INTRODUCTION
⌅The dynamics of marine top predator colonies are complex, as they are typically structured in space, so changes over time in a given colony are often explained by the regional dispersal of individuals among patches of the same metapopulation (see e.g. Oro and Pradel 1999 Oro D., Pradel R. 1999. Recruitment of Audouin’s gull to the Ebro Delta colony at metapopulation level in the western Mediterranean. Mar. Ecol. Prog. Ser. 180: 267-273. https://doi.org/10.3354/meps180267 ). Hence, colonization and extinction of individual seabird breeding sites cannot be explained by considering only local factors of individual colonies in isolation from regional conditions (see e.g. Spendelow et al. 1995 Spendelow J.A., Nichols J.D., Nisbet I.C.T., et al. 1995. Estimating annual survival and movement rates of adult within a metapopulation of Roseate Terns. Ecology 76: 2415-2428. https://doi.org/10.2307/2265817 , Fernández-Chacón et al. 2013 Fernández-Chacón A., Genovart M., Pradel R., et al. 2013. When to stay, when to disperse and where to go: survival and dispersal patterns in a spatially structured seabird population. Ecography 36:1117-1126. https://doi.org/10.1111/j.1600-0587.2013.00246.x , Goyert et al. 2018 Goyert H.F., Garton E.O., Poe A.J. 2018. Effects of climate change and environmental variability on the carrying capacity of Alaskan seabird populations. The Auk 135: 975-991. https://doi.org/10.1642/AUK-18-37.1 ). A special case is that of edge populations, as they are typically small and distant from other large colonies that could buffer local population declines by immigration when environmental conditions become unfavourable at the regional scale ( Fagan et al. 1998 Fagan W.F., Stephen R., Cosber C. 1998. How habitat edges change species interactions. Am. Nat. 153: 165-182. https://doi.org/10.1086/303162 ).
Common guillemots (Uria aalge) in the Iberian Peninsula (henceforth Iberia) collapsed during the second half of the 20th century. It has been suggested that this precipitous decline was due to local anthropogenic causes (i.e. direct human persecution and mortality in fishing gear) ( Munilla et al. 2007 Munilla I., Díez C., Velando A. 2007. Are edge bird populations doomed to extinction? A retrospective analysis of the common guillemot collapse in Iberia. Biol. Conserv. 137: 359-371. https://doi.org/10.1016/j.biocon.2007.02.023 ). However, Iberian guillemots may have been involved in metapopulation dynamics (including colonization and extinction of local patches) for centuries or millennia as a result of environmental variability. An alternative hypothesis is that changes in food availability not only affected breeding success or true survival but that food scarcity could also have forced guillemots to permanently emigrate from Iberia.
The presence of guillemots in Iberia has been traced back to the upper Pleistocene rocks of Gibraltar, Málaga, Valencia, Asturias and the Basque Country ( Cortés et al. 1980 Cortés J.E., Fynlaison J.C., Mosquera M.A., García E.F.J. 1980. The birds of Gibraltar. Gibraltar Bookshop, Gibraltar., Carrasquilla 1993 Carrasquilla F. H. 1993. Catálogo provisional de los yacimientos con aves del Cuaternario de la Península Ibérica. Archaeofauna 2: 231-275., Elorza 2014 Elorza M. 2014. Explotación de aves marinas en el Tardiglaciar del Golfo de Bizkaia: las aves de Santa Catalina. Kobie. Bizkaiko Arkeologi Indusketak -Excavaciones Arqueológicas en Bizkaia 4: 263-296., Fig. 1A ) and to the Holocene rocks of Málaga and the Basque Country ( Eastham 1986 Eastham A. 1986. The birds of Cueva de Nerja. In: Jordá J.F. (ed.): La prehistoria de la Cueva de Nerja (Málaga). Trabajos sobre la Cueva de Nerja 1: 109-131. Patronato de la Cueva de Nerja, Málaga, Spain., Elorza and Marco 1993 Elorza M., Marco A. S. 1993. Postglacial fossil Great Auk and associated avian fauna from the Biscay Bay. Munibe Antropologia-Arkeologia 45: 179-185.). In Brittany, France, guillemots were also present during postglacial times, 5500 to 2500 years before present (BP) ( Fig. 1A ) ( Tresset 2005 Tresset A. 2005. L’avifaune des sites mésolithiques et néolithiques (5500 à 2500 av. J.-C.) de Bretagne: implications ethnologiques et biogéographiques. Rev. Paléobiol. 84: 95.). However, fossils cannot reveal whether the species was wintering or breeding around that time.
