North Atlantic Oscillation drives the annual occurrence of an isolated, peripheral population of the brown seaweed Fucus guiryi in the Western Mediterranean Sea

Sampling locations and surveys

Study sites were Punta Calaburras (36°30′28″N, 004°38′08″W) and Tarifa (36°00′03″N, 005°36′37″W) (Fig. 1A). Both locations have a similar hot-summer Mediterranean climate (group Csa, Köppen-Geiger climate classification system). Overall mean precipitation ranges from 0 or 1 mm in July, to 93 mm or 146 mm in November, at Punta Calaburras or Tarifa, respectively. Air temperature is also similar at both study sites (overall minimum mean in January of 12.5 °C in Punta Calaburras, and of 11.8 °C in Tarifa; overall maximum mean in August of 25.2 °C in Punta Calaburras vs. 23.4 °C in Tarifa). Surface water temperature ranges from 15–16 °C in March to 22–24 °C in August in both locations. Moreover, although environmental conditions at both sampling points differ at short-term, a biogeographical subregion (from the Strait of Gibraltar to Punta Calaburras) has been recognized based on the oceanographic conditions in the northern Mediterranean-Atlantic transition zone. This subregion shows a close agreement with the species composition of the littoral and sublittoral benthic communities (Bermejo et al., 2015). Consequently, we assumed that comparison of both populations of F. guiryi is worthwhile.

The sampling location at Punta Calaburras consists of rocks (schists and gneiss) protruding from the mean sea level up to 30 cm at low spring tide, occupying a surface area of ca. 150 m². The thalli, when they are present, proliferate over almost all the rocks. Consequently, all emergent rocks were visually surveyed during low spring tides, during ca. 1 h searches across all microhabitats. The survey allowed us to detect macrothalli but not possible resilient, microscopic stages. The presence of thalli of F. guiryi was checked annually in February–March, and in July-August (Table 1), from 1990 to 2015. Tarifa presents a subhorizontal sandstone platform where F. guiryi proliferates. The location was studied in midsummer 2011. To provide consistency in the comparisons with the population at Punta Calaburras, thalli of the alga growing up to 30 cm above low spring tide level were collected and analyzed for developmental instability and photosynthetic performance (see below).

We observed thalli of F. guiryi at Punta Calaburras every year of the survey in February–March, but thalli did not always persist into midsummer. The time series of the occurrence of thalli in midsummer was initially analysed taking into account two aspects: the distribution of the presence of thalli and the tendency of the occurrence. The annual occurrence distribution was checked with the test due to Prahl (1999) for a stationary Poisson process (randomly distributed throughout the time). The tendency of the annual occurrence distribution was analysed by the Laplace test (Cox & Lewis, 1978).

Analysis of the annual occurrence as a function of oceanographic and atmospheric variables

The relationship between the presence/absence in midsummer of F. guiryi thalli at Punta Calaburras from 1990 to 2015, and the temperature, rainfall, NAO, AO and SST, was addressed by binary logistic regressions, widely used for establishing relationships between environmental independent variables and the probability response of target variables (Zuur, Ieno & Smith, 2007). Temperature and rainfall mean monthly data were obtained from the Agencia Española de Meteorología (Fuengirola station, 4 km to the east of Punta Calaburras). Of the atmospheric oscillations, NAO is the most important mechanism responsible for the interannual climate variability in SW Europe, particularly during the winter (Hurrell, 1995; Hurrell et al., 2003). The AO also affects the overall mean of weather conditions in SW Europe. According to Thompson & Wallace (1998) the AO explains anomalies in the Arctic region according to the polar vortex. Thus, when the AO index is positive (characterized by a strengthening of the polar vortex), surface pressure is low in the polar region, and the opposite occurs when the index is negative. Monthly AO and NAO index values were obtained from the free-access web sites of the US National Oceanic and Atmospheric Administration, http://www.cpc.ncep.noaa.gov/products/precip/CWlink/daily_ao_index/ao.shtml and http://www.cpc.ncep.noaa.gov/products/precip/CWlink/pna/nao.shtml, respectively. Finally, the power of the Atlantic current entering the Mediterranean Sea was estimated by the SST values close to Punta Calaburras (the higher the flow of AJ water, the lower SST is; Parrilla & Kinder, 1987). Data of SST were obtained from the free-access web site from the Centro Oceanográfico de Málaga (sede Fuengirola), Instituto Español de Oceanografía, http://www.ma.ieo.es/gcc/playafuengirola_taireyagua_anomalias.xls.

