Why Some Birds Migrate (And Others Don’t)

Why do some birds migrate short distances whilst others undertake extraordinary journeys across continents and oceans, whilst many others do not migrate at all?

by GrrlScientist for Forbes | @GrrlScientist

A common swift (Apus apus) flying over Barcelona, Spain. This species is a long-distance migrant, flying annually from southern Africa to northern and central Europe to breed.
pau.artigas / CC-BY-SA 2.0)

Yesterday afternoon, I heard the high-pitched screams of common swifts, Apus apus, for the first time this year as they rocketed across the skies over my head, capturing and consuming flying insects on-the-wing. These beloved birds are long-distance migrants that winter in the southern portion of Africa, and whose appearance throughout northern Europe in early days of May heralds the arrival of spring.

But 40% of bird species are migratory, so slightly more than half of all bird species stay put. For example, I’ve been devotedly feeding a variety of wild bird species throughout the long sub-Arctic winter. Some of them are now moving on to their breeding grounds, whilst most remain here year-round.

“Why do some birds migrate short distances while some others undertake extraordinary journeys across continents, and many others do not migrate at all? [This] question has been puzzling me since I started my PhD,” writes macroecologist Marius Somveille in a blog essay that was published at the same time that the embargo lifted on his new paper, published by Nature Ecology and Evolution.

Indeed, these comprehensive questions have long perplexed scientists, ornithologists and bird watchers. To answer these questions, Dr. Somveille first initiated this research whilst finishing his PhD at Cambridge University and he mostly completed this work during his first postdoc at Oxford University.

Identifying global biodiversity patterns

To do this work, Dr. Somveille and his collaborators, Ana Rodrigues, an ecologist at the Université de Montpellier, and Andrea Manica, an evolutionary ecologist at Cambridge, mapped global patterns for the seasonal distributions of migratory birds. These patterns were then used to statistically test several hypotheses about the processes that underly global species distribution patterns. The analyses indicated that seasonal regions have higher numbers of migratory birds and that they migrate to the closest area they can find that best meets their needs.

“However, it didn’t explain why some species migrate and some don’t,” Dr. Somveille said. “It occurred to me that, ultimately, whether or not migration is a good strategy for a species depends also on what the other species are doing. To investigate this, I needed a mechanistic model.”

Migration is mostly about energy optimisation

But where might Dr. Somveille and his collaborators start looking for an explanation that might help develop a testable mechanistic model? Energy looked like a good place to start: It has long been known that energy plays a critical role in bird migration. Basically, birds’ underlying strategy is to maximize the energy that is seasonally available to them whilst minimizing their energetic costs for obtaining it.

“I started simple; modelling the year-round energy balance of average bird species that I then placed into the environment so as to maximise energy-efficiency,” Dr. Somveille explained.

But energy optimization is more universal than just regulating migration in birds: energy constraints appear to be a key driver in global biodiversity distribution patterns in general. This hypothesis, known as the species-energy relationship (ref), provided Dr. Somveille and his collaborators with a direct link between species-level migratory strategies and the overall species distribution patterns.

“It led me to Lotka’s maximum power principle [PDF], according to which bird species’ distributions should optimise a yearly energetic balance given the environment and competition with other species. Energy-efficiency was an ideal candidate mechanism around which to design the model,” Dr. Somveille explained.

Dr. Somveille and his collaborators constructed a model based upon the assumptions that any particular species’ energetic costs are comprised of thermoregulation, reproduction and migration, as well as basal energetic costs for simply being alive (Figure 1a). The energy available to any particular species during any particular season is the energy that has been produced in a region but has not acquired by any other species living there.

“I estimated parameter values at first directly from the literature (before a fitting procedure was performed later on),” Dr. Somveille said.