The modern presence of breeding guillemots in NW Spain and Portugal is only known to have certainly occurred since the late 19th century and early 20th century ( Tait 1924 Tait W.C. 1924. The birds of Portugal. Nature 114: 8. https://doi.org/10.1038/114008b0 ). One of the possible scenarios is that guillemots were present as breeders in southern Europe during the cold and productive upper Pleistocene and beginnings of the Holocene, but that they became rare or intermittent breeders in Iberia during most of the Holocene, with warmer and less productive waters (the Holocene Climate Optimum took place between 9000 and 5000 years BP, with a thermal maximum around 8000 years BP).
From this perspective the most recent of these recolonization events of Atlantic Iberia could have happened in the final decades of the 19th century and the early 20th century (when the oldest written references to breeding guillemots are dated), ending at some unknown point between the 1960s and 1970s (1962-1973 according to available counts), and this is the focus of the present study. Apparent mortality, identified as the main mechanism of the guillemot collapse in Iberia ( Munilla et al. 2007 Munilla I., Díez C., Velando A. 2007. Are edge bird populations doomed to extinction? A retrospective analysis of the common guillemot collapse in Iberia. Biol. Conserv. 137: 359-371. https://doi.org/10.1016/j.biocon.2007.02.023 ), could hence reflect true mortality but also emigration/immigration to other metapopulation patches as a result of deteriorating regional environmental conditions ( Martínez-Abraín 2015 Martínez-Abraín A. 2015. Are edge bird populations doomed to extinction: A response to Munilla et al. Biol. Conserv. 191: 843-844. https://doi.org/10.1016/j.biocon.2015.07.014 , but see Munilla and Velando 2015 Munilla I., Velando A. 2015. The Iberian guillemot population crash: a plea for action at the margins. Biol. Conserv. 100: 842. https://doi.org/10.1016/j.biocon.2015.07.015 ).
Because guillemots are known to feed on clupeids, and the study region supports a rich clupeid fishery, we propose as a working hypothesis that they were facultative specialists on clupeids in Iberia (i.e. they specialized locally in one prey type though it could changed from site to site), and that the crash of small pelagic fisheries likely caused a rapid collapse of guillemots. The diet of breeding guillemots in Atlantic Iberia is unknown, but the chicks and adults of the species typically consume fish from the families Ammodytidae, Clupeidae and Gadidae in European colonies (see e.g. Barrett and Krasnov 1996 Barrett R. T., Krasnov Y. V. 1996. Recent responses to changes in stocks of prey species by seabirds breeding in the southern Barents Sea. ICES J. Mar. Sci. 53:713-722. https://doi.org/10.1006/jmsc.1996.0090 , Anderson et al. 2013 Anderson H.B., Evans P.G.H., Potts J.M., et al. 2013. The diet of common guillemot (Uria aalge) chicks at colonies in the UK, 2006-2011: evidence for changing prey communities in the North Sea. Ibis 156: 23-34. https://doi.org/10.1111/ibi.12099 , Budge et al. 2011 Budge J., Barret R.T., Pedersen T. 2011. Optimal foraging in chick-raising Common Guillemots (Uria aalge). J. Ornithol. 152: 253-259. https://doi.org/10.1007/s10336-010-0578-9 ). In some colonies they consume a high percentage of two Clupeidae: Atlantic herring (Clupea arengus) and European sprat (Sprattus sprattus) (99% in Wales and 96% in the Baltic Sea; Harris and Wanless 1984 Harris M.P., Wanless S. 1984. Fish fed to young guillemots, Uria aalge, and used in display on the Isle of May, Scotland. J. Zool. 207: 441-458. https://doi.org/10.1111/j.1469-7998.1985.tb04942.x , Öesterblom and Olsson 2002 Öesterblom H., Olsson O. 2002. Changes in feeding behaviour and reproductive success in the Common Guillemot Uria aalge on the island of Stora Karlsö. Ornis Svecica: 1-2: 53-62.). Moreover, a high percentage of the winter diet of guillemots in France is known to be composed of anchovies (Engraulis encrasicolus) and sardines (Sardina pilchardus) ( Pasquet 1988 Pasquet E. 1988. A study of the diet of common guillemot Uria aalge and razorbill Alca torda, wintering in French waters. Alauda 56: 8-21.). Direct, although anecdotal, evidence of sardine consumption by 20th-century Iberian guillemots comes from the report that out of 12 collection skins of birds captured in Galicia during the breeding seasons of 1964 and 1965 and preserved at the Doñana Biological Station, one of them contained a sardine in its stomach (Carlos Urdiales, pers. com.).
