henson et al. (Henson et al., 2009), the timing of the phytoplankton spring bloom is linked to the NAO, with a relatively early bloom when the NAO is negative.
According to Hátún et al. (Hátún et al., 2009), a weak North Atlantic subpolar gyre may lead to an increase of saline and warm water south of Iceland, which in turn may positively affect the annual mean phytoplankton biomass and negatively the abundance of Calanus finmarchicus in the Irminger Sea. We observed high Chl a concentrations (Fig. 3B), a delay in the phytoplankton spring bloom (Fig. 3A) and low euphausiid numbers (Fig. 5) in years when the subpolar gyre was weak (2000–2010) (Hátún et al., 2009; Larsen et al., 2012). As copepods are part of the diet of euphausiids (Falk-Petersen et al., 2000; Dalpadado et al., 2008b), the decline in the stocks of C. finmarchicus may have also contributed to the decline in euphausiid populations.
As stated above, T. longicaudata is the most abundant euphausiid species in the North Atlantic. Einarsson (Einarsson, 1945) proposed that the upper temperature limit at which the density of T. longicaudata is limited is 15°C, and reproduction would be optimal at a temperature of <12°C. Given the significant positive effect of temperature on euphausiid densities (Fig. 7F), the observed decrease in densities in the CPR areas in the 2000s (Fig. 5) while surface temperatures increased (Fig. 2) is somewhat surprising. As Chl a appears to affect more the long-term variability of the euphausiids than temperature (Table I), it may be argued that the effects of temperature were overridden by the effect of Chl a. However, the fact that Chl a was generally increasing in the CPR areas in the 2000s (Fig. 3B) while the abundance of euphausiids was decreasing (Fig. 5) seems to contradict this. In addition to temperature and Chl a, the onset of the spring bloom also affected the euphausiid long-term changes (Tables I and III). In all the CPR areas, there was a gradual delay in the onset of the spring bloom from around 2006 (Fig. 3A). In the beginning of the time series, blooming began between late April and mid-May, whereas in the end between mid-May and early June. The delay may have translated into a weaker synchrony between the timing of first feeding euphausiid larvae in surface layers and the timing of the spring bloom. We believe that the warming of the surface waters during recent years may thus have been a factor resulting in a weaker temporal synchrony of the developing young euphausiids with the phytoplankton bloom that in turn may have led to the reduced euphausiid population sizes, thus suggesting that large-scale climatic factors may have altered conditions during the most important period in the life cycle.
We acknowledge that top-down effects also play a role in regulating abundance. For example, the annual consumption of euphausiids by fin whales, estimated to be ∼2.5 million tonnes (Sigurjónsson and Víkingsson, 1997), may be sufficient to affect the stock sizes in the region. Predation pressure from fin whales may have increased due to the 10% increase in fin whale numbers from 1987 to 2001 in the area between Iceland and Greenland (Víkingsson et al., 2009). A recent study conducted in Icelandic waters on the diet composition of minke whales has shown a lower incidence of euphausiids in their diet in the mid-2000s than in the early 1980s (Víkingsson et al., 2014). This may, in turn, reflect the declining number of euphausiids in the area.
Changes in the abundance and distribution pattern of several marine species around Iceland have been related to climatic changes (Astthorsson et al., 2007), for instance, the northward migration of many southern rare and vagrant species (Valdimarsson et al., 2012) and the unprecedented feeding migrations of mackerel to the sea areas around Iceland during summer in recent years (Astthorsson et al., 2012). We observed a more or less general decline of euphausiids in the open sea areas southwest, south and east of Iceland, partly dictated by the weakened synchrony between the developing young euphausiids and the phytoplankton spring bloom. These findings are in general agreement with other studies on the long-term changes of zooplankton, such as C. finmarchicus, in the open waters of the North Atlantic (e.g. Planque and Taylor, 1998; Hátún et al., 2009). They support the hypothesis that bottom-up regulation, driven by climate change forcing, is a major factor affecting interannual changes in the abundance of euphausiids.
As outlined in the section Introduction, euphausiids are a key group of organisms in the North Atlantic ecosystem where they link production at lower trophic levels to top predators, some of which are commercially utilized. Still the availability of time series data on their abundance is hampered by the fact that the samplers used are primarily designed to capture smaller zooplankton taxa. Future studies should aim at developing more efficient sampling methodologies to monitor euphausiid standing stock, e.g. by applying fine meshed pelagic trawls and acoustic techniques. Given the ecological role of euphausiids as mediators of matter and energy flux between key groups of the ecosystem, the further studies should monitor euphausiid abundance at the species level, focusing on trophic interactions, and the underlying mechanisms for the observed relationships, in order to increase understanding of how climate change will likely impact the marine ecosystem.
FUNDING
This work was supported by the Marine Research Institute (project 15.19) and by the EURO-BASIN—European Union Basin-scale Analysis, Synthesis and Integration (FP7 Contract No. 264933). Funding for A.S.A.F. was provided by the Norden Top-level Research Initiative sub-programme ‘Effect Studies and Adaptation to Climate Change’ through the Nordic Centre for Research on Marine Ecosystems and Resources under Climate Change (NorMER).
ACKNOWLEDGEMENTS
We thank the Marine Research Institute and the EURO-BASIN for funding this study. We would like to thank all the colleagues at MRI who have taken part in the collection of data and analysis of samples. Furthermore, we thank H. Valdimarsson for providing data from the Icelandic spring surveys on hydrography. We wish to thank the staff at SAHFOS for analysis of the CPR samples and for making the data available for this study. Finally, we thank the two anonymous referees for many helpful comments on an earlier version of this paper.
REFERENCES
Author notes
Corresponding editor: Roger Harris