Edited by: Mary C. Fabrizio, William & Mary’s Virginia Institute of Marine Science, United States
Reviewed by: Bruno Díaz López, Bottlenose Dolphin Research Institute (BDRI), Spain; Leigh Gabriela Torres, Oregon State University, United States
This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
Marine mammals have been proposed as ecosystem sentinels due to their conspicuous nature, wide ranging distribution, and capacity to respond to changes in ecosystem structure and functioning. In southern European Atlantic waters, their response to climate variability has been little explored, partly because of the inherent difficulty of investigating higher trophic levels and long lifespan animals. Here, we analyzed spatio-temporal patterns from 1994 to 2018 of one of the most abundant cetaceans in the area, the common dolphin (
The global mean surface temperature has increased by approximately 1∘C from pre-industrial levels (
Marine mammals, as wide ranging top predators, amplify trophic information across multiple spatiotemporal scales and can therefore act as sentinels of ecosystems’ responses to climate variability and change (
Combining data from multiple sampling programs can help overcome this problem (
Until now, this estimator has been mainly applied to commercially important fish stocks (
Spatial distribution of common dolphin sightings (displayed in segments of up to 10 km) over the BoB for the 1994–2018 period. Circle sizes are proportional to group size, while solid gray lines indicate the isobaths. Sightings in yellow represent the ferry data used to check model fit.
Such productivity and diversity, however, might be altered by climate change in the near future, as rising temperatures (0.26∘C per decade;
Advancement of both MSFD criteria in this region is therefore necessary, especially when it is known that projections of climate change impacts on cetaceans at large spatial scales (e.g., global;
The hypothesis that climate change may affect top predators through climate influences on their ectothermic prey has been often suggested (
Given the multiple drivers potentially influencing cetacean spatio-temporal patterns, understanding the role of each of them is key for a better anticipating of future responses. For that reason, in this study we used a 25-year-long temporal series (1994–2018) to test the effect of prey biomasses, oceanographic conditions and climate indices on the abundance and distribution of the common dolphin (
Cetacean data analyzed in this study, despite focusing on the BoB, belong to a large compilation made by
In order to match with the spatial resolution of the environmental data that we examined in later steps (see “
Observations of common dolphin were analyzed by means of a spatio-temporal delta-generalized linear mixed model (delta-GLMM), referred to here as a VAST model (
Another feature of the VAST model is that it decomposes spatio-temporal patterns in available point-count data into multiple additive components:
A temporal main effect (“intercepts”) representing changes in median abundance over time;
A spatial main effect (“spatial component”) representing the average spatial distribution during the modeled interval;
An interaction of space and time (“spatio-temporal component”) representing variation in distribution among years;
Density covariates, representing the impact of environmental conditions on expected density;
Catchability (a.k.a. detectability) covariates, representing the impact of environmental and/or sampling conditions on expected sampling data, but which do not reflect variation in population density and hence are “partialled out” prior to predicting densities.
Each of these components can be included in each of two linear predictors, and these two linear predictors are then transformed
In our case, we treated year as a fixed effect (default VAST setting), such that there is no shrinkage in overall abundance across years. We modeled spatial and spatio-temporal variation as random effects to help account for multidimensional factors that are not included directly in the model but that can affect the density and distribution of the modeled species (
Detectability covariates were not considered here, because Beaufort sea-state and platform height were included in
The annual abundances of common dolphin predicted for the study area were then analyzed by means of a linear regression to identify significant temporal trends and compared by means of a correlation test with an observed abundance index to check model fit. The observed abundance index was based on the encounter rate (individuals/km) of common dolphin estimated from monthly at-sea observations taken by a team of experienced observers in a constant effort-based systematic sampling scheme, i.e., the Pride of Bilbao ferry (
An additional analysis with predicted abundances was also conducted to identify areas in which significant spatio-temporal changes occurred over the study period. For that, predicted abundances per grid cell were analyzed as a function of year by means of a linear regression. The slope and the
Shifts in distribution were summarized by calculating the centroid of the distribution for a given year (termed center of gravity, CoG) after having predicted the density associated with every knot and year in the previous step. By means of the SDF estimator implemented in the VAST model, the CoG was calculated for the BoB population domain and standardized by the total abundance predicted for the study area, so that our analysis focused on changes in distribution after controlling for changes in total abundance (
To understand spatio-temporal patterns, three main groups of drivers were analyzed (
Summary of the local oceanographic, regional climatic and regional prey variables used in this study accompanied by a little description and the source from which they were obtained.
