Compensation between species occurs when as one species’ abundance or biomass goes up, some other species’ abundances go down so that the summed abundance or biomass at the community level stays relatively steady. This property is believed to have ramifications for how biodiversity affects ecosystem functioning: if species diversity is to maintain some stability of biomass at the ecosystem level, we expect such compensation to arise. The opposite of compensation is synchrony, the most extreme form of which is all species fluctuating in unison. In a recent Ecology & Evolution article¹, we ask whether compensation or synchrony is found in wetland bird counts, using a long-term survey at the Teich Reserve, Arcachon Bay. We find that although synchrony dominates, compensation occurs at long temporal scales and broad functional scales: compensation is found by “zooming out” temporally and functionally.

In many datasets so far, ecologists have found more synchrony than compensation. People have looked at communities (sets of multiple species) of plants, of plankton, of mammals as well as some insect communities. Although compensation can occur, it is quite rare and synchrony usually dominates. Why? A first idea is that there may be a dominant frequency at which most species numbers fluctuate, for example whenever there is an annual cycle due to seasonal temperature variation. Thus researchers have looked at statistics that extract effects at that specific frequency, and looked at effects possibly with lower or higher periodicities. And yet compensation is still rare.

We were interested in the Teich reserve for several reasons. First, bird monitoring started in 1973², birds are counted every month (actually, several times during the month but data were aggregated at monthly scale), and we had data up to 2016 when we started analyzing it. A second reason is that multiple factors potentially leading to compensation were present:

• many bird species are at the same trophic level, sometimes eating the same food and potentially competing

• there has been a progressive change in the management of water levels (more pronounced around 2006), which led to some groups being favoured, like waders over ducks.

These two causes can generate compensation at the whole-community level. But let’s say that we separate our community of wetland birds into waders and waterfowl (ducks, swans and geese). We could expect that competition will generate compensation within a group (e.g., within waders, since they eat similar foods) and a change in the environment will lead to compensation between groups (e.g. birds in group 1 react similarly to the environment while birds in group 2 react in an opposite way). In our case, both things probably happen at the same time, so what should we expect at the whole-community level? Well, as usual it’s complicated. To answer that, we have used statistics and a large number of simulations presented as supplementary material. We have found that when there’s both competition and a change in the environment, a strong response to the environment can make compensation less likely at the whole-community level. The reason is, we then have two groups whose numbers react in the same way but the strong response to the environment tends to synchronize member species within a group.

This explains why it is difficult to see compensation by just looking at the community level. Still, in our case, there might be some compensation at long timescales (above 5 years), which may translate the fact that the community as a whole is slowly changing in response to a change in the environment. However, compensation can also be found at the annual scale, in winter, when summing the numbers of waders, summing the numbers of waterfowl, and then looking at the correlation between them, which is negative.

Another reason why it is hard to find compensation is that some species intrinsically fluctuate much more than the rest. It is a classic fact of ecology that in any community, a handful of species represent most of the individuals that we see, while many rare species have low numbers. However, in this context of skewed abundance distribution, an inequality in the level of fluctuation also matters. If one species fluctuates to a degree that the sum of all other similar species fluctuates less, then no increase in the other species will manage to offset a decrease in the numbers of the dominant one. In other words, compensation is mathematically unfeasible unless we average some of these fluctuations over longer time intervals (another reason why longer timescales are useful). We have also shown this with simulations, and the phenomenon may appear for some species in our data (e.g. within calidrids, as the dunlin dominates in both mean abundance and variation). While you may think this is just a question of using biomass rather than abundance it is not–our results are robust to using biomass instead.

Thus, there is some compensation in the data but

• it appears only at long time scales that average some of the short-term variation in common species

• compensation occurs more frequently between than within functional groups like waders or waterfowl

While the study area is arguably small (120 ha) for birds, and this is only one wetland, these patterns might be more general. There are very many ecosystems where migration generates massive influxes of some dominant species with highly variable numbers. This happens up to the polar regions where some of these birds go breeding. Very few ecosystems are in fact truly closed or well-approximated as such. Therefore we might expect that few ecosystems will easily show compensation unless one averages data over large time intervals and broad groups of species. Or that one picks specific pairs of species that have good reasons to replace one another, although in this case as well, as we show using cormorants vs grey herons and little egrets, differences in the amplitude of abundance variation between species can make compensation difficult to find.

Above I have mentioned the scientific results that we have obtained in this study, largely theoretically-driven. There are two interesting practical questions a birdwatcher could ask:

• why didn’t you study a more practical issue, this is a reserve after all?

• how come that you started in 2016 and publish this in 2022?

These are two fair questions. First, we initially had in mind to study a more practical question, the effect of management and disturbances on these bird communities (perhaps we still might!). One difficulty was that the main management practice consists in setting the water levels, which is done through locks (the area was formerly a fish reservoir³). Unfortunately, we have very good records of the number of birds in the reserve but no records of the water levels within the reserve (at least not on a monthly basis from the 80s!). We did study synchrony before and after 2006 though, when the water level management changed quite a bit, but did not find marked results. And there are other disturbance factors like hunting (I’ve personally seen hunters very close to the reserve) that we have very little precise information about. Thus while it was possible to ask a theoretical question assuming that there were unknown changes in the environment in addition to an overall shift in water levels, asking a more management-oriented question would have required much more data on the pressures experienced by birds, in order to understand their responses. Also, the reserve is connected to other areas of Arcachon Bay which we should consider for management. In other words, there would be ample opportunities for more analysis from a management perspective but we should dig up more data!

About the second question: peer-review has been really, really long. These measures of compensation and synchrony are tricky to interpret and we had to convince referees that our results could not be artefacts. Some of the statistical tests had to be re-coded (one the statistical packages in R produced uncertainties measures that did not look right, and we re-coded them from scratch using “surrogate” time series). We did lots of simulations to make sure that our approach to analysing data was sound, even though this paper was initially meant to be just our first analysis of the dataset, paving the way for further work. This makes for a much more statistically-driven paper that we initially intended, which is great in some ways (check out our supplement with the simulations!). Also, we started in a journal where they switched editors at some point, which interrupted the peer-review process. The article was then forwarded to Ecology and Evolution where the peer-review process started anew, as new referees, even when forwarded previous comments, always have new questions. We have learned much about compensation and synchrony in that period, but Claude Feigné, who was the scientific manager of the reserve, had time to retire before this paper was published! This is a good illustration of the often conflicting timescales of scientific research vs conservation and management. Although flipping perspectives, one could also say that the data has been waiting since the 80s for such a time series analysis.

Summer 2022 update: some of the early data may need to be corrected for a subset of the community. We’ll look into this as early as possible.