Collapses of large seabird colonies resulting from sudden crashes in its main local prey have been commonly reported (e.g. Cury et al. 2011 Cury P.M., Boyd I., Bonhommeau S., et al. 2011. Global seabird response to forage fish depletion: one-third for the birds. Science 334: 1703-1706. https://doi.org/10.1126/science.1212928 , Erikstad et al. 2013 Erikstad K.E., Reiertsen T.K., Barrett R.T., et al. 2013. Seabird-fish interactions: the fall and rise of a common guillemot Uria aalge population. Mar. Ecol. Prog. Ser. 475:267-276. https://doi.org/10.3354/meps10084 , Sherley et al. 2013 Sherley R.B., Underhill L.G., Barham B.J., et al. 2013. Influence of local and regional prey availability on breeding performance of African penguins Spheniscus demersus. Mar. Ecol. Prog. Ser. 473: 291-301. https://doi.org/10.3354/meps10070 ). Crashes in prey populations can cause the decline of predatory seabird populations by negatively influencing apparent survival (i.e. either true mortality or dispersal) or by heavily reducing offspring production and recruitment (e.g. Sandvik et al. 2012 Sandvik H., Erikstad K.E., Saether B-E. 2012. Climate affects seabird population dynamics both via reproduction and adult survival. Mar. Ecol. Prog. Ser. 454: 273-284. https://doi.org/10.3354/meps09558 , Waugh et al. 2015 Waugh S.M., Barbraud C., Adams L., Freeman A. 2015. Modelling the demography and population dynamics of a subtropical seabird, and the influence of environmental factors. Condor 117: 147-164. https://doi.org/10.1650/CONDOR-14-141.1 , Fayet et al. 2021 Fayet A.L., Clucas G.V., Anker-Nilssen T., et al. 2021. Local prey shortages drive foraging costs and breeding success in a declining seabird, the Atlantic puffin. J. Anim. Ecol. 90: 1164. https://doi.org/10.1111/1365-2656.13442 ). Declines due to decreases in apparent survival happen faster than those caused by reduced fecundity and/or recruitment, which typically show a pattern of delayed effects arising from the late-maturing age of long-lived seabirds ( Erikstad et al. 2013 Erikstad K.E., Reiertsen T.K., Barrett R.T., et al. 2013. Seabird-fish interactions: the fall and rise of a common guillemot Uria aalge population. Mar. Ecol. Prog. Ser. 475:267-276. https://doi.org/10.3354/meps10084 , Meade et al. 2013 Meade J., Hatchwell B.J., Blanchard J.L., Birkhead TR. 2013. The population increase of common guillemots (Uria aalge) on Skomer Island is explained by intrinsic demographic properties. J. Avian Biol. 44: 55-61. https://doi.org/10.1111/j.1600-048X.2012.05742.x , Reynolds et al. 2019 Reynolds S.J., Hughes B.J., Wearn C.P., et al. 2019. Long-term dietary shift and population decline of a pelagic seabird: A health check on the tropical Atlantic? Global Change Biol. 25: 1383-1394. https://doi.org/10.1111/gcb.14560 ). Breeding dispersal is not common in guillemots, as is to be expected in long-lived species in which territorial vacancies are scarce ( Harts et al. 2016 Harts A.M.F., Jaatinen K., Kokko H. 2016. Evolution of natal and breeding dispersal: when is a territory and asset worth protecting? Behav. Ecol. 27: 287-294. https://doi.org/10.1093/beheco/arv148 ). However, a catastrophic crash in staple prey species can force massive and rapid runaway dispersal of adults, as sometimes occurs in nesting colonial birds when a predator is involved (see Oro 2020 Oro D. 2020. Perturbation, behavioural feedbacks, and population dynamics in social animals. Oxford University Press, Oxford. https://doi.org/10.1093/oso/9780198849834.001.0001 ). Additionally, sink seabird populations can experience fast declines if the rescue from source colonies by dispersal is discontinued ( Bonnaud et al. 2009 Bonnaud E., Bourgeois K., Vidal E., et al. 2009. How can the Yelkouan shearwater survive feral cat predation? A meta-population structure as a solution? Pop. Ecol. 51:261-270. https://doi.org/10.1007/s10144-008-0134-0 , Sanz-Aguilar et al. 2016 Sanz-Aguilar A., Igual J.M., Tavecchia G., et al. 2016. When immigration masks threats: The rescue effect of a Scopoli’s shearwater colony in the western Mediterranean as a case study. Biol. Conserv. 198: 33-36. https://doi.org/10.1016/j.biocon.2016.03.034 , Seward et al. 2019 Seward A., Ratcliffe N., Newton S., et al. 2019. Metapopulation dynamics of roseate terns: Sources, sinks and implications for conservation management decisions. J. Anim. Ecol. 88:138-153. https://doi.org/10.1111/1365-2656.12904 ).
