Spatial patterns of Anchoveta (Engraulis ringens) eggs and larvae in relation to pCO2 in the Peruvian upwelling system

Large and productive fisheries occur in regions experiencing or projected to experience ocean acidification. Anchoveta (Engraulis ringens) constitute the world's largest single-species fishery and live in one of the ocean's highest pCO2 regions. We investigated the relationship of the distribution and abundance of Anchoveta eggs and larvae to natural gradients in pCO2 in the Peruvian upwelling system. Eggs and larvae, zooplankton, and data on temperature, salinity, chlorophyll a and pCO2 were collected during a cruise off Peru in 2013. pCO2 ranged from 167–1392 µatm and explained variability in egg presence, an index of spawning habitat. Zooplankton abundance explained variability in the abundance of small larvae. Within the main spawning and larva habitats (6–10°S), eggs were found in cool, low-salinity, and both extremely low (less than 200 µatm) and high (more than 900 µatm) pCO2 waters, and larvae were collected in warmer, higher salinity, and moderate (400–600 µatm) pCO2 waters. Our data support the hypothesis that Anchoveta preferentially spawned at high pCO2 and these eggs had lower survival. Enhanced understanding of the influence of pCO2 on Anchoveta spawning and larva mortality, together with pCO2 measurements, may enable predictions of ocean acidification effects on Anchoveta and inform adaptive fisheries management.


Introduction
Ocean acidification is the increase in the partial pressure of CO 2 (pCO 2 ) and decrease in pH and CaCO 3 saturation state caused by the rapid addition of atmospheric CO 2 to the ocean from deforestation and fossil fuel burning [1]. While declining CaCO 3 saturation state may make it more difficult for calcifying organisms to secrete their exoskeletons [1,2], increasing pCO 2 may be particularly challenging for fish [3]. Elevated pCO 2 will reduce the outward partial pressure gradient of CO 2 across the gills and skin of fish, and can lead to respiratory acidosis [3,4]. Many of the reported effects of elevated pCO 2 on fish are assumed to be the downstream consequence of compensatory processes to restore internal pH homeostasis during acidosis.
Several species of fish are susceptible to elevated pCO 2 during the early life-history stages (i.e. eggs and larvae), experiencing changes to physiology, development, growth, behaviour, central neural processing and mortality (see review by [5]). Some of these changes have the potential to reduce fitness and chance of survival. Importantly, high and variable mortality during the larval stage, often driven by adverse environmental conditions, greatly influences recruitment variability [6]. Given the susceptibility of young fish to ocean acidification and their role in shaping fisheries, it is essential to understand the impacts of ocean acidification on this important life stage. Naturally high-pCO 2 areas, such as eastern boundary upwelling systems (EBUS) and CO 2 vents, offer the opportunity to investigate the current relationship between organisms and the CO 2 of their environment and pose questions about future changes with ocean acidification [7][8][9].
Anchovies (Engraulis spp.) are small pelagic fish that occur worldwide in temperate regions of high productivity, particularly in the coastal upwelling areas of EBUS [10,11]. Wind-driven upwelling brings cold, nutrient-rich, high-pCO 2 waters to the surface and creates a spatial and temporal mosaic in pCO 2 [7,12]. Anchovy populations around the globe undergo large fluctuations in biomass in response to environmental changes on interannual to centennial timescales [9,[13][14][15]. Notable examples are the collapse of the Pacific sardine (Sardinops sagax) fishery off California in 1947 and Anchoveta (Engraulis ringens) fishery off Peru in 1972 due to changing ocean conditions and overfishing [11,14]. The low level of nucleotide diversity and shallow coalescence of mitochondrial DNA genealogies of anchovies indicate periodic regional population collapses have occurred in the past in response to changes in oceanographic processes [10]. Rapid evolutionary adaptation is more likely to occur in populations with high levels of existing genetic variation and large population size [7]. Therefore, the strong influence of the environment on the biomass and recruitment combined with the genetic structure suggest that anchovy populations may be especially vulnerable to climate change effects [9].
Anchoveta (Engraulis ringens) inhabit the Humboldt Current System and play an important ecological role as a midtrophiclevel species [16], and support the world's largest single-species fishery [17]. Of the three stocks, the north-central Peru (NCP, 4-15 8S) stock is located within the highly productive and high-pCO 2 Peruvian upwelling system and dominates the landings [14]. The Peruvian upwelling system experiences elevated pCO 2 year-round, with concentrations exceeding those of other EBUSs. Measurements of pCO 2 in the coastal region can reach 1500 matm [18][19][20]. Coastal upwelling systems have a lowered buffering capacity to offset acidification and are at the forefront of observable climate change [18,21]. 'Hotspots' of acidification ( pCO 2 . 1000 matm) are predicted to occur in major fishery zones by mid-century when atmospheric CO 2 is projected to reach 650 matm [21].
