Reproductive success of egg-burying species such as sea turtles depends upon the stability of the nest environment [1], [2]. For example, it is suggested that crocodilians and turtles survived the mass extinction at the end of the Cretaceous Period in part because they laid eggs in substrates that provided relatively stable temperature and humidity throughout incubation [3], [4]. Variations in temperature and humidity occur in the underground nest environment and are greater in freshwater turtle and lizard nests that are shallower than sea turtle nests [5]. Water exchange and incubation temperature influence hatching success of eggs and fitness of hatchlings. High temperature and dry substrates are unfavorable for egg development and hatchling quality [2], [5]–[8] and can cause the loss of an entire reproductive season [9]. Thus, changes in local climate can have important effects on reproductive success of reptiles.

El Niño Southern Oscillation (ENSO) has profound effects on many ecological processes [10], [11]. It alternates from “El Niño” periods when sea surface temperature (SST) in the eastern tropical Pacific is warmer than normal, to “La Niña” when SST is cooler than normal, both of variable duration and strength, with ENSO neutral conditions in between. ENSO affects local weather in diverse ways. Warm El Niño events are associated with increased precipitation in the Galápagos Islands [12], [13] and floods in South America, whereas they are associated with drought in Australia and Indonesia [14].

Leatherback turtles (Dermochelys coriacea) nest on tropical and subtropical beaches around the world, but are critically in danger of extinction [15] due to anthropogenic impacts such as egg poaching and bycatch in industrial and artisanal fisheries [16], [17], [18]. Populations of leatherback turtles have been greatly reduced in the eastern Pacific and Playa Grande, in northwestern Costa Rica remains one of the last major nesting sites in the region [19], [20]. Female leatherbacks in the eastern Pacific nest on average 7 times in a season and return to the nesting beach every 3.7 years [20], [21]. This remigration interval can be prolonged during El Niño conditions or reduced during La Niña conditions [22], [23]. Eggs are buried ∼80 cm deep in the sand, incubate for ∼60 days [24] and, after hatching, hatchlings emerge synchronously from the nest [25]. Temperature of the sand affects the ability of hatchlings to emerge from the nest [26]. Furthermore, leatherback turtles, as well as other reptiles, have temperature-dependent sex determination (TSD) and El Niño causes a female-biased sex ratio [24], [27]. Climate change may threaten survival of leatherback populations even if other factors driving population declines are removed.

Monthly precipitation in North Pacific Costa Rica is influenced by ENSO. Rainy seasons during El Niño events occur but levels of rain are low, causing droughts during the following dry season [28]. The rainy season extends from May to November with September and October being the rainiest months, followed by the dry season from December to April (Fig. 1). The leatherback nesting season starts in October and ends at the end of February. Because the incubation period lasts two months, many clutches are still incubating during the peak of the dry season when sand temperature is maximal, which reduces hatching success and emergence of hatchlings from the nest [7], [9], [26].

The objective of this study was to test the effects of climatic conditions on the incubation of leatherback eggs and emergence of hatchlings from the nest. We found that hatching success of eggs and emergence rate of hatchlings from the nest were strongly affected by the prevailing climatic conditions. Our projections showed that hatchling output will decline throughout the 21^st century under projections of climate change.

We conducted the study at Playa Grande, Parque Nacional Marino Las Baulas (10°20 N, 85°51 W), located on the North Pacific coast of Costa Rica. We included six nesting seasons in the analyses, from 2004–2005 to 2009–2010. Each season included clutches laid from October of each year to January of the next. Clutches laid in February were removed from the analyses because of low sample sizes. We marked nests during egg oviposition and excavated them two days after hatchlings emerged. In some instances we could not mark them because we missed the turtle, we found her when she was already covering the nest, or we lost the nest because of inaccurate triangulations. We excavated and included a total of 814 nests in the analyses (31% of the total estimated number of clutches laid between 2004–2005 and 2009–2010) from 283 females (70% of the females) (Table 1).

We estimated hatching success as the percentage of eggs within a clutch that completed development, and emergence rate as the percentage of hatchlings that successfully emerged from the nest within two nights of the initial emergence event. We used the formula H = S/(S+U) to estimate hatching success (H), where S was number of empty eggshells (more than 50% of the shell remained [29]) and U was the number of unhatched eggs as previously described [26]. We used the formula R = (S−(L+D))/S to estimate emergence rate (R), where L and D were live and dead hatchlings found in the nest respectively at excavation time. We estimated mean monthly hatching success and mean monthly emergence rate by month eggs were laid (October - January).

We obtained local climate data (monthly precipitation and air temperature) from a weather station located at the Daniel Oduber Quirós International Airport in Liberia, Costa Rica (approximately 50 km from study site). Through a regression analysis, we determined the effects of ambient temperature during the two incubating months and precipitation accumulated over different periods of time (during the two months of incubation, one, two and three months before eggs were laid and total precipitation during rainy season), on both hatching success and emergence rate. We then used the regression equations from the analyses of precipitation and ambient temperature that gave the strongest signal on monthly hatching success and emergence rate to produce a hindcast of mean annual estimates from 1976 to 2010, and average annual values considering the temporal distribution of nests (on average 13%, 27%, 35% and 25% of clutches are laid in October, November, December and January respectively). Finally, we compared the hindcast of hatching success and emergence rate to the September - October Multivariate ENSO Index (MEI) for the same time period. This index serves as an indicator of the strength of the El Niño and La Niña events. The MEI is calculated bimonthly and based on six observed variables over the tropical Pacific: sea level pressure, surface zonal and meridional wind components, sea surface temperature, surface air temperature, and cloudiness [30]. The MEI data were available at http://www.esrl.noaa.gov/psd/people/klaus.wolter/MEI/table.html.