Here we explore possible associations between the incomplete time series of breeding guillemot numbers in NW Iberia and the complete time series of sardine and anchovy landings at the regional level, used as a proxy of overall stock size. We expected to find some strong relationship between the time series of predator and likely prey species, in a similar way to that recently found by Martínez-Abraín et al. (2019) Martínez-Abraín A., Santidrián Tomillo P., et al. 2019. Delayed predator-prey collapses: the case of black-legged kittiwakes and Iberian sardines. Mar. Ecol. Prog. Ser. 631: 201-207. https://doi.org/10.3354/meps13164 for the NW Iberian collapse of the black-legged kittiwake (Rissa tridactyla) population.
METHODS
⌅We first compiled the most comprehensive time series available for common guillemot breeding numbers in Galicia (NW Spain, Fig. 1B ) from 1961 to 2015 (55 years). Guillemot information available for 1961 is based on an estimate made by Bárcena (1985) Bárcena F. 1985. Localización e inventario de las colonias de Arao común, Uria aalge Pontopp., en las costas de Galicia: determinación de las posibles causas de su desaparición. Boletín de la Estación Central de Ecología 28: 19-28. from surveys of sailors and fishermen along the coast of Galicia. This estimate is consistent with the few quantitative data available before the 1960s that characterize the pre-collapse situation, showing relatively large numbers such as the 660 individuals on Sisargas Islands counted by Bernis (1948) Bernis F. 1948. Las aves de las islas Sisargas en julio. Bol. Soc. Esp. Hist. Nat. 46: 647-814. and the ca. 6000 pairs in Berlengas Islands, Portugal estimated in 1939 ( Lockley 1952 Lockley R.M. 1952. Note on the birds of the islands of the Berlenga (Portugal) and Desertas and Baixo (Madeira) and the selvages. Ibis 94: 144-157. https://doi.org/10.1111/j.1474-919X.1952.tb01795.x ). The next available counts after the 1961 estimate come from 1974-1976, and only provide partial information from four of the eight known breeding sites ( Bermejo and Rodríguez-Silvar 1983 Bermejo A., Rodríguez-Silvar J. 1983. Situación del arao común, Uria aalge ibericus, como especie nidificante de Galicia. Alytes I: 343-346., Bárcena 1985 Bárcena F. 1985. Localización e inventario de las colonias de Arao común, Uria aalge Pontopp., en las costas de Galicia: determinación de las posibles causas de su desaparición. Boletín de la Estación Central de Ecología 28: 19-28.). Although the total number of pairs for that point in time should be somewhat larger, the data reflect unequivocally a reduction of two orders of magnitude in the number of individuals (from ca. 3000 to only 40). The next available counts correspond to the 1980s (Bárcena 1985). From then on, systematic censuses for the three remaining guillemot colonies became more frequent and were taken from Mouriño et al. (2004) Mouriño J., Arcos F., Alcalde A. 2004. Arao Común Uria aalge. In: Madroño A., González C., Atienza J.C. (eds): Libro Rojo de las Aves Reproductoras de España, pp. 261-264. Dirección General para la Biodiversidad-SEO/BirdLife. Madrid., adding our own unpublished data (Jorge Mouriño) since 2003. We provide a more detailed summary of guillemot count data in Appendix 1 , indicating the period, the method used, the number of birds counted, the colony, the number of colonies and the data source. Guillemot counts in colonies were performed from May to June.