Anchoveta respond to environmental fluctuations by altering their habitat use and reproductive strategy. For example, during El Niñ o events, Anchoveta migrate further south and nearer to the coast to seek refuge from warm temperatures [13]. Fecundity and spawning frequency are reduced, and the spawning season extended [13,22]. Unlike adults, the early life-history stages of many fish species are planktonic and largely unable to make behavioural modifications to escape stressful environmental conditions. Furthermore, the peak spawning season for Anchoveta (August-November) occurs during maximum upwelling activity [19,23], resulting in the spawning and development of eggs and larvae at high pCO 2 .
We use the Peruvian upwelling system, with its naturally high pCO 2 and large Anchoveta population, as a natural experiment to investigate the relationship of pelagic fish eggs and larvae to pCO 2 . We posed the following questions: (i) What is the surface water pCO 2 in the spawning habitat? (ii) Are eggs and larvae found in areas of high pCO 2 ? (iii) Can we use differences in spawning and larva habitats to make inferences about the effects of pCO 2 on mortality? To address these questions, we examined the distribution and abundance of Anchoveta eggs and larvae across an inshore-offshore gradient of pCO 2 off Peru during the spawning and upwelling season in 2013.

Material and methods (a) Cruise information
Eggs and larvae of the NCP stock of Anchoveta and oceanographic data were collected between 3 8S and 12 8S during a 30 day cruise in August-September 2013. The cruise was conducted by the Instituto del Mar del Perú (Imarpe) and comprised parallel transects extending from the coast to approximately 90 nm (140 km) offshore (figure 1).

(b) Oceanographic data
Temperature, salinity and pCO 2 were measured continuously and recorded at 1 minute intervals from the vessel's seawater system at 3 m depth using a thermosalinograph (Sea-Bird Electronics Inc., Bellevue, WA, Model SBE-45) and flow-through sensor based on membrane equilibration and non-dispersive infrared spectrometry (KM Contros GmbH, Kiel, Germany, Model HydroC w CO 2 FT), respectively. Three seawater samples were taken for the analysis of total alkalinity and dissolved inorganic carbon (Andrew Dickson, Scripps Institution of Oceanography, La Jolla, CA, USA) to validate flow-through pCO 2 measurements. An 8 day composite of surface chlorophyll a (mg m 23 ) within a 15 km radius of egg and larva stations was obtained from the MODIS Aqua Ocean Color sensor (http://coastwatch.pfeg.noaa.gov). Chlorophyll a is a proxy of primary productivity and, in this study, an indication of the availability of phytoplankton prey for young larvae since Anchoveta larva prey on phytoplankton until 2.5-4.0 mm in standard length (SL) [24,25].

(c) Eggs and larvae
Eggs were collected continuously at a depth of 3 m in 20 minute intervals using the Continuous Underway Fish Egg Sampler (CUFES), which concentrates and filters eggs from a flow of seawater through 330-mm mesh [26]. Volumetric concentration (eggs min 21 and eggs m 23 ) at 3 m is highly correlated with areal abundance (eggs m 22 ) [26,27]. Eggs were counted onboard and preserved in 2% formalin-seawater.
Larvae ranging from 2 to 15 mm were collected on station using a Hensen net of 60 cm diameter and 330 mm mesh towed vertically from a depth of 50 m [28]. Larvae were counted, measured to the nearest 0.1 mm, and preserved in 70% ethanol. Larvae measuring less than 5 mm in length were used for subsequent analyses because they were more abundant than larger larvae and their collection was more contemporaneous with oceanographic measurements. Lengths were not adjusted for shrinkage and counts were not adjusted for variable retention rate as there is evidence for both complete [29,30] and partial (0.63) retention [31] of anchovy larvae smaller than 5 mm caught using 330 mm mesh. Because small larvae (less than 5 mm) can actively swim only approximately 50% of the time and at speeds that are significantly lower than typical current speeds [25,32], their behavioural contribution to horizontal movement is negligible and they are not expected to avoid the net. However, bias in abundance data may be present.