Temperature | °C | Mean annual temperature between 0 and 100 m depth | The Iberian Biscay Irish Ocean Reanalysis Model | |
Chlorophyll |
Mg/m3 | Mean annual chlorophyll between 0 and 100 m depth | ||
NAO | − | Both NAO and EA are estimated from the difference of atmospheric pressure at sea level between the Icelandic Low and Azores High, but the anomaly centers of the EA pattern are displaced southeastward to the approximate nodal lines of the NAO pattern | NOAA (National Oceanic and Atmospheric Administration) | |
EA | − | |||
AMO | − | Average anomalies of sea surface temperatures | ||
Anchovy | Tons | Mean spawning stock biomass in subarea 8 (Bay of Biscay) | ICES (International Council for The Exploration of Seas): stock assessment models | |
Sardine | Tons | Mean spawning stock biomass in division 8.c and 9.a (Cantabrian Sea and Atlantic Iberian waters) | ||
Mackerel | Tons | Mean spawning stock in subareas 1–8 and 14, and in Division 9.a (the Northeast Atlantic and adjacent waters) | ||
Horse mackerel | Tons | Mean spawning stock biomass in Subarea 8 and divisions 2.a, 4.a, 5.b, 6.a, 7.a–c., and 7.e–k (the Northeast Atlantic) | ||
Blue whiting | Tons | Mean spawning stock biomass in subareas 1–9, 12, and 14 (Northeast Atlantic and adjacent waters) |
Local oceanographic conditions integrated at 100 m depth, specifically temperature and chlorophyll
Regional climate indices, specifically North Atlantic Oscillation (NAO), East Atlantic Pattern (EA), and Atlantic Multidecadal Oscillation (AMO) climate indices (details in
Regional biomasses of potential prey species, based on the assumption that climate change will affect cetaceans distribution through changes in their prey (
Temperature and Chl-a values were sourced from the Iberian Biscay Irish Ocean Reanalysis Model available at the Marine Environmental Monitoring Systems,
For modeling purposes, local temperature and Chl-a variables were included as quadratic forms in the model to allow for non-linear responses (
As a preliminary analysis, potential drivers were correlated with the abundance and CoG of common dolphin obtained in the previous baseline spatio-temporal model. Then, covariates-based modeling was performed in two different ways to identify the most parsimonious drivers and to uncover the relative contribution of covariates:
Univariate spatio-temporal models were fitted for each variable using the same configuration as in the baseline spatio-temporal model. Univariate models were then compared with the baseline model by means of the AIC (
Univariate models were fitted for each variable after setting the spatio-temporal variation (i.e., spatio-temporal random effects) to 0. This was done to remove the contribution of random effects and isolate the effect of the covariates since in VAST, random fields can also account for changes in distribution over time by capturing the residual spatial patterns that cannot be attributed to the fixed effect (
A total of 1728 sightings of common dolphin collected in 21 different surveys were analyzed (
Sampling effort (number of segments of up to 10 km) as a function of year and longitude
The common dolphin abundance estimated by the baseline spatio-temporal model showed a significant increase (
Abundance of common dolphin in the BoB predicted by the baseline spatio-temporal model with standard deviation (shaded area), the linear trend, and its significance.
