This study was performed in Gran Canaria Island (Fig. 1), which is located at the centre of the Canarian Archipelago, with ca. 45 km diameter and a maximum elevation of 1950 m above sea level (Carracedo et al. 2002). Sediments and rocky reefs mainly compose the nearshore bottoms, with a high variability in the distribution and complexity of these habitats, which can be colonized by a range of canopy-forming species (Tuya and Haroun 2006, Tuya et al. 2014).

A total of 629 specimens were collected, from January to December 2012 at Telde (east of the island, Fig. 1), through bottom trapping carried out by the local artisanal fleet. Traps were deployed between 18 and 30 m depth, either individually or in strings of 2-3 traps along a fishing rope. The number of days on which traps were placed on the seafloor varied between 3 and 10. For each individual, the total length (TL, mm), total weight (TW, g), and gutted weight (GW, g) were recorded. The sex and stage of sexual maturation (EMS) were recorded by a macroscopic examination of the gonads. Gonads were first removed and weighed to the nearest 0.01 g (GNW). Maturity stages were classified as immature (I), developing (II), spawning (III), regressing (IV) and regenerating (V) (Brown-Peterson et al. 2011). The sagittae otoliths were used to determine fish age. They were extracted from all individuals, cleaned and stored dry in plastic vials.

The right otolith was selected to determine fish age. As otolith weight is considered as an indicator of fish growth rate (e.g. Pawson 1990, Fletcher 1991, Cardinale et al. 2000), the criterion was based on the absence of significant differences in mean weight between left and right otoliths (Student t-test, t=0.0098, p=0.992, n=200).

First, whole otoliths were immersed in a 1:1 glycerin–alcohol solution and observed under a stereomicroscope (NIKON SMZ 1000) using reflected light and a dark background. This method, however, did not provide reliable age estimates, so a random subsample of 200 otoliths was selected to perform the study through sectioning. For that, whole otoliths were embedded in epoxy resin and two or three transverse sections (1 mm thick) were cut through the central region with a slow-speed circular saw (Buehler, ISOMET-TM) to obtain a section that included the otolith core. Otolith sections were then mounted on a glass slide using Crystalbond as a mounting media and polished using decreasing grit abrasive paper (3M Lapping Film). Sections that included the core were examined under a compound microscope with transmitted light (Axioplan, Zeiss; Carl Zeiss Inc., Oberkochen, Germany) connected to a digital camera (ProgRresTM C10 plus; Jenoptik, Jena, Germany). Under transmitted light, the core and opaque bands appear as dark rings, and the wider translucent bands as clear or hyaline rings. The count path of the annuli was from the nucleus towards the tip of the inner face next to the sulcus, where the deposition of seasonal rings appeared clearly defined (Fig. 2a).

To assess the precision of readings, one experienced reader counted opaque bands without knowledge of fish size at least twice. To minimize reading bias, the two readings were separated 2-3 months after randomization in the process of readings. When readings differed, a third reading was taken. An otolith was only considered unreadable, and therefore excluded from the analysis, when the differences between readings did not improve after this procedure. The coefficient of variation (CV=SD⁄mean) was used to measure precision of annuli counts together with a paired t-test to statistically compare differences between readings (Chang 1982, Campana 2001). To assess the yearly pattern of deposition of otolith annuli, the appearance of each otolith margin was recorded as opaque or translucent. The timing of annulus formation was examined by plotting the percentage occurrence of otoliths with a peripheral opaque band as a function of the sampling month (Morales-Nin 1992). January 1 (peak spawning; see results) was considered the birthdate, hence their annuli count was assigned to equal age. Finally, the von Bertalanffy growth equation was used to describe the growth of the species; it was fitted to the observed individual length-at-age data, rather than the frequently used mean length-at-age, to show individual growth variability:

L = L_∞ [1 – е^{–k (t – t0)}]

where L is the TL (cm), L_∞ is the asymptotic length, k is the growth coefficient, t is the age (years) and t₀ is the hypothetical age at which length is zero. To better estimate growth parameters, ages of smallest individuals (undifferentiated) were used for both sexes (García-Mederos et al. 2010). Hotelling’s T2 test was used to compare growth parameters between males and females (Cerrato 1990, Gordo 1996).

