Electronic Supplementary Material

“Breeding synchrony in colonial birds:

from local stress to global harmony”

Roger Jovani1,2* and Volker Grimm1

1UFZ, Helmholtz Centre for Environmental Research – UFZ

Department of Ecological Modelling

Permoserstr. 15

04318 Leipzig

Germany

2Department of Conservation Biology

Estación Biológica de Doñana, C.S.I.C.

Apdo. 1056,

E-41080 Sevilla

Spain

* Author for correspondence

Fax: 0049 341 3500

E-mail:

(A) Details of robustness tests

The effect of stochasticity in stress level dynamics

Background: In real colonies stresses that affect one breeding pair easily will affect also their neighbours. For instance, one or a group of floaters could produce a cascade of interactions in a particular zone of a colony affecting several nearby nests simultaneously. Here we have tested the effect of an even less favourable scenario regarding potential mechanisms counter-acting synchronization: we introduce stochasticity affecting the stress level dynamics of randomly chosen females. Examples of these stresses could be fights with floaters, sporadic fights with other non-neighbours, internal temporal physiological disequilibrium, uncertainty in the perception of the neighbour’s stress levels, or just individual variability in stress tolerance.

Methods: We introduced stochasticity in the daily update of the stress level in the following way:

(Eq. A1) OSLt+1 = [(1 - NR) * OSLt] + (NR * meanNSLt) – SD + noise

This stochastic stress affects a random subset of the females (p) each day and noise takes a random value between SD and 2*SD. Thus, when a female suffers stochastic stress the calming owed to day length enlargement (SD) vanishes and the female has a stress increase up to SD. Note that large values of p are unrealistic simply because reproduction is impossible if each day many females are too stressed to ever calm down for egg-laying. We have simulated stochastic stresses affecting up to one out of four females each day (p from 0 to 0.25), and we have run simulations at different NR values.

Results: Even low neighbour relevance (NR > 0.1) was found to be sufficient to counteract the desynchronising effect of local stochasticity even when this affected 25% of all individuals every day (Figure A1).

Note that for NR=0 the desynchronization of breeding was huge. Thus, since some level of stochasticity surely happens on the field, this result supports the idea that without some sort of local interactions between neighbours it would be difficult to synchronize breeding of the whole colony.

Figure A1. Breeding synchrony (=standard deviation of laying dates) depending on neighbour relevance, NR, and the daily probability of an individual to be affected by a stochastic effect (noise) on its stress level (OSL), (Eq. A1).

The effect of spatial heterogeneity

Background: The spatial patterns of breeding dates from the basic model (figure 3) may suggest that the spread of “synchrony waves” are necessary to synchronise the breeding of the whole colony. Thus, it could be assumed that the model could be sensitive to the existence of barriers of stress information between neighbours leading to the disruption of laying waves in the colony.

Methods: We introduced a variable number of empty nest-sites that were unavailable for nesting. This is a real situation in natural colonies: potential nest-sites may be not used in a given breeding season and bushes, big stones, or large slopes may even prevent the construction of nests in certain places within colonies. We used a random distribution of empty nest sites to simulate the most desynchronizing scenarios. This is because patches of empty nest sites would leave occupied nest sites unaffected. Linear chains of empty nest sites would simply create isolated subgroups (subcolonies). Subcolonies are known from field studies to be out of phase because there is not interaction between subcolonies, which are affected by some trigger factors independently.

Results: We found the model to be very robust even to high proportios of empty sites, e.g. 50% (Figure A2).

Figure A2. Effect of empty and unavailable nest sites (red) on synchronization; 50% of the nest sites were assumed to be empty. Top panels: neighbour relevance NR = 0; bottom panels: NR = 0.2. Compare this figure with figure 2 and 3 of the main paper.

Effect of dynamics of colony formation

Background: There are several bird species where the colony builds more or less at once, as we have implemented in the model. That is, although each individual could have different initial stress levels, all birds start interacting at the same time. However, there are many other species for which this is not the case. For instance, it is common that adult birds are the first on arriving to the colony and after the colony is formed there are still many other (mainly juvenile) birds that arrive at the colony even when the first birds are already breeding. This situation may potentially desynchronize the breeding of the whole colony because late arrivals may occur when other birds are already breeding.

Methods: We simulated an extreme case were arrival dates to the colony followed an exponential decay, thus creating a scenario where many females arrive at the colony when most of the birds in the colony have already laid. This was achieved by settling 10% of the remaining empty nests in the colony each day. Females just arrived to the colony occupied a random empty nest and a randomly chosen stress level between 100 and 200. We run these simulations with a different amount of birds being not in the colony at the beginning of the breeding season.

Results: Even with no individual present on day 0, the spatial and temporal effect of small NR values (NR = 0.2; figure A3, bottom panels) was huge compared to NR = 0 (figure A3, top panels). Note that breeding is delayed (and synchronized) for NR = 0.2 compared to NR= 0. This is because birds that are already tranquil and ready to breed interact with birds that arrived later and have a higher stress level, leading to an overall delay of breeding. Moreover, note that the symmetric normal-like shape of the histogram of laying dates shown in figure 2 is lost and a long tail appears, corresponding to those late arrivals to the colony.

Figure A3. Effect of colony formation (see text) on breeding synchrony. Top panels: neighbour relevance NR = 0; bottom panels: NR = 0.2. Compare this figure with figure 2 and 3 of the main paper.

Effect of colony size

Background: Literature on colonial breeding synchrony confers relevance to colony size because Darling (1938) proposed that birds would become stimulated (thus advancing and synchronizing laying) when breeding in huge colonies, but not in small ones. It can be thought that something similar could occur with the reciprocal stress modulation hypothesis proposed here. We have thus explored the effect of smaller colony sizes than the one used in the Basic Model.

Methods: Our Basic Model was run in a simulated colony of 10,000 females. Here, we simulated smaller colony sizes of 9, 100 and 1,024 females (10 and 1,000 can not be simulated in a square colony). We tested it with NR = 0.2. This can be tested in the RobustnessTests.nlogo model by changing the “Settings” in the model interface.

Results: Irrespective of colony size, breeding synchrony is much larger in colonies where individuals give some relevance to neighbours (NR = 0.2) than in colonies where individuals do not care about the stress level of their neighbours (NR = 0) (figure A4).

(A) NR=0 NR=0.2

(B)

(C) NR=0 NR=0.2

Figure A4. Effect of colony size on breeding synchrony for colony sizes of 9, 100 and 1,024 females in A, B and C, respectively. Left panels: no neighbour relevance (NR = 0); Right panels: NR = 0.2. I t can be easily appreciated that increasing NR from 0 to 0.2 increases breeding synchrony irrespective of colony size.

B) Source code

Two *.nlogo files are included as Electronic Supplementary Material. They can be easily run after freely downloading NetLogo 3.1.4 (Wilensky 1999) from http://ccl.northwestern.edu/netlogo/download.shtml (make sure you download version 3.1.4; the most recent version [January 2008] is 4.0.2).

The models have a friendly “Interface” where it is easy to change the values of the variables such as NR, SD, noise, etc. and see in real time their effect on individual stress level dynamics, laying date histograms and spatial patterns of laying synchrony. Moreover, under “Procedures” you can find the program code.

The “BasicModel” is the one that was used for the results shown in the main paper and “RobustnessTests” is the one used to test the robustness of the model, which results are shown in the Electronic Supplementary Material. We recommend to start with the “BasicModel” before exploring the “RobustnessTests”.

Copyright note: The two programs are free of use for science and education, but not for commercial purposes. If you use the programs for science, please make sure to refer to this article.