Fig. 1 | Model description. a, The model is built from the following three main components: species’ energetic costs (a function of the location of breeding and non breeding ranges, comprising thermoregulation, reproduction and migration costs); energy supply (derived from the NDVI, and variable across space and seasons); and 1,000 simulated range options (the same size as the average bird seasonal range size). Integrating these three components, the model is applied through a sequence of simulation steps whereby a virtual world with the same geography and seasonality as Earth is progressively filled with virtual species. b, At the start of the simulation (T0), the virtual world is empty of bird species (R = 0) and the energy available is equal to the energy supply (EA = ES). In each simulation step (Ti; sub-steps 1 to 4) a new virtual species is added to the virtual world, selected among 1,040,000 candidate species (each being a pair of a breeding and a non-breeding range options) by being the most energy-efficient distribution (lowest ratio between energetic costs and the energy available remaining given the n species already present EA = ES — nEC). As this new species is added (R = n + 1), the energy available. EA is further depleted in the corresponding breeding and non-breeding ranges. The simulation ends (Tend) when the virtual world is nearly saturated with simulated species (EA ≈ 0 in at least one season). The different size shading of the bird shadows indicate that the corresponding energetic costs (see top-left panel of the figure) have different values for different candidate distributions. The grey shading indicates that the two maps within it represent the state of the system at a given step. The purple shading highlights that the explanation on the right-hand side represents what happens during one simulation step.

“I did not really know what outcome to expect.”

Energy optimisation model accurately predicts five observed real-world patterns

Dr. Somveille and his collaborators ran many simulations of their model, starting with an empty virtual world with the same seasonality and geography as the real world, and they progressively filled it with virtual bird species, one at a time. The virtual bird species added to each model simulation was chosen based upon having the the most favourable cost–benefit energetic balance amongst the many available candidate virtual species (Figure 1b). Virtual bird species were added until the virtual world became nearly saturated. In total, 7,783 virtual species were simulated.

Then, virtual seasonal bird species distributions were compared to real-world observations for five empirical patterns: 1. richness in breeding migrants (Figure 2a, b, c, d); 2. richness in non-breeding migrants (Figure 2e, f, g, h); 3. richness in residents (Figure 2i, j, k, l); 4. seasonal difference in richness (Figure 2m, n, o, p); and 5.proportion of migrants (Figure 2q, r, s, t).

Fig. 2 | contrast between empirical patterns in the global spatial distribution of terrestrial birds across seasons and the same patterns simulated through the overall best-fit model. a–d, Richness in breeding migrants. e–h, Richness in non-breeding migrants. i–l, Richness in residents. m–p, Seasonal difference in richness. q–t, Proportion of migrants. Latitudinal trends (c,g,k,o,s) were obtained using Nadaraya–Watson kernel regression estimates (using the ksmooth function from the stats package in R). In the scatterplots of the relationship between the empirical and simulated patterns (d,h,l,p,t), goodness-of-fit was computed using the sum of squared residuals from the 1:1 line (in red). In r, land hexagons with zero-simulated species (for which the proportion of migrants could not be calculated) are shaded in grey. A total of 7,783 virtual species were simulated.

The best-fit model accurately predicted all five observed species parameters.

“It was astonishing to see that the model, based on simplifying assumptions (e.g. no difference between species) and simple rules, could already re-create well all the previously described patterns associated with the global seasonal distribution of birds!” Dr. Somveille said enthusiastically.

I think it is interesting that the model also captured a peculiarity in the global pattern of bird migration: that is, bird migration is mainly a Northern Hemisphere phenomenon (ref).

Species are globally distributed in the most energy-efficient way

“The study indicates that birds are distributed throughout the world in the most energy-efficient way, and that migration helps many species to optimise further their energy budget,” Dr. Somveille explained.

Not only does this model support a mechanistic explanation for the species–energy relationship hypothesis, it also provides a general explanation for migration as a mechanism that allows birds to optimize their energy budget throughout the seasons, and it takes into account competition from other species.