A multiple linear regression model was used to evaluate whether age determination was predicted by otolith weight, otolith diameter and fish size (Boehlert 1985, Schwamborn and Ferreira 2002). All variables used in the multiple linear regression were log-transformed to conform the assumptions of linearity, i.e. normality and homogeneity of variances (Zar 1996).

The spawning pattern was assessed from monthly changes in the frequency of the maturity stages and the gonadosomatic index (West 1990):

GSI = 100 (GNW ⁄ GW)

Size at first maturity was based on the examination of males and females in mature phases (phase III, phase IV, or phase V) and immature individuals collected during the spawning period. The TL of all individuals was used to estimate the size at first maturity (L₅₀), defined as the size at which 50% of all fish are at sexually mature phases. Maturity curves were adjusted using the logistic curve (Pope et al. 1975):

P = 100 / (1 + exp (a + bTL))

where P is the percentage of mature individuals as a function of size class (TL), and a and b are specific parameters which can change during the life cycle. A logarithmic transformation was initially applied to calculate the parameters a and b by means of linear regression. An Analysis of Covariance (ANCOVA) was used to compare the curves of maturity between sexes.

The sex ratio of the population (males: females), and the sex ratio by size intervals (10 mm) were determined. Sex ratios were tested statistically for significant deviations from the expected 1:1 ratio via chi-square tests (α=0.05).

Spatial and temporal variation in the abundance of S. lurida was studied from visual censuses carried out on a monthly basis between January and December 2012, at two randomly selected localities (Fig. 1). One locality is in the northwest (NW, Agaete, 28°5’57.58”N, 15°42’33.48”W) and the other in the east of the island (E, Telde, 27°59’21.68”N, 15°22’12.96”W); visual counts were performed between 0.2 and 5.8 m depth. Only individuals >2 cm (TL) were counted to optimize the in situ identification. On each sampling occasion, n=5 replicated 25 m transects were haphazardly surveyed during daylight hours. The abundance of fish within 2 m of either side of each transect (100 m²) was recorded on waterproof paper by the same diver, according to standard procedures for the study region (Boyra et al. 2004, Tuya et al. 2004). Individuals were categorized as juveniles (TL<4 cm), sub-adults (TL 6-8 cm) and adults (TL>10 cm) based on morphological characteristics, fundamentally related to their colouration and body size (Mapstone and Wood 1975). This approach was subsequently endorsed by our data (see results). Concurrently, the diver counted on his way back the number of large (>1 m) and small (<1 m) topographic elements of the rocky substrate (i.e. cracks, crevices, caves, holes per 100 m²), and visually estimated the percentage cover of algae, following standardized procedures via the Linear Point Intercept sampling technique (Ohlhorst et al. 1988). The type of substrate was recorded every metre along the 25 m long transect (i.e. 25 points per transect). Water temperature on the bottom was registered using an underwater thermometer.

Differences in the total abundance of juveniles, sub-adults and adults among months (fixed factor) and localities (random factor) were tested through ANOVAs from square root transformed data; this was necessary to avoid heterogeneous variances. Multiple linear regressions tested whether the number of small and large topographic elements (100 m⁻²), the type of substrate, and the algal cover (per transect) affected abundances of the damselfish over time. To retain variables with good explanatory power, the Akaike information criterion (AIC) routine was used as a selection criterion for each model (the smaller the value the better the model, Anderson and Legendre 1999), and the contribution of each independent variable to each model was described with partial r² values. Collinearity diagnoses among independent (predictive) variables were carried out through Spearman-rank-correlations. All analyses were based on a ‘forward’ selection procedure.

Fish ranged in size from 56 to 157 mm TL, and weighed between 3.9 and 76.6 g TW. A total of 419 males, 113 females and 110 immature specimens (gonads were characterized by small, thin and translucent filaments) were identified. Significant differences (Student t-test, t=11.22, p=0.002) were found between the mean sizes of sexes; males were larger than females (Table 1). The size and weight distributions differed significantly between males and females (Kolmogorov-Smirnov test, length: d=6.00, p<0.0001; weight: d=5.67, p<0.0001).