Spatio-temporal changes in the abundance of common dolphin (predicted by the baseline model) illustrated by means of the change rate (the slope of the linear regression). Hatched areas indicate those areas where change rate is not significant (
The CoG also showed a high interannual variability, but no significant trend was found over time in either of the two axes (
The variation in the center of gravity (CoG) of common dolphin expressed in eastings
Neither the annual temperature nor the Chl-a concentration integrated at 100 m depth revealed a significant (
The correlation between the potential drivers and the CoG (easting and northings) of common dolphin only showed weak relationships. In contrast, predicted abundance revealed several strong relationships (
Pearson correlation among the common dolphin’s predicted abundance, CoG and potential drivers. Circle sizes are proportional to the correlation coefficient, which is indicated inside the circles. Non-significant correlations (
For covariates-based models, the AIC score showed that the most substantial decrease was for the NAO index and regional prey species biomasses (especially anchovy and sardine). Local Chl-a concentration, as well as horse mackerel and mackerel, only contributed slightly, while remaining drivers (temperature, AMO, EA, blue whiting and prey species biomass index) were not relevant in terms of AIC (
Model terms.
No covariates | 27814.85 | 0 | |
Temperature | 27820.78 | 5.93 | |
Chlorophyll | 27811.99 | ||
NAO | 27806.3 | ||
EA | 27816.38 | 1.53 | |
AMO | 27817.57 | 2.72 | |
Anchovy | 27807.76 | ||
Sardine | 27809.77 | ||
Mackerel | 27812.81 | ||
Horse mackerel | 27812.63 | ||
Blue whiting | 27816.69 | 1.84 | |
Biomass index | 27814.12 | –0.73 |
Similarly, covariate-only models (with no random effects) showed that the NAO index and prey species biomasses were able to explain the increase in region-wide abundance of common dolphin (
Abundance estimates predicted by the baseline spatio-temporal model (black line) and by the covariates-based model (with no random effects, colored line) so that the contribution made by each variable
In the case of CoG, only Chl-a and temperature contributed to explain the observed variability but, even then, only in a very small proportion (
Center of gravity estimates predicted by the baseline spatio-temporal model (black line) and by the temperature and chlorophyll-based models (with no random effects, colored line), expressed in easting
The evaluation of the spatio-temporal patterns of common dolphin in the BoB agrees with the MSFD aiming to assess the abundance and distribution of species in European waters. Surveys providing information on species distribution and abundance in this region, however, have shown significant shifts in the spatial distribution of observations, which make necessary the application of methods such as VAST to account for uneven sampling effort.
The modeling of common dolphin sightings revealed a significant increase in abundance, which is in agreement with previous studies conducted in the BoB (
In addition, the predicted abundance estimates were found to be quite coherent with those obtained in previous surveys conducted in summer 2012 in the BoB (
The increasing trend in abundance found in this study for the BoB, however, does not necessarily imply an overall population increase at the Northeast Atlantic level (i.e., species whole distribution range), and instead, could be due to the arrival of individuals from unsampled areas. That is why the results found in this study should be treated with caution and never be used to downplay the effects of incidental capture on common dolphin, especially when recent estimates suggest that the bycatch in the BoB is unsustainable for the population as a whole (
Local environmental variables, such as temperature and Chl-a used in this study, are often unable to capture complex associations between environment and ecological process due to time lags in species responses coupled with the non-linear intrinsic nature of population dynamics (
This can be particularly true for Chl-a and cetaceans species that feed on zooplanktivorous fishes, since the abundance of the latter has been related to a period of zooplankton grazing and a phytoplankton decay (
In this study, however, predictors were introduced at an annual scale to match the available temporal scales of both prey and climatic indices, which prevented its incorporation in a lagged phase and likely led to the low contribution of Chl-a in explaining the spatio-temporal patterns of common dolphin. Similarly, the lack of importance shown by temperature could be also a consequence of this annual resolution or could instead suggest that, within the core of the species range, temperature is not such an important variable to explain its abundance and distribution.