Mean (±SD) annual hatching success and mean (±SD) annual emergence rate of leatherback clutches at Playa Grande for 2004–2005 to 2009–2010 were 0.47±0.25 and 0.82±0.22 respectively.

Climate variability influenced hatching success and emergence rate of leatherback clutches. Low precipitation and high temperatures were detrimental to both eggs during development and hatchlings during emergence (Fig. 2). Mean monthly hatching success was influenced by precipitation accumulated in the two months before eggs were laid, precipitation in October (immediately prior to the onset of the dry season), and average ambient temperature during the two months of incubation, The relationship was defined by a stepwise multiple regression equation:

where x₁ = precipitation accumulated in the two months before eggs were laid, x₂ = precipitation in October and x₃ = mean ambient temperature during the two months of incubation (R² = 0.81, n = 24, P<0.001, Fig. 2A). The standardized partial regression coefficients were 0.46, 0.45 and - 0.24 respectively. Therefore, precipitation in the two previous months, precipitation in October and ambient temperature during incubation explained 40, 39 and 21% respectively of the hatching success variation accounted for by the model.

Monthly emergence rate was best explained by the average ambient temperature during the two months of incubation and the total precipitation in September and October. The relationship was defined by a stepwise multiple regression equation:x₁ = mean ambient temperature during the incubation months and x₂ = precipitation accumulated in September – October (R² = 0.65, n = 24, P<0.001, Fig. 2B). The standardized partial regression coefficients were - 0.62 and 0.44 respectively. Thus, air temperature during incubation and total precipitation in September and October explained 58 and 42% respectively of the variation in emergence rate explained by the model. Therefore, precipitation and ambient temperatures, characterized by a strong interannual variability, affected (1) eggs during development and (2) hatchlings during emergence through the ∼80 cm sand column.

We found that MEI was significantly correlated with hatching success (quadratic equation: y = 0.45–0.09 x+0.01 x², R² = 0.52, n = 34, P<0.001) and emergence rate (quadratic equation: y = 0.83–0.06 x −0.01 x², R² = 0.52, n = 34, P<0.001), with positive values of MEI (El Niño) corresponding to years of low hatching success and emergence rates and negative values of MEI (La Niña) corresponding to years of high hatching success and emergence rates (Fig. 3A,B). Our regression model estimated the lowest hatching success and emergence rate for 1987, a year of strong El Niño conditions, while estimating the highest hatching success for 1998 and 2007, and highest emergence rate for 1999, all during La Niña events (Fig. 3C,D).

Local climate had a strong effect on leatherback eggs and on hatchlings that must dig through 80 cm of sand. Dry and warm conditions at Playa Grande were harmful to both eggs and hatchlings. High temperatures during incubation also create lethal conditions for olive ridley turtle (Lepidochelys olivacea) eggs in Pacific Costa Rica during the dry season [9] and decrease emergence rate of green turtles (Chelonia mydas) in Caribbean Costa Rica [35].

Precipitation creates suitable conditions of humidity inside the nest and cool temperatures favor development and prevent sand desiccation in this region. However, higher levels of precipitation can result in increased mortality of eggs or hatchlings as seen at other locations [36]. The strong interannual variability of rainfall at Playa Grande results in “boom-and-bust” cycles of recruitment in the population. Therefore, cool and pluvial La Niña years are extremely important for periods of enhanced leatherback neonate recruitment. Additionally, because leatherbacks exhibit temperature-dependent sex determination and air temperatures during La Niña events are cooler, these events are critical for the production of male hatchlings [37]. Within a season, during both El Niño and La Niña events, wet and cool conditions at the beginning of the nesting season result in a higher production of hatchlings than at the end of the season when warm and dry conditions persist [26]. Leatherbacks at Playa Grande typically nest in the open beach away from the vegetation and do not select for different nest sites along the sea to vegetation axis [38] and there is no shade available along the beach. Sea turtles in this type of tropical breeding site will have a lower fitness if they nest during the months when high temperatures would be detrimental for eggs and hatchlings. However, some nesting still occurs during the off season. Likewise during El Niño years, turtles still nest but in lower than average numbers. Historically, it is possible that some turtles benefited from low nest density found on the beach at these times, before declines started.

Hatching success and emergence rate, however, are influenced by variations in nest microclimate that are not fully captured by the MEI or the coarse monthly precipitation and temperature averages examined herein. For example, high temperatures inhibit emergence of sea turtle hatchlings [7] and high temperatures in the period just before emergence are associated with lower emergence rate [26], [35]. An example of where such dependence meteorological variations not captured by the metrics considered herein may have played a role is the 1997 El Niño. Despite being the strongest ENSO event recorded, the 1997 El Niño resulted in low annual hatching success and emergence rates but not lower than in 1987. This difference may be explained by the shorter time-scale meteorological variations or higher levels of rain accumulated in September and October of 1997 (271 and 182 mm) compared to 1987 (213 mm and 167 mm). Likewise, the 1988 La Niña was a very strong event but resulted in lower hatching success than in 1998 and 2007, when higher levels of rain were registered. Further exploration of the impact of meteorological variations over finer temporal and spatial scales than those captured by the metrics considered, may explain more variance in hatching and emergence success. However, the performance of the coarse-scale meteorological forcing considered herein as predictors (e.g., R² = 0.81 and R² = 0.65 for hatching and emergence respectively) suggests that they provide useful, robust indicators of climate effects. The utility of monthly precipitation and temperature metrics as predictors of hatching and emergence rates also facilitates integration with climate projection, which are archived as monthly outputs.