Secondly, we contrasted the guillemot time series with the information on sardine and anchovy landings for the period 1960-2015. Fish landings (May-October for sardines and April-June for anchovies) were obtained from the International Council for the Exploration of the Sea (ICES) official databases and used areas 8bc (anchovies) and 8c (sardines) ( ICES 2017 International Council for the Exploration of the Sea (ICES). 2017. Report of the working group on southern horse mackerel, anchovy and sardine (WGHANSA), 24-29 June 2017, Bilbao, Spain. ICES CM 2017/ACOM 17. ICES, Copenhagen.). In 2017 targeted anchovies were mostly Age 1 and Age 2 (around 13 cm in modal length) and all were sexually mature; targeted sardines also belonged to ages 1 and 2 and had a modal length of 16 cm ( ICES 2017 International Council for the Exploration of the Sea (ICES). 2017. Report of the working group on southern horse mackerel, anchovy and sardine (WGHANSA), 24-29 June 2017, Bilbao, Spain. ICES CM 2017/ACOM 17. ICES, Copenhagen.). These mean prey sizes are among the upper range of staple prey sizes delivered to chicks in northern latitudes (e.g. mean Sprattus length 11.5 cm; range 10.6-13.1 cm in Sweden; Enekvist 2003 Enekvist E. 2003. Energy intake of common guillemot, Uria aalge, chicks at Stora Karlsö, Sweeden. Master Thesis, Gotland University. http://www.diva-portal.org/smash/get/diva2:279689/FULLTEXT01.pdf (last accessed 10/08/2022).). Fish landings were assumed to be a good proxy of total clupeid stock size in the study region because fishing boats are added to the fishery when the stock size is large and removed when the stock shrinks. In addition, although the horsepower of fishing boats has increased over time, clupeid landings have shown an almost monotonic declining trend (anchovies since the 1960s and sardine since the early 1990s). Hence, if landings clearly follow a declining trend despite the increase in fishing effort or horsepower, the trend must reflect an actual decline, even stronger than that reflected by our uncorrected landings. Our approach can be considered conservative.
We fitted general linear models to ln-transformed guillemot counts and ln-transformed sardine/anchovy landings. A strong relationship (assessed by means of the coefficient of determination r2 in comparison with models fitted to untransformed data) indicates an exponential relationship between the untransformed variables. Models were fitted using the software environment R and graphs were plotted using the ggplot2 library in R ( https://www.r-project.org/ ).
We analysed the occurrence of regime shifts in the anchovy landings time series following Rodionov (2004) Rodionov S.N. 2004. A sequential algorithm for testing climate regime shifts. Geophys Res Lett 31: L09204. https://doi.org/10.1029/2004GL019448 , a technique based on sequential t-test analysis. To detect regime shifts, we used an add-in for an Excel spreadsheet developed by the National Oceanic and Atmospheric Administration (www.beringclimate.noaa.gov/regimes/index.html). The cut-off length was set at 10 years, following Cabrero et al. (2019) Cabrero A., González-Nuevo G., Gago J., Cabanas, J.M. 2019. Study of sardine (Sardina pilchardus) regime shifts in the Iberian Atlantic waters. Fish. Oceanogr. 28: 305−316. https://doi.org/10.1111/fog.12410 in a study of Iberian sardines. Huber’s weight parameter was set at 1 (standard deviations) to account for the outliers by a weighting factor. Statistical significance was set at α=0.05. Pre-whitening was performed before regime shift detection to prevent autocorrelation in time series that could increase the number of incorrectly identified regime shifts. Red noise was modelled with a first-order autoregressive model, and its coefficients were estimated by means of the ordinary least squares method. Since there were no counts of guillemots between 1961 and 1974, we were not able to run the regime shift tool for the common guillemot time series, as we did with both the sardine and anchovy landings time series. The use of a tool to fill in the missing values in the guillemot time series of counts was deemed unadvisable because of the large number of consecutive missing values (e.g. from a single count in 1961 to the next one in 1974). The large number of missing values in the guillemot time series also made it impossible to use more ad hoc analytical tools such as time-series analysis tools for the detrending of time series or the study of time delays.
Maps were generated using the free Maptool program available at www.seaturtle.org .
RESULTS
⌅The abrupt collapse of guillemot numbers in an undetermined year between 1962 and 1973 can be deduced from Figure 2A . The time series of sardine and anchovy landings showed that a regime shift took place in sardines in 1968 and in anchovies in 1969 ( Fig. 2B ). These dates of prey crashes (1968 and 1969) might narrow down the collapse date of guillemots to the period 1968-1970 if the response of guillemots to prey scarcity was fast (i.e. the year after the sardine crash or the year after the anchovy crash, which would be two years after the sardine crash). Regime shifts in the sardine landings time series were also detected in 1980, 1988, 1998 and 2012, following the decadal-scale periodicity of sardine abundance cycles reported previously in the specialized literature ( Cabrero et al. 2019 Cabrero A., González-Nuevo G., Gago J., Cabanas, J.M. 2019. Study of sardine (Sardina pilchardus) regime shifts in the Iberian Atlantic waters. Fish. Oceanogr. 28: 305−316. https://doi.org/10.1111/fog.12410 ). A second regime shift in anchovy landings was detected in 1979. All those regime shifts were not considered in our analyses as they happened when only a few guillemot individuals were remaining.