The remaining zooplankton were preserved in 2% formalinseawater buffered with borax. Zooplankton volume was measured using the displacement method following the removal of large rspb.royalsocietypublishing.org Proc. R. Soc. B 284: 20170509 gelatinous organisms [28] and is used here as an index for the abundance of zooplankton that prey on Anchoveta larvae. Maps of the concentration of eggs and larvae with pCO 2 and salinity interpolations to a 0.18 grid were created using MATLAB (The Mathworks, Inc., Natick, MA, USA).

(d) pCO 2 data processing
The HydroC w CO 2 FT automatically performed zero-CO 2 gas measurements every 12 h. Zero-CO 2 gas readings and pre-and post-calibration information were used to apply a drift correction based on absolute sensor runtime [33]. The response time (RT) of the sensor varied due to a variable flow of the ship's seawater system and fouling. To account for the effect of variable RTs, two corrections were performed based on signal recovery times following zero-CO 2 gas measurements [34], a 'slow/fouled' RT of 1200 s and 'fast/unfouled' RT of 300 s. The final RT-corrected series was created from these two datasets. Time periods of variable seawater flow were removed from the final data as their quality was unknown.
A linear regression (R 2 ¼ 0.99) of temperature measurements at 3 m depth obtained from Niskin casts during sampling stations (n ¼ 50) and temperature measurements from the thermosalinograph was used to convert the RT-corrected pCO 2 at the sensor to pCO 2 in situ at 3 m [35]. pCO 2 measurements from the HydroC w CO 2 FT and those estimated from DIC and TA measurements using CO 2 Calc (http://pubs.usgs.gov/of/ 2010/1280/) had an average deviation of 2.2%. We attribute an uncertainty of 1.5% for the majority of measurements, but acknowledge that over short periods and in times of large pCO 2 gradients, the uncertainty could be as high as 10% due to discrepancies between the sensor's actual RT and that assumed during processing.

(e) Spatial generalized linear mixed models
We constructed a set of candidate models (electronic supplementary material, table S1) to evaluate the relative influence of temperature, salinity, pCO 2 , satellite chlorophyll a and zooplankton displacement volume on the presence of eggs and abundance of small larvae (SL ,5 mm). Egg presence was modelled to capture the areal extent of the spawning habitat and because 60% of CUFES samples contained zero eggs [36,37].
We performed logistic regressions with a binomial distribution and logit link to model the effects of temperature, salinity, pCO 2 and chlorophyll a on egg presence. Larva abundance was modelled using generalized linear models with a Poisson distribution and log link. Zooplankton displacement volume was included as an additional predictor variable. Models included either temperature or salinity, and either temperature or pCO 2 due to strong correlations between these variables (Pearson's jrj . 0.6; electronic supplementary material, table S2). Quadratic terms were included in models because the probability of encountering eggs of anchovy can peak across a range of oceanographic conditions [36,38] and dome-shaped relationships may be indicative of an 'optimal environmental window' [39].
Egg and larvae data were standardized by subtracting the mean and dividing by the standard deviation prior to model fitting. Spatial autocorrelation, detected by global Moran's I, was accounted for as a random effect through the use of spatial generalized linear mixed models (SGLMM) using the function corrHLfit in the package 'spaMM' in R v. 3.1.2 (R Core Team 2013) [40].
The relative plausibility of the candidate SGLMMs was determined using Akaike's Information Criterion adjusted for small sample sizes (AICc; electronic supplementary material, table S3) [41]. We calculated the DAICc and scaled the models by their Akaike weight. Parameter estimates for each variable in models with Akaike weights greater than 10% of the model with the lowest AIC were averaged to account for model selection uncertainty using the R package 'AICcmodavg' [41]. Ninety-five per cent confidence intervals were constructed around parameter estimates for each predictor variable in the model. Parameters were interpreted as significant if confidence intervals did not overlap zero [41].
We constructed partial-effects plots to illustrate the effect of individual predictor variables on the probability of egg capture and predicted number of larvae. Partial effects were calculated by allowing the variable of interest to take on measured values while all other predictor variables in the model were fixed at their median value [37]. Predicted probabilities and counts were averaged within 0.5 unit bins to provide a clearer picture of the central relationships.   [19]. In general, pCO 2 was high near to the coast where wind-driven coastal upwelling occurred and decreased offshore. Approximately 74% of the measurements during 2004-2006 exceeded atmospheric pCO 2 (378 matm; http:// www.esrl.noaa.gov/gmd/ccgg/globalview/index.html) by more than 100 matm and 8% had values that were more than twice atmospheric pCO 2 [19]. In comparison, 84% of our data exceeded atmospheric pCO 2 (393 matm) by more than 100 matm and 23% were greater than twice atmospheric pCO 2 (756 matm). However, the majority of high values ( pCO 2 . 1000 matm) in 2004-2006 were observed further south (14-16 8S) than the main spawning area and our study region. Our findings indicate that the spatial extent of high pCO 2 water in the main spawning habitat, as well as the maximal concentration of pCO 2 in this water, was greater in 2013 compared with 7-9 years ago.

