ESM for the article «Age-specific survival and annual variation in survival offemalechamois differ between populations»
Supplementary Data Description
Table S1. Description of the data and the study sites in France (Bauges) and in Switzerland (Swiss National Park, SNP).
Table S2. Sample sizes.
Supplementary Methods
Table S3. Results of models dealing with trap-happiness.
Table S4. Results of models dealing with resighting probabilities (p).
Supplementary Results
Table S5. Estimates of some of the models presented in Table 2. The estimates are back-transformed unless otherwise noted.
Table S6. Model selection to describe the effect of site (Val Trupchun and Il Fuorn) on age-specific survival of female chamois in the Swiss National Park (SNP).
Figure S1.Age-specificsurvivaloffemalechamoisinthetwopopulationsoftheSwissNationalPark(Val Trupchunand Il Fuorn).
Figure S2. Age- and year-specific survival of female chamois in the Bauges population.
Figure S3. Age- and year-specific survival of female chamois in the Swiss National Park (SNP) population.
Supplementary References
Supplementary Data Description
Table S1. Description of the data and the study sites in France (Bauges) and in Switzerland (Swiss National Park, SNP)
Bauges (Armenaz-Pécloz) / SNP (Val Trupchun) / SNP (Il Fuorn)Characteristics of the sites
Altitude / 1400 to 2200 m / 1750 to 3000 m
Climateⱡ / Temperature = 9.22 ± 0.46°C *
Rainfall = 1361.8 ± 210.6 mm * / Temperature = 0.98 ± 0.48°C **
Rainfall = 882.5 mm ± 213.2 mm ***
Rock / Calcareous / Calcareous / Dolomite
Other wild ungulates/predators / - / Red deer, ibex, eagle for kids / Red deer, ibex, eagle for kids
Characteristics of the chamois
# of females captured / 238 (1991-2012) / 40 (1996-2012) / 89 (1995-2012)
# of observations / 1044
(39 censored) / 169
(0 censored) / 534
(0 censored)
Mean body mass † / 27.3 kg / 26.9 kg / 23.8 kg
Mean ± SE horn growth
between 0 and 3 years † / 160.7 ± 14.0 mm / 164.7 ± 4.5 mm / 155.3 ± 2.4 mm
Age range at capture / 1-12 / 0-13 / 0-23
Last age from observations / 20 / 19 / 26
ⱡ Annual mean temperature and cumulative amount of rainfall ± SD (period 1996-2012)
*Lescheraines weather station (590 m a.s.l., 45°42'30"N, 06°06’06’’E),
** Buffalora weather station (1968 m a.s.l., 46°39'00.0"N 10°16'00.0"E)
*** Punt la Drossa weather station (1710 m a.s.l., 46°39'00.0"N 10°11'00.0"E)
† Measures at first capture (we excluded individuals younger than 3 years old) published elsewhere for the Bauges population (Bleu et al. 2014