The overall relationship between guillemot numbers and sardine landings was exponential (see Appendix 2 for the results of a linear model fitted to ln-transformed data, r2 adj=0.52 compared with non-transformed data, r2 adj=0.27).
The overall relationship between guillemot numbers and anchovy landings was linear, but completely dependent on the large 1960s large estimate of guillemots (see Appendix 2 for the results of a general linear model fitted to untransformed data, r2 adj=0.72, compared with ln-transformed data, r2 adj=0.13).
When the high 1961 pre-collapse guillemot count was excluded, general linear models fitted to both transformed (r2 adj=0.61; Fig. 3 ) and untransformed guillemot and sardine data (r2 adj=0.60) provided similar results, suggesting a linear relationship between the two variables at the scale of tens of guillemots or fewer.
However, we detected no relationship between the amount of anchovies landed and the number of guillemots after the collapse using either ln-transformed (r2 adj=0.04) or untransformed data (r2 adj=0.14).
As an exercise to assess the possibility of getting spurious results in our analyses, we detrended the guillemot and sardine landings time series from 2003 to 2015 (a continuous guillemot time series without missing data). The linear relationship was not statistically significant (F=1.6; 1 and 11 degrees of freedom; p-value=0.23). However, the detrended guillemot time series was shorter than the one we analysed (from 1974 on) and included guillemot abundance only at the scale of individuals, an order of magnitude smaller than the analysed time series with counts of up to 80 individuals.
| Parameter | Year | RSI | Conf |
|---|---|---|---|
| Sardine landings | 1968 | -0.77 | 1.0×10 -6 |
| Sardine landings | 1980 | 0.58 | 3.8×10-5 |
| Sardine landings | 1988 | -1.0 | 1.9×10-5 |
| Sardine landings | 1998 | -0.4 | 2.9×10-7 |
| Sardine landings | 2012 | -1.1 | 6.2×10-6 |
| Anchovy landings | 1969 | -0.36 | 0.0012 |
| Anchovy landings | 1979 | -0.62 | 0.0002 |
DISCUSSION
⌅Our results allowed us to reduce the current uncertainty about the dates of the guillemot collapse in Atlantic Iberia, suggesting that it most likely coincided with or immediately followed the crash of sardines and anchovies (1968-1970). A rapid response of guillemots to the abrupt decline of the two species of small pelagic fish seems feasible considering the colonization/extinction dynamics of the breeding population of black-legged kittiwakes that occurred two decades later in the same area. Kittiwakes colonized NW Spain and Portugal in the mid-1970s ( Rodríguez-Silvar and Bermejo 1975 Rodríguez-Silvar J., Bermejo A. 1975. Primera nidificación de la gaviota tridáctila (Rissa tridactyla) en el SW de Europa. Ardeola 21: 409−414.), coinciding with a temporal recovery of the sardine stock during the 1970s and 1980s (see Fig. 3B for a peak of sardine landings in that period; Cendrero 2002 Cendrero O. 2002. Sardine and anchovy crisis in northern Spain: natural variations or an effect of human activities? ICES Mar. Sci. 215: 279-285., Cabrero et al. 2019 Cabrero A., González-Nuevo G., Gago J., Cabanas, J.M. 2019. Study of sardine (Sardina pilchardus) regime shifts in the Iberian Atlantic waters. Fish. Oceanogr. 28: 305−316. https://doi.org/10.1111/fog.12410 ). The kittiwake colonies, which were also located at the southernmost limit of their European distribution, collapsed two decades after their formation, coinciding with a severe crash of Iberian sardines in 1991 (see Martínez-Abraín et al. 2019 Martínez-Abraín A., Santidrián Tomillo P., et al. 2019. Delayed predator-prey collapses: the case of black-legged kittiwakes and Iberian sardines. Mar. Ecol. Prog. Ser. 631: 201-207. https://doi.org/10.3354/meps13164 ) that has been permanent ever since. Guillemots and kittiwakes shared a similar nesting habitat along the Atlantic Iberian coast (i.e. cliffs and caves on small islets or large rock pinnacles), and even nested on the same locations at times (i.e. the Sisargas Islands, Cape Vilán; Fig. 1A ). In both kittiwakes and guillemots, small pelagic fish abundance may have played a key role in regulating (bottom-up) metapopulation dynamics at the southernmost edge of European distribution.
Cases of rapid collapse of guillemot populations associated with crashes in their staple prey have been previously reported. For example, an 80% decline in a colony with 245000 guillemot pairs occurred in only one year on Bear Island, located on the Svalbard archipelago, Barents Sea (Norway). In this case, the decline was linked to low abundance indices of fish prey ( Anker-Nilssen and Barrett 1991 Anker-Nilssen T., Barrett R.T. 1991. Status of seabirds in northern Norway. British Birds 84: 329-341., Erikstad et al. 2013 Erikstad K.E., Reiertsen T.K., Barrett R.T., et al. 2013. Seabird-fish interactions: the fall and rise of a common guillemot Uria aalge population. Mar. Ecol. Prog. Ser. 475:267-276. https://doi.org/10.3354/meps10084 ). Similarly, Gaston et al. (2009) Gaston A.J., Bertram D.F., Boyne A.W., et al. 2009. Changes in Canadian seabird populations and ecology since 1970 in relation to changes in oceanography and food webs. Environ. Rev. 17: 267-286. https://doi.org/10.1139/A09-013 found that long- and even short-term trends of Canadian seabirds (including guillemots) were strongly influenced by physical changes in the marine environment surrounding colonies, through food-web effects.
Local crashes of small pelagic fish can be driven by a severe primary productivity crisis arising from oceanographic changes (see e.g. Goyert et al. 2018 Goyert H.F., Garton E.O., Poe A.J. 2018. Effects of climate change and environmental variability on the carrying capacity of Alaskan seabird populations. The Auk 135: 975-991. https://doi.org/10.1642/AUK-18-37.1 ), sometimes in interaction with overfishing. Other causes include recruitment problems after intense catches ( Cendrero 2002 Cendrero O. 2002. Sardine and anchovy crisis in northern Spain: natural variations or an effect of human activities? ICES Mar. Sci. 215: 279-285.) and mass displacements of fish due to changes in sea water temperature ( Junquera 1986 Junquera S. 1986. Pèche de l’anchois (Engraulis encrasicholus) dans le Golfe de Gascogne et sur le litoral atlantique de Galice depuis 1920. Variations quantitatives. Revue des Travaux de l’Institut des Pêches Maritimes 48: 133-142., 1991 Junquera S. 1991. Study of the population diversity of anchovy (Engraulis encrasicolus L., 1758) (Pisces, Engraulidae) by means of the canonical analysis of morphological characters and biological parameters, Doctoral thesis. University of Oviedo, Spain. 222 pp (In Spanish).).
Iberian sardine stock dynamics in the second half of the 20th century were linked with changes in oceanographic conditions in interaction with overfishing ( Holling 1973 Holling C.S. 1973. Resilience and stability of ecological systems. Ann. Rev. Ecol. Sys. 4: 1-23. https://doi.org/10.1146/annurev.es.04.110173.000245 , Cendrero 2002 Cendrero O. 2002. Sardine and anchovy crisis in northern Spain: natural variations or an effect of human activities? ICES Mar. Sci. 215: 279-285.). The Iberian anchovy crash in the late 1960s seemingly resulted from the shrinkage of the anchovy distribution (i.e. a permanent emigration of anchovies towards the eastern side of the Bay of Biscay) as a result of cold-water marine upwellings (see Junquera 1986 Junquera S. 1986. Pèche de l’anchois (Engraulis encrasicholus) dans le Golfe de Gascogne et sur le litoral atlantique de Galice depuis 1920. Variations quantitatives. Revue des Travaux de l’Institut des Pêches Maritimes 48: 133-142., 1991 Junquera S. 1991. Study of the population diversity of anchovy (Engraulis encrasicolus L., 1758) (Pisces, Engraulidae) by means of the canonical analysis of morphological characters and biological parameters, Doctoral thesis. University of Oviedo, Spain. 222 pp (In Spanish)., Uriarte et al. 1996 Uriarte A., Prouzet P., Villamor B. 1996. Bay of Biscay and Ibero Atlantic anchovy populations and their fisheries. Sci. Mar. 60 (Supl. 2): 237-255., Cendrero 2002 Cendrero O. 2002. Sardine and anchovy crisis in northern Spain: natural variations or an effect of human activities? ICES Mar. Sci. 215: 279-285.). This might well explain the strong linear relationship detected between anchovy landings and guillemot numbers when the whole time series was used versus the absence of relationship when only the guillemot time series without the first 1960s datum was considered. Simply, anchovies emigrated en masse out of the study area, and hence could not influence any further guillemot dynamics after their collapse because of the low anchovy densities. Additionally, the period in which anchovies are more abundant and hence are fished (April-June) only overlaps partially with the breeding period of guillemots (May-July), whereas the period of sardine abundance (May-October) is longer and overlaps better.