“[W]hat surprised me most about the findings is how a model based on basic energetic principles and ignoring the specific ecologies of species can go such a long way towards explaining the seasonal distribution patterns of all the (land) bird species on the planet,” Dr. Somveille said in email.

This model also provides the foundations for predicting how land-use changes could affect global species distributions as humans out-compete flora and fauna for ever-larger portions of energy.

“It is likely that migratory patterns will be dramatically impacted by human activities (and they already are),” Dr. Somveille elaborated in email.

Canada geese (Branta canadensis) flying in a V-shaped formation. Although Canada geese are a migratory species, some populations do not migrate at all, instead remaining close to parks and golf courses year-round, where there is abundant food and water.
John Benson / Creative Commons Attribution 2.0 Generic license.)

“Migration is a flexible behavior that can be seen as an ecological adjustment, and humans can alter the environment in so many ways — modifying the energy available in the environment positively, such as with bird feeders or landfills, or negatively, such as with urbanisation, but also create barriers to migration or impacting the climate, which affect energetic costs,” Dr. Somveille said.

“I think that the response of migrations/migratory species will be complex,” Dr. Somveille continued. “Some species might shorten their migration while others migrate further. Some species might stop migrating while others start migrating. Models like the one that we developed based on energy optimisation could be very useful for making such predictions under various global change scenarios.”

Although the model was developed using data from terrestrial bird species, Dr. Somveille and his collaborators think it can be applied to other migratory species, such as fishes and whales, as well as to resident species, such as amphibians or even plants, by taking into account that sedentary species may cope with seasons through hibernation or dormancy.

“This is good news for making predictions because there is no reason to believe that birds would not distribute according to the same principle at different times,” Dr. Somveille said. “Our mechanistic model could therefore be useful for predicting the global seasonal distribution of birds back to the last glacial period for instance, but also in a future with global change.”


Marius Somveille, Ana S. L. Rodrigues, and Andrea Manica (2018). Energy efficiency drives the global seasonal distribution of birds, Nature Ecology & Evolution, published online on 7 May 2018 ahead of print | doi:10.1038/s41559–018–0556–9

Also cited:

Karl L. Evans, Philip H. Warren and Kevin J. Gaston (2005). Species–energy relationships at the macroecological scale: a review of the mechanisms, Biological Reviews 80(1):1–25 | doi:10.1017/S1464793104006517

Alfred J. Lotka (1922). Natural selection as a physical principle, Proceedings of the National Academy of Sciences of the United States of America, 8:151–154 | PDF

Marius Somveille, Andrea Manica, Stuart H. M. Butchart, and Ana S. L. Rodrigues (2013). Mapping Global Diversity Patterns for Migratory Birds, PLoS ONE8(8):e70907 | doi:10.1371/journal.pone.0070907

Read more about migration:

GrrlScientist. “How Migratory Birds Solve The Longitude Problem”, Forbes, 22 August 2017. (Medium link.)

GrrlScientist. “Can We Save Europe’s Migratory Birds?”, Forbes, 7 November 2016. (Medium link.)

GrrlScientist. “Watch: Songbirds Return to North America”, The Guardian, 4 July 2015. (Medium link.)

GrrlScientist. “Butterbutt biology: what mitochondria teach us about bird migration”, The Guardian, 3 October 2013. (Medium link.)

GrrlScientist. “Wise old birds teach migration route to young whooping cranes”, The Guardian, 2 September 2013. (Medium link.)

GrrlScientist. “Watch: Discovering dragonflies that cross oceans”, The Guardian, 6 April 2011. (Medium link.)

GrrlScientist. “Fly Me to the Moon: The Incredible Migratory Journey of the Arctic Tern”, ScienceBlogs, 13 January 2010. (Medium link.)

GrrlScientist. “Will The Great Animal Migrations Disappear Forever?” ScienceBlogs, 30 July2008. (Medium link.)

Originally published at Forbes on 7 May 2018.