For the analysis, we tested the monthly values of environmental variables from the same month as well as the overall mean figures from two to six previous months. We assessed the significance of the variables in the model using the Wald test (Wald, 1943), the calibration of the model using the Hosmer & Lemeshow test (Hosmer & Lemeshow, 1980), its discrimination capacity using the area under the curve (AUC) of the receiving operator characteristics (Lobo, Jiménez-Valverde & Real, 2008), and its explanatory power using the Nagelkerke R² (Nagelkerke, 1991).

Additionally, we used relevant probability levels to assess the environmental conditions that favoured the presence of F. guiryi thalli, the opening gap between the values considered as clearly probable (p > 0.6) or clearly improbable (p < 0.4). It must be taken into account that p = 0.5 means that the presence or the absence of the thalli have a similar probability. We then compared the correct classification rate of the models for years clearly probable and clearly improbable for a presence of thalli, and simultaneously we identified the levels of the environmental variables associated with the relevant probability levels.

Developmental instability

All the species in the genus Fucus exhibit self-symmetry, i.e., symmetry across scale (Corbit & Garbary, 1995). We estimated the developmental instability in F. guiryi individuals by deviations of the self-symmetry of thalli, by using the box-counting procedure (Mandelbrot, 1983; Iannaccone & Khokha, 1996). At Punta Calaburras, a single thallus was collected from the center of a randomly placed 15 × 15 cm quadrat, from each of the 20 different protruding rocks (those reaching the maximum height over the sea surface). At Tarifa, a 5 m length horizontal transect was located parallel to the sea surface and a thallus was collected every 20 cm from the center of a quadrat of 15 × 15 cm. At both study sites, the sampling stopped when the twenty-first thallus lacking grazing marks was collected. This precaution was taken because grazing marks alter the thallus shape and, consequently, the self-symmetry; thus, individuals showing no damage were selected (around 15% of thalli showed herbivore marks, especially by the fish Sarpa salpa). Curiously, grazing marks were exclusively found on thalli collected in Tarifa, possibly due to the fact that the thalli inhabiting Punta Calaburras are located in the upper part of the tidal range of the area. Thalli were placed between two transparent acetate foils, avoiding overlapping the fronds, and they were scanned in TIFF format (300 ppi). The scanned images were superimposed on grids with exponentially increasing box sizes (0.125, 0.25, 0.5, 1, 2 and 4 cm²). The number of boxes in which at least part of the thallus occurred were counted using an image analysis system Visilog 6.3 (Noesis, France). Twenty independent thalli were processed from each location. Because the overall positioning of the boxes can affect the results of a box count (Walsh & Watterson, 1993; Schulze, Hutchings & Simpson, 2008), the counting of the boxes was carried out three times (named “replicates”), repositioning the thalli over the acetate foils with the grids every time. We then regressed the natural log of the number of occupied boxes against the natural log of the size of each box. The absolute value of the slope of the regression line is the fractal dimension (a measure of the space filled by the individual). Developmental instability is the degree to which the individual failed to fit the idealized phenotype, and is measured as the standard error of the estimate (S_Y⋅X, computed as the square root of the residual mean square of the anova regression of the linear fit). The value of S_Y⋅X is an overall indication of the accuracy with which the fitted regression function predicts the dependence of Y on X. Under non-stressful conditions, all points should lie on the regression line.

A two-level nested anova (model: y = overall mean + locations + replicates [locations] + error) was performed to compare the S_Y⋅X values. The factor “locations” correspond to the Tarifa and Punta Calaburras populations, whereas the factor “replicates [locations]” corresponds to the three independent measurements of S_Y⋅X of each thallus from both locations. The homogeneity of variances was previously checked with Bartlett’s test.

Measurement of natural solar radiation and temperature in air and water

The measurements were carried out on 16th July 2011 at Tarifa, and on the following day at Punta Calaburras. Daily changes in PAR (λ = 400–700 nm) were measured using a LI-190R PAR sensor connected to a LI-1400 data logger (LI-COR, Lincoln, NE, USA). The ultraviolet A (λ = 315–400 nm) and ultraviolet B (λ = 280–315 nm) bands were measured using a RM12 device (Dr. Gröbel, Ettlingen, Germany) connected to the respective UVA and UVA sensors. Measurements were made every 30 min, and data were fit to a single sinusoid with the free software PAST ver. 2.17 (Hammer, Harper & Ryan, 2001); the daily doses of each channel were calculated by integrating the area under the sinusoidal curves. Air temperature (±0.1 °C) in the shade was measured with a sensor connected to a LI-1400 data logger. Seawater temperature (±0.1 °C) was measured with the temperature sensor of a Crison2 OXI-92 (Crison, Barcelona, Spain) oxymeter.