The overall ratio of males to females was 1:0.26; the hypothesis of uniformity between sexes was therefore rejected (χ² =180.5, p<0.05). Females were found in most size intervals, but males were more abundant in the larger size interval (130-140 mm) (Table 2).

The readings coincided for 171 otoliths (85%) and were dissimilar for the remaining 29 (15%). There was no significant difference in age estimation between readings (CV=2.7%; t=–0.094, p>0.05). The growth rings, opaque and translucent, were usually well visible (Fig. 2). Marginal zone analysis showed a pattern of alternating narrow translucent zones and wide opaque zones, forming one annulus per year. Otoliths with an opaque edge (faster growth) were more abundant (52.9-87.5%) from November to March (spawning season, see below), while otoliths with a translucent edge (slower growth) were more common during the remaining months (Fig. 3).

Fish ranged in age from 0 to 18 years; the age of most fishes was between 8 and 10 years. Females reached higher maximum ages (18 years) than males (14 years). Significant differences in mean sizes between sexes were obtained from V and X age classes (Supplementary material, Table S1). A considerable variability in the length-age relationship indicated considerable differences in individual growth. The growth curve obtained by age-at-length data was well described through a von Bertalanffy growth fit, attaining a determination coefficient (r²) of 0.439 for males and 0.707 for females (Fig. 4, Table 3). Significant differences were found between the von Bertalanffy growth curves of both sexes (Hotelling’s T²-test, = 68.654>T₀² _0.05,3.181 = 8.515).

Partial regression coefficients indicated that radius diameter (r²-partial=0.775, t=13.728, p<0.001) and otolith weight (partial r²=0.5575, t=6.367, p<0.001) were significant predictors of fish age (Fig. 5). By contrast, fish length did not contribute to explaining variation (partial r²=0.338, t=–0.728, p=0.338).

Males in phase III were observed from November to May, with a maximum peak in February (Fig. 6a). The maximum occurrence of spawning females (phase III) also occurred from November to March, including a peak in February (Fig. 6b). The presence of regressing females (phase IV) was observed from February to August. Females at immature, regressing and developing/regenerating phases (V, I and II, respectively) were found throughout the entire year. A slight increase in phase III females was observed in May, which would likely result in spawning activity during June, suggesting the possibility of a secondary breeding season (Fig. 6b). For males, the GSI increased from November to May, including a peak in January (0.258) (Fig. 6c). The GSI of females also increased from November to May, with a peak of maximum activity in February (3.897) (Fig. 6d). Overall, integration of EMS and GSI results with maturity phases throughout the year indicated a spawning season from November to May. The maturity curves were clearly different between males and females (ANCOVA, F=34.811, p<0.001); the size at which 50% of individuals are mature was 103.44 mm TL for males and 84.71 mm TL for females (Fig. 7).

The abundance of juveniles, sub-adults and adults varied inconsistently over time between localities (ANOVA: Locality × Month, p<0.05, Table 4). However, juvenile abundances were significantly higher in April at both localities (9.0±2.6 ind/100 m² at Agaete; 11.0±2.7 ind/100 m² at Telde, Fig. 8a, b). The abundance of sub-adults attained higher mean values in July at Agaete (19.6±2.8 ind/100 m²), while the larger values were observed between November and January (14.2±0.7 ind/100 m², 13.2±2.8 ind/100 m², respectively) at Telde (Fig. 8c,d). Adults peaked in abundance at similar times at both localities; November (22.4±1.8 ind/100 m²) at Agaete and November-December (28.2±2.1 ind/100 m², 23.8±1.7 ind/100 m², respectively) at Telde (Fig. 8e, f).

The percentage of sandy cover in the case of juveniles and adults, and the percentage cover of algae for sub-adults, accounted for the largest contribution to variability in fish abundances at Agaete. However, the number of small topographic elements was the largest contributor to variation in the abundance of sub-adults and adults at Telde (Supplementary material, Table S2). For the majority of specimens, the best models provided by the AIC routine included both the number of small topographic elements with algae and with sandy cover, i.e. patterns of abundance were, in most cases, affected by these two descriptors of the habitat (Table S2). Juvenile abundance was significantly predicted by the presence of large topographic elements at Telde (Table S2), i.e. juveniles were particularly abundant on high relief areas.