On the contrary, regional indices of climate, spanning several months and considering wider areas of influence, are less disturbed by local variability and very often outperform locally estimated environmental variables (
The results found in this study are a good example of this, as the NAO climate index was found to be the best predictor explaining the abundance of common dolphin according to AIC scores. Specifically, results showed a positive relationship between both, meaning that common dolphin abundance is enhanced during positive phases of NAO, which are characterized by colder and drier conditions over Mediterranean regions, central and southern Europe (e.g., BoB), and warmer and wetter conditions in northern Europe (
Although the NAO index and similar climate indices have been previously related to the abundance of wide ranging predators in the BoB (
Common dolphins are assumed to be opportunistic predators that feed on a wide variety of species, although a preference for energy-rich species, such as the anchovy, sardine, mackerel and horse mackerel investigated in this study, has been suggested (
However, not all potential prey species were included and differences in the distribution of stocks may have also affected the results. In fact, only anchovy’s biomass had been estimated exclusively for the BoB. Remaining species biomasses were either estimated using adjacent areas (i.e., Iberian sardine) or distribution areas that extended considerably the observations range of common dolphin (i.e., blue whiting, mackerel and in a lesser extent horse mackerel), which could have contributed, for example, to the higher prominence of anchovy detected in this study.
The common dolphin is considered a warm-temperate species, and accordingly, its range is expected to expand in response to increasing water temperature (
The prey variables considered in this study, however, could not explain much of the observed spatio-temporal variability of the CoG as a result of being introduced as a biomass index that changed across time but not across space, and hence, could not confirm or reject the hypothesized prey-driven distribution. Whether top predator abundance and distribution is driven by the environment or prey is a much debated question in ecology (
Climate change is believed to affect marine mammals through changes in their physical environment but also in their prey. However, many studies aimed at understanding climate impacts often employ environmental characteristics as proxies for prey distribution. In this study, we incorporated both environmental and prey variables estimated at local and regional scale and explored the relative importance of each of them in explaining the spatio-temporal variability in common dolphin data. Although we could not attribute much of the detected distributional shifts to the variables considered in this study, we could conclude that, in the BoB, climate indices and prey species biomasses can play an important role in driving the abundance patterns of the common dolphin.
Further research on climate change effects on common dolphin, however, should focus on comprising the whole distribution range of the species, given the increasingly feasible possibility for combining surveys across areas and regions provided by methods such as those used here. This way, we could address important knowledge gaps that have not been solved here, for example, if the increasing trend found in abundance is due to the arrival of new individuals or it is the result of an overall population growth. Answering to this question will undoubtedly help understand population dynamics and bycatch implications, but meanwhile, we reiterate our call for caution when interpreting the abundance patterns predicted in this study.
Any requests for survey data should be addressed to their owners. In future, some survey data may become open access. Please contact PE (
AA, MLo, and GC conceived and designed the study. AA applied the models and wrote the main text of the manuscript. JW collated and standardized survey data while MP and JT helped with the modeling approach. The remaining co-authors contributed survey data and revised the manuscript. All authors contributed to the article and approved the submitted version.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
This research was funded by the Basque and the Spanish Government through CHALLENGES (CTM2013-47032-R) and EPELECO projects and LIFE-IP URBAN KLIMA 2050 project (Grant agreement no. LIFE18 IPC/ES/000001) which has received funding from European Union’s LIFE program. AA has benefited from a Basque Government scholarship (PRE_2016_1_0134) and MLo from Ramón y Cajal funding (Ministerio de Ciencia e Innovación, RYC-2012-09897).
We thank the multiple organizations, institutions and surveys that provided data (ATLANCET, BIOMAN, CEMMA, CODA, ESAS, EVHOE, IBTS, IFAW, IWDG, JUVENA, KOSMOS, MARINELIFE, ORCA, PELACUS, PELGAS, SAMM, SCANS, SPEA, SWF, WDC; details in
The Supplementary Material for this article can be found online at: