Climate change increases exposure to risks associated with late breeding in a tropical bird Supplementary material Rainfall and timing of breeding – preliminary analysis We wished to describe the relationship between rainfall and the timing of breeding (first egg laying date). Previous work on kestrels has shown that the frequency of rainfall in the July to September (spring) period is correlated with the timing of breeding – birds breed earlier in dryer springs (Nicoll 2004). In order to determine the specific period that influenced the timing of breeding within this three month time-window, we constructed a series of models that added different rainfall variables to a null model. This null model was defined using a linear mixed modelling framework (LMM), with female identity as a random effect to account for repeated observations from individual females breeding in different years. We fitted a global model to the timing of breeding data including individual and environmental variables other than rainfall (see Supplementary Table 1 for a list of candidate predictor variables), then simplified this global model using the update method in R (Crawley 2007). This produced a minimum adequate model (MAM, Model 1), which was then used as the null model for exploring the effect of rainfall on the timing of breeding. We explored two types of rainfall variable – total rainfall (mm) and the number of rain-days. To determine the important rainfall variable(s) and time period related to the timing of breeding, we set up a series of four global models, including the null model described above (Model 1), and three models differing in their time windows: Model 2 analysing the complete three month period (July to September), Model 3 analysing two month periods (July to August, August to September, mid-July to mid -September) and Model 4 analysing one month periods (July, August, September, mid-July to mid-August and mid-August to mid-September). Each model was then simplified using the step AIC/update function in R (Crawley 2007), and the AIC of the final MAMs compared. The results show that timing of breeding is closely related to the number of rain-days in August (Supplementary Table 2) – birds breed earlier when it is dryer in August. Supplementary Table 1: Definition of explanatory variables included in the generalised linear models Variable name Variable type Definition Female age Categorical Age of female* Male age Categorical Age of male partner* Pair PBE Categorical Previous breeding experience of the pair Pair PBEAS Categorical Previous breeding experience of pair at particular territory Season Categorical Year / season of breeding attempt Territory Categorical Breeding territory Density Continuous Total population density in the season of breeding * Age split into three groups - Birds aged 1 year old; birds aged 2-7 years and birds aged 8 or more years (Categorisation based on previous exploratory analyses) Supplementary Table 2: Linear mixed models run to determine significant rainfall variable affecting the timing of breeding. All models had female identity as a random effect and selection between models involving different time windows was done using AIC, first within global models and then across the minimum adequate models. The final selected model is given in bold. N = 243 breeding attempts (first clutches of birds breeding in two or more years only) Model 1 Fixed effects tested in global models Male age group & female age group Fixed effects in minimum adequate model Parameter estimates (±SE) P value Intercept 56.60 ± 2.60 <0.0001 Male age (2-7 years old) -16.48 ± 2.64 <0.0001 Male age ( ≥ 8 years) -10.76 ± 4.08 0.0092 Female age (2-7 -6.07 ± 2.88 0.0371 Log Likelihood AIC AICc ΔAIC AIC weights -809.76 1633.49 1634.07 5.78 0.04 years old) 2 3 Male age group, female age group, July to September rainfall & July to September rain days Male age group, female age group, July to August rainfall, July to August rain days, August to September rainfall, August to September rain days, mid July to mid September rainfall & mid July to mid Female age ( ≥ 8 years) -5.08 ± 4.23 0.2327 Intercept 45.64 ± 7.45 <0.0001 July – September rain days 0.18 ± 0.12 0.0121 Male age (2-7 years old) -16.31 ± 2.63 <0.0001 Male age ( ≥ 8 years) -10.34 ± 4.08 0.0123 Female age (2-7 years old) -6.88 ± 2.92 0.0199 Female age ( ≥ 8 years) -7.26 ± 4.48 0.1071 Intercept 41.03 ± 7.51 <0.0001 July – August rain days 0.36 ± 0.16 0.0293 Male age (2-7 years old) -16.36 ± 2.61 <0.0001 Male age ( ≥ 8 -10.65 ± 4.04 0.0095 -808.6 1633.2 1633.93 5.49 0.05 -807.42 1630.81 1631.58 3.1 0.16 September rain days 4 Male age group, female age group, July rainfall, July rain days, August rainfall, August rain days, September rainfall, September rain days, mid July to mid August rainfall, mid July to mid August rain days, mid August to mid September rainfall & mid August to mid September rain days years) Female age (2-7 years old) -7.04 ± 2.88 0.016 Female age ( ≥ 8 years) -7.30 ± 4.33 0.094 Intercept 44.56 ± 4.97 <0.0001 August rain days 0.61 ± 0.22 0.0055 Male age (2-7 years old) -16.37 ± 2.59 <0.0001 Male age ( ≥ 8 years) -10.90 ± 4.02 0.0075 Female age (2-7 years old) -7.11 ± 2.85 0.0138 Female age ( ≥ 8 years) -7.12 ± 4.23 0.0949 -805.86 1627.71 1628.45 0 0.75 References cited in supplementary material Crawley, M. J. 2007 The R Book. Chichester: John Wiley & Sons Ltd. Nicoll, M. A. C. 2004 The ecology and management of a re-introduced population of the Mauritius kestrel (Falco punctatus). In The ecology and management of a re-introduced population of the Mauritius kestrel (Falco punctatus). Unpublished PhD thesis., vol. PhD: University of Reading.