In vivo measurements of chlorophyll a fluorescence

A day-long record (from 06.00 to 18.00, Coordinated Universal Time –UTC-) of the photosynthetic performance of F. guiryi was carried out on the same days as the solar radiation measurements. The sampling day was selected to correspond to a spring tide, with a maximum tidal height ca. 1.2 m at Tarifa and ca. 0.4 m at Punta Calaburras (Fig. 2). Because the weather was sunny without clouds and with similar air and seawater temperatures (see ‘Results’), we assumed that the photosynthetic performance measured on the two consecutive days at the different locations should be comparable. Five independent thalli were randomly collected before sunset from the higher eulittoral of Tarifa and Punta Calaburras. For this purpose, we initially randomly collected 20 thalli at each location, as explained in the Developmental instability section above. Then, the thalli were randomly numbered from 1 to 20, and we selected five thalli whose numbers correspond to those that had been generated with a random integer generator accessible at a free-access web site: https://www.random.org/integers/. The thalli were placed in a 25 L white polyvinylchloride tank in natural seawater from the collection site. The tank was placed close to the attachment site of the alga, in an unshaded place. To avoid nutrient depletion and changes in temperature in the tank, seawater was renewed completely every 10 min. The measurements were carried out every 2 h from sunrise to sunset; from 10.00 to 14.00 UTC; an extra set of measurements was carried out on thalli exposed to air from 09.30 UTC.

Fresh mass (FM) on thalli in situ was measured with an ELB120 portable analytical balance (±0.01 g) (Shimadzu, Kyoto, Japan) after thalli were blotted dry with paper towel. The same samples were transported to the laboratory and dried at 60 °C for 48 h to determine their dry mass (DM). The percentage of water content of thalli was determined as: ${\displaystyle \text{\%}\text{Water content}=\left(\frac{\mathrm{FM}-\mathrm{DM}}{\mathrm{FM}}\right)\times 100.}$

Chlorophyll fluorescence was measured using a portable pulse amplitude modulated PAM-2000 fluorimeter (Walz, Effeltrich, Germany) following Schreiber, Schliwa & Bilger (1986). The optimal or potential quantum efficiency (F_v∕F_m) was measured in thalli exposed to darkness for 30 min.

The relative electron transport rate (ETR_rel) was estimated as: ${\displaystyle ET{R}_{\mathrm{rel}}={\Phi }_{\mathrm{PSII}}\times I}$ where I is the incident irradiance of PAR (λ = 400–700 nm) and Φ_PSII is the quantum yield of PSII photochemistry.

The contribution of the location and the time of day, on water content of thalli (in air), and F_v∕F_m, Φ_PSII and ETR_rel (in air and water), was analysed by a two-way, model I anova. Differences, when obtained, were checked by the Student-Newman Keuls (SNK) procedure.The homogeneity of variances was previously checked with the Bartlett’s test. The Pearson’s correlation coefficient was computed for the relationships between hydration of thalli and photosynthetic performance parameters.

Statistical software analysis

The exp and tendency tests in the time series were performed using the free software PAST ver. 2.17 (Hammer, Harper & Ryan, 2001) accessible at http://nhm2.uio.no/norlex/past/download.html. The remaining statistical analyses were carried out using R Core Team (2013).

Analysis of the time series of occurrence

The occurrences in midsummer of thalli of F. guiryi at Punta Calaburras through the years 1990 to 2015 were clustered (M = 0.96; M-expected = 0.36, p < 0.0001) from 1990–1994 and 1999–2004, with sporadic occurrences in 2006 and 2011 (Table 1). A trend in the occurrence of thalli in midsummer throughout of the time series was not detected (U-Laplace test = − 3.8 × 10⁻¹⁵, p = 1).

We found a significant positive relationship between the NAO for the months from April to June (NAO₃) of each year and the probability of the presence of F. guiryi (χ² = 13.530, df = 1, p = 0.0002; Wald’s test = 5.994, df = 1, p = 0.014; Table 2) according to the logit y function: ${\displaystyle y=3.418\times {\mathrm{NAO}}_3+0.239.}$ The 95% confidence limits for the intercept and for the explanatory variable NAO₃ were [−0.779, 1.263] and [0.682, 6.138], respectively; that is to say, a logit y function in which the intercept is deleted could also be used because its contribution to the model was not significant (the confidence interval includes the figure 0). This model was well calibrated (Hosmer and Lemeshow’s test = 4.145, df = 7, p = 0.7661), meaning that the differences between observed and predicted frequencies were not significant. The overall ability of discrimination and the general explanatory power of the model were high (AUC = 0.876 and Nagelkerke R² = 0.541, respectively).