(b) Spawning habitat characterization
Eggs were found throughout the sampling region in high abundance. A total of 236 220 eggs were collected in 867 CUFES samples, with a mean of 27 eggs m 23 and maximum of 2000 eggs m 23 (figure 1a). Maximal egg concentration was 1-3 orders of magnitude greater than that of the centralsouth Chile stock of E. ringens (33-428S) [14,42], E. mordax in the California Current [26,36] and E. encrasicolus in the Benguela Current [43], and reflects the large spawning stock biomass off Peru.
The main spawning habitat was located between 7-108S (figure 1a), consistent with other years [23,44 -46]. This area is characterized by a wider continental shelf, increased stability of the physical environment, higher retention rates, and better feeding conditions for larvae, all factors that likely contribute to making it a preferred spawning location [23,32,45]. CUFES samples with the highest concentrations of eggs (more than 1000 eggs m 23 ) were found near to the coast in recently upwelled water with a mean salinity and pCO 2 of 34.90 + 0.01 (mean + s.d.) and 930 + 211 matm, respectively (figures 1a and 2). Spawning also occurred offshore in lower pCO 2 water (figure 1a).
Egg frequency of occurrence was maximal in water of high pCO 2 (1000-1100 matm), cold temperature (15-168C) and relatively low salinity (34.85-34.90) ( figure 3). Chlorophyll a estimated from satellite imagery ranged from 0.2 to 16.7 mg m 23 and eggs occurred most frequently in areas with high chlorophyll a concentrations. These oceanographic conditions are characteristic of the productive cold coastal water mass (CCW) that dominates coastal upwelling [47].
pCO 2 was the only statistically significant variable to predict egg presence (table 1). The relationship of pCO 2 to egg presence was positive and quadratic (table 1; figure 4a). The probability of collecting eggs increased from 0.30 to 0.97 as pCO 2 increased from the mean of 641 matm to 1198 matm (figure 4a), indicative of the fact that spawning was largely associated with upwelled water high in pCO 2 . The high probability of finding eggs in the lowest pCO 2 waters was initially unexpected, but corroborates the offshore spawning that was observed in low pCO 2 water (figure 1a).
To our knowledge, this is the first report of pCO 2 as a variable that significantly characterizes the spawning habitat of an anchovy species. Temperature and salinity, often with chlorophyll a concentration, have been identified as important factors in the characterization of spawning habitat for

(c) Larva habitat characterization
A total of 1157 Anchoveta larvae were collected at 74 stations, of which 683 measured less than 5 mm in SL and had corresponding pCO 2 data for subsequent analyses. Larvae were concentrated between 7 and 108S, and samples had a mean of 10 larvae m 23 and a maximum of 51 larvae m 23 (figure 1b). Compared with eggs, larvae were more abundant in lower pCO 2 (400-500 matm), warmer (17-188C), and more saline (34.90-35.00) water (figure 3). Selection for pCO 2 was not evident, with 48% of larvae found between 400-600 matm, corresponding to the pCO 2 range most sampled (figure 3). Chlorophyll a concentration ranged from 0.3 to 5.8 mg m 23 and zooplankton displacement volume ranged from 0.2 to 7.5 mg m 23 .
Larvae were found within the CCW, but to a larger extent within the mixed coastal-subtropical water mass (MCS), characterized by higher temperatures and salinity [42]. Only 17% of larvae were found at salinity ,34.90, where the majority of eggs were collected. Mean salinity and pCO 2 were 34.98 + 0.06 and 552 + 183 matm, respectively, for the largest samples containing more than 40 larvae m 23 ( figure 2).