Relatively rapid recoveries can take place in large guillemot populations as a result of the exchange of individuals among colonies, making them more resilient to prey crashes than small colonies. For example, a recovery from 36000 pairs to 95000 pairs was reported in Bear Island in just two years by Anker-Nilssen and Barrett (1991) Anker-Nilssen T., Barrett R.T. 1991. Status of seabirds in northern Norway. British Birds 84: 329-341.. However, rapid recoveries do not always occur. For example, the guillemot population of Hornøya Island (Troms og Finnmark, Barents Sea, Norway) took 20 years to recover to the numbers reached before the collapse ( Eriskstad et al. 2013 Erikstad K.E., Reiertsen T.K., Barrett R.T., et al. 2013. Seabird-fish interactions: the fall and rise of a common guillemot Uria aalge population. Mar. Ecol. Prog. Ser. 475:267-276. https://doi.org/10.3354/meps10084 ).
Small edge populations are less likely to recover from abrupt environmental changes, as they experience both Allee effects (i.e. inverse density-dependence or depensation) and poor rescue by immigration ( Schippers et al. 2011 Schippers P., Stienen E W.M., Schotman A.G.M., et al. 2011. The consequences of being colonial: Allee effects in metapopulations of seabirds. Ecol. Model. 222: 3061-3070. https://doi.org/10.1016/j.ecolmodel.2011.05.022 , Sanz-Aguilar et al. 2016 Sanz-Aguilar A., Igual J.M., Tavecchia G., et al. 2016. When immigration masks threats: The rescue effect of a Scopoli’s shearwater colony in the western Mediterranean as a case study. Biol. Conserv. 198: 33-36. https://doi.org/10.1016/j.biocon.2016.03.034 ). This is especially the case in populations in which the geographic range does not occur along a continuous distribution but shows a large gap in relation to the bulk of the population. This may have been the case of the common guillemot populations in Spain and Portugal, which were relatively small colonies (one recorded in Portugal and nine in Spain) and showed a considerable geographical gap between them and the French Bretagne and southern British populations (500-1000 km away; see Fig 1b ). Probably owing to this marginal situation, the local Iberian guillemot populations were not rescued by immigration. Surprisingly, the guillemot time series showed a stable state of quasi-extinction during the last 13 years of study (2003-2015), even scoring 0 pairs in one year (2012) and then recovering a few pairs in subsequent years (2013-2015). This pattern was already described and discussed by Oro (2020) Oro D. 2020. Perturbation, behavioural feedbacks, and population dynamics in social animals. Oxford University Press, Oxford. https://doi.org/10.1093/oso/9780198849834.001.0001 using the same guillemot time series and some other series.
Current sardine densities in the study area are far from the large figures of decades past ( Cabrero et al. 2019 Cabrero A., González-Nuevo G., Gago J., Cabanas, J.M. 2019. Study of sardine (Sardina pilchardus) regime shifts in the Iberian Atlantic waters. Fish. Oceanogr. 28: 305−316. https://doi.org/10.1111/fog.12410 ). The same is true for anchovies, whose crashes in the 1960s to 1970s have been permanent until now (i.e. multi-decadal hysteresis). Only during the last few fishing seasons (2021), have anchovy catches seemingly shown some recovery along the western Bay of Biscay (according to media news on captures of anchovies in the western Cantabrian Sea). If sardine and/or anchovies were to again reach suitable densities to sustain a new recolonization of guillemots, conditions in the large northern European guillemot colonies (either as good as to provide a surplus or as bad as to cause long-distance dispersal) would be critical for the return of guillemots to southern European latitudes (see Oro 2020 Oro D. 2020. Perturbation, behavioural feedbacks, and population dynamics in social animals. Oxford University Press, Oxford. https://doi.org/10.1093/oso/9780198849834.001.0001 ). However, current sea warming conditions promote reduced primary productivity by increasing the vertical stratification of the water column, which can make it difficult to recover past clupeid densities.
A word of caution finally to account for the statistically non-significant results of our exercise with detrended data from 2003 to 2015 (continuous time series without missing data). Our results would be much more reliable if performed on complete and detrended time series, but this was not feasible because of the large number of missing data in the guillemot time series. Our results make biological sense, although the possibility of getting some spurious results (cause-effect relationships) when working with poor data sets is always present.