Appendix A. Testing the "natal dispersal" model assumptions.
We checked the three assumptions of the "natal dispersal" model, i.e., (i) sex-biased natal dispersal, (ii) spatial variation in breeding habitat quality at the scale of natal dispersal, and (iii) temporal predictability of habitat quality (Julliard 2000).
We used natal dispersal data collected between 1997 and 2000 in the main study area, the Bremgartenwald, on our study species, the great tit, obtained from records of ringed nestlings that recruited within the local breeding population. Within the Bremgartenwald until year 2000, nest boxes were distributed continuously in space, thus allowing us to obtain unconstrained dispersal distances. The dispersal distance was computed as the straight distance (in m) between the natal nest box and the first breeding attempt nest box. We tested whether this distance differed between sexes, after accounting for natal conditions (year, hatching date, fledgling mass and body condition, and brood size at fledging). See Appendix B.
To investigate spatial and temporal variability in local breeding habitat quality, we used the main breeding characteristics as measures of habitat quality: nest box occupancy rate, clutch size, laying date, probability of breeding failure, fledgling number for successful nests, and fledgling body mass and condition (defined as the ratio of body mass over tarsus length at day 14). We used breeding data collected on great tits in the Bremgartenwald during the period 19932000 in these analyses. Within the Bremgartenwald, we defined square zones based on the mean natal dispersal distance of the most philopatric sex (i.e., approximately 450 m × 450 m; see below). We then checked spatial and temporal variation in breeding characteristics by (i) testing the influence of zone, year and their interaction (defined as random factors) on these values, and (ii) measuring their temporal autocorrelation. Nest box occupancy rate and probability of breeding failure were analysed with logistic regressions, laying date, clutch size and fledgling number and body condition with GLMs, using mixed models (proc mixed in SAS - Littell et al. 1996). Autocorrelation coefficients (Moran’s I) and significance (measuring temporal predictability) were obtained using the R software (Legendre and Vaudor 1991).
Spatial variation in breeding habitat quality: All measures of local breeding habitat quality varied in space at the scale of natal dispersal distances (see Appendix C), as reflected by significant effects of zone or zone × year interaction. More importantly, the zone × year interaction was significant for nearly all variables of reproductive success, showing that the relative quality of each zone varied through time at a large temporal scale (see below for year to year predictability), i.e. the same zones were not always the best or the worst quality zones. This implies that tits have to keep on sampling their environment and gathering information on where the best sites are at a given time.
Temporal predictability of breeding habitat quality: Occupation rate and clutch size were temporally autocorrelated (one-year time lag: occupation rate: n = 168, Moran’s I = 0.712, P < 0.001; clutch size: n = 162, Moran’s I = 0.284, P = 0.004 - see Appendix D). We found no temporal autocorrelation in fledgling number (n = 162, Moran’s I = -0.1296, P = 0.140), and mean fledgling body mass (n = 162, Moran’s I = -0.0540, P = 0.348) or condition (n = 160, Moran’s I = -0.0026, P = 0.530). However, these variables measure late breeding success, and nesting attempts in this population have been largely manipulated during these years (brood size manipulations, artificial nest parasite infestations, etc.), which is likely to have blurred any temporal autocorrelation. From results on early breeding traits, i.e. nest site choice and clutch size, we can conclude that the environment is at least partially predictable at the spatial scale of natal dispersal in the main study area, the Bremgartenwald.
Julliard, R. 2000. Sex-specific dispersal in spatially varying environments leads to habitat-dependent evolutionary stable sex ratios. Behavioral Ecology 11:421428.
Legendre, P., and A. Vaudor. 1991. Le progiciel R. Analyse multidimensionnelle, analyse spatiale. Université de Montréal, Montréal, Canada.
Littell, R. C., G. A. Milliken, W. W. Stroup, and R. D. Wolfinger. 1996. SAS System for Mixed Models. BBU Press, SAS Institute, Cary, North Carolina, USA.