Larvae of various lengths (ages) were widespread throughout the larva habitat (figure 5). As an example, 52% of the larvae that were collected between 120 and 140 km were between 2-3 mm, 31% between 3-4 mm and 17% between 4-5 mm. Similarly, large larvae were also found near shore, with 23% of the larvae between 3-4 mm at stations 20-40 km from the coastline ( figure 5).   figure 4b). Larva abundance increased as zooplankton volume reached the mean value of 3.0 ml m 23 and then decreased exponentially as volume increased (figure 4b). Because zooplankton were collected using a 330 mm mesh, zooplankton volume provides an estimate of the abundance of zooplankton predators of larvae. Anchoveta larvae transition from a diet of phytoplankton to zooplankton around 4.3 mm, but the main prey, copepod nauplii (40-90 mm in diameter) [25], are too small to have been captured in zooplankton collections. Therefore, the decrease in larva abundance at high zooplankton volume is consistent with higher predation rates by zooplankton predators on larvae. Similarly, the occurrence of Anchoveta larvae off Chile was negatively correlated with predatory zooplankton [27] and positively correlated with small microplankton [42].

(d) Factors influencing habitat characterizations
The spawning and larva habitats of the NCP stock of Anchoveta were similar in latitudinal range, but distinct in oceanographic conditions. Samples with the highest concentrations of eggs and larvae were collected in waters of significantly different salinity and pCO 2 (t-test, p-value , 0.001), corresponding to different water masses. Egg frequency of occurrence, probability of collection and concentration were maximal at high pCO 2 and low salinity in CCW. Contrarily, larva frequency of occurrence and concentration were highest at low pCO 2 and high salinity in MCS. There are several factors to consider that could explain the distinct habitat characterizations for eggs and larvae.
Although advection has the potential to transport individuals offshore, natal locations [50], oceanographic modelling and our data indicate that offshore advection is minimal in our study area. Regional hydrodynamic and individual-based models, which account for egg buoyancy and larval vertical migration, concluded that retention is upward of 70% in the main spawning and larva habitats (7-10 8S) [23,51]. Over an 8 day period drifting at 4 cm s 21 , eggs and larvae are transported less than 28 km in simulations [23]. In addition, the relative strength of alongshore currents in the winter would support equatorward rather than offshore transport [51]. Lastly, direct observation of our own data also does not support significant offshore transport during development. Both eggs and larvae of various lengths were found throughout the sampling region, from near-shore stations to those further offshore (figures 1 and 5). Indeed, it is thought that spawning in locations that favour retention is a fundamental life-history characteristic of Anchoveta in the Humboldt Current [51,52].
In general, pCO 2 is high along the coast where waters are upwelled and decreases offshore [18,19,53]. As upwelled water ages, pCO 2 declines primarily due to the uptake of inorganic carbon by phytoplankton for primary production [19,53]. Drifters in the Peruvian and Mauritanian upwelling systems measured an average decline in pCO 2 of less than 200 matm over an approximately one-week period of offshore transport [19,53]. We collected and analysed data of eggs, which hatch in approximately 2 days at 178C [44], and larvae of SL ,5 mm, corresponding to first-feeding larvae approximately one week in age [25]. Despite variability in pCO 2 on the event time scale from changes in wind, upwelling and primary production, we anticipate a decline in pCO 2 of 200 matm during the 5-7 day development of eggs into young larvae.
We hypothesize that the differences we observed in the distributions of eggs and larvae result from low egg and larva survival at high pCO 2 . Our results suggest that eggs spawned at high pCO 2 (more than 900 matm) suffered higher mortality than those spawned at lower pCO 2 , contributing to the absence of large concentrations of larvae at pCO 2 . 700 matm. Despite the majority of spawning having occurred at high pCO 2 , a larger fraction of eggs spawned further offshore in waters of higher salinity and low-to-intermediate pCO 2 appear to have survived to the larval stage. Although this study was not specifically designed to measure mortality, observations of distinct egg and larva habitats, coupled with the significance of pCO 2 as the variable characterizing spawning habitat in SGLMMs, are consistent with the hypothesis that the concentration of larvae relative to eggs in different water masses is representative of regional differences in mortality rate associated with pCO 2 .
We acknowledge that there are variables that we did not measure that may have affected the distribution and abundance of eggs and larvae, and whose mechanistic relationship may have been captured by the environmental variables we measured. Oxygen is strongly correlated with pCO 2 [54] and an important variable affecting fish distributions in the Humboldt Current [55]. However, while the correlation of O 2 and pCO 2 is strong at depth, O 2 in the surface ocean is near saturation despite elevated pCO 2 . Of the 47 samples of surface (0 m depth) O 2 taken during the cruise, 15 of these were taken in seawater with pCO 2 . 800 matm and O 2 ranged from 2.15 to 5.18 ml l 21 . Additionally, variables related to water column structure and horizontal flow have increased the power of models to predict anchovy spawning habitat in the California Current [36,37]. Lastly, the inshore-offshore gradient in upwelling that drives the spatial pattern in pCO 2 , O 2 and horizontal flow, also influences nutrient concentrations and the community composition and size structure of phytoplankton and zooplankton. All of these factors are likely to affect the distribution, abundance and survival of eggs and young larvae. Therefore, measurements of these variables in addition to pCO 2 are needed during future cruises to rigorously test our hypothesis that pCO 2 is an important driver of mortality.  Using systems such as the Peruvian upwelling that are naturally high in pCO 2 as a window into a future acidified ocean can add value to laboratory experiments. By exploring the spatial patterns of Anchoveta eggs and larvae to pCO 2 in their natural habitat, the entirety of the whole ecosystem, including predator -prey dynamics and other important environmental variables, is taken into consideration. While controlled laboratory experiments may not show a significant effect of elevated pCO 2 on the survival of eggs of Atlantic herring (Clupea harengus), mortality in nature might increase due to a decrease in growth rate that would prolong exposure of vulnerable larvae to predators [56]. Similarly, the lack of effect of elevated pCO 2 on Baltic cod (Gadus morhua) egg and larva survival in laboratory experiments could be attributed to adaptation since spawning occurs at high pCO 2 [57]. However, the capacity for a species to adapt may be influenced by the source (i.e. upwelling, eutrophication) and history of high pCO 2 in their natural habitat. Initially, our results appear to contradict the fact that anchovy across EBUSs thrive during La Niñ a years, characterized by strong upwelling and presumably high-pCO 2 waters. Upon closer examination, our findings do not conflict with this observation and highlight the important role of spatial and temporal scale. For example, our data do not indicate that the Anchoveta population will be reduced during high-pCO 2 years (La Niñ a events). Rather, our data show that the survivors into the larva stage may originate from eggs spawned further offshore in lower pCO 2 waters during years of strong upwelling with high pCO 2 inshore. In fact, there is precedent for differences in survival with respect to spatial scale, with survivors to six months of age originating from eggs spawned less than 75 km and also more than 150 km offshore in the years 1999, 2003, 2005 and 2007 [58].
As the extent and concentration of high-pCO 2 water increases, ocean acidification has the potential to influence variability in mortality on seasonal and inter-annual time scales. The peak spawning season for Anchoveta is from August-November, when upwelling and abundance of high-pCO 2 water are at a maximum [19,23]. During 1953-1981, recruitment into the fishery at six months was on average larger for the smaller spawning season of February-March than the peak spawning season from August -November [59]. While it was hypothesized that differences in recruitment success could be related to the effects of turbulence on larval food concentration [59], the influence of high-pCO 2 water on mortality could become a contributing factor in the near future. Changes in mortality on spatial and temporal scales could have widespread effects on the population and fishery.
Anchoveta, Alaska pollock (Theragra chalcogramma), Skipjack tuna (Katsuwonus pelamis), sardines (Sardinops spp.) and Atlantic herring (Clupea harengus) collectively comprise 19% of the world's global marine fish catch [17]. The major fisheries of the world are largely found in naturally high-pCO 2 regions that are expected to experience ocean acidification earlier and more strongly than other areas of the world's ocean [21]. Thus, understanding the effects of ocean acidification on marine fish is important for the management and sustainability of fisheries in the future. Enhanced understanding requires the long-term monitoring of pCO 2 concurrent with the collection of eggs and larvae. Furthermore, time series of observations spanning multiple decades are required in order to differentiate the impacts of climate variability from those of climate change [9].
The management strategies of Anchoveta and Pacific sardine (Sardinops sagax) incorporate climate variability to some degree, and may serve as a model for the sustainable management of fisheries in the face of climate change. For example, Imarpe's EUREKA Program enables managers to repurpose the fishing fleet for a rapid stock assessment of Anchoveta during the onset of an El Niñ o event [60]. The harvest control rule for Pacific sardine (Sardinops sagax) depends on temperature in the southern California Current System [61]. These progressive management strategies are testimony to the importance of, and ability to, consider the impacts of climate on the sustainability of commercial fisheries. In the future, EUREKA cruises and harvest control rules may be triggered by extreme upwelling or La Niñ a events that are characterized by high pCO 2 waters. It is timely to consider how climate change, particularly ocean acidification, may be incorporated into fishery management strategies.
Data accessibility. The datasets supporting this article have been uploaded as part of the electronic supplementary material.