According to the logit y function, the probability p of the occurrence of F. guiryi thalli in midsummer computed as exp^y∕(1 + exp^y), was calculated (Fig. 3). Based on relevant p values, we estimated the correct classification of years in which the NAO₃ index favoured the presence or the absence of F. guiryi thalli. The model clearly identified three of four highly probable years (p > 0.6, corresponding to NAO₃ > 0.048) for the presence of F. guiryi thalli, and simultaneously, all of the clearly improbable years (p > 0.4, corresponding to NAO₃ <  − 0.189) were correctly assigned.

Developmental instability

The standard error of the regression (S_Y⋅X) derived from the box-counting method was used as a proxy for developmental instability in F. guiryi. The comparison of the S_Y⋅X values showed that the replicates [locations] were not significantly different (nested anova; F = 0.0002, df = 4 and 114, p = 1.000) suggesting that the reproducibility of the method was reliable. The S_Y⋅X values ranged from 0.025 to 0.162 (overall mean = 0.094 ± 0.038) in the thalli from Tarifa, and from 0.037 to 0.153 (overall mean = 0.090 ± 0.025) in the algae collected at Punta Calaburras, being significantly higher, at an α level of 0.05, at Tarifa than at Punta Calaburras (nested anova; F = 14.041, df = 1 and 4, p = 0.040).

In situ temperature, solar radiation and photosynthetic performance

At Tarifa, the temperature of the air on 16th July 2011 increased from 18.2 °C in early morning to an overall mean of 27.4 ± 0.3 °C between 12.30 and 14.30 UTC, and then declined throughout the afternoon. The temperature of the seawater did not change significantly throughout the day, with an overall mean value of 19.5 ± 0.1 °C. The air temperature records at Punta Calaburras were 21.1 °C in early morning and a maximum of 28.2 °C at noon; the seawater temperature ranged from 19.1 to 19.8 ° C.

The daily profile of the solar irradiance recorded at Tarifa showed a symmetrical pattern centered at noon, typical for a clear blue sky (Fig. 4). Daily doses of solar radiation were 9228.25, 1109.70 and 13.03 kJ m⁻² for PAR, ultraviolet A and ultraviolet B, respectively (Fig. 4). Solar radiation data recorded at Punta Calaburras the following day were similar (data not shown), with doses differing <±3%.

The F_v∕F_m figures ranged from 0.674 ± 0.035 to 0.732 ± 0.034 during the day in permanently submerged thalli (Fig. 5A). The F_v∕F_m values were similar at both locations but a highly significant effect of time of day was detected (Table 3). The interaction between sampling location and time of day was not significant (Table 3). Under simulated emerged conditions, the values of F_v∕F_m significantly decreased from 10:00 to 14:00 (Fig. 5A), with a greater decrease in thalli from Punta Calaburras than those from Tarifa (Table 3). A significant interaction between locations and time of day was also found (Table 3).

The Φ_PSII figures ranged from 0.307 ± 0.023 to 0.732 ± 0.028 during the day in permanently submerged thalli (Fig. 5B). The Φ_PSII values were similar at both locations (Table 3) but a highly significant effect of time of day was detected (Table 3). The interaction between sampling location and time of day was not significant (Table 3). Under simulated emerged conditions the values of Φ_PSII significantly decreased from 10:00 to 14:00 (Fig. 5B; Table 3), with a greater decrease in thalli from Calaburras than those from Tarifa (Fig. 5B; Table 3). A significant interaction between locations and time of day was also found (Table 3).

The ETR_rel figures ranged from 175.3 ± 11.4 to 671.7 ± 34.0 during the day in permanently submerged thalli (Fig. 5C). The ETR_rel values were similar at both locations (Table 3) but a highly significant effect of time of day was detected (Table 3). The interaction between sampling location and time of day was not significant (Table 3). Under simulated emerged conditions the values of ETR_rel significantly decreased from 10:00 to 14:00 (Fig. 5C), with a greater decrease in thalli from Calaburras than those from Tarifa (Fig. 5C; Table 3). A significant interaction between locations and time of day was also found (Table 3).

The water content in algal fronds decreased when they were exposed to air (Fig. 5D), with a greater decrease in thalli from Calaburras than those from Tarifa (Fig. 5D; Table 3). A significant interaction between location and time of day was also found (Table 3).

We found that the hydration level significantly correlated (p < 0.0001, n = 15) both with F_v∕F_m (r = 0.9407) and Φ_PSII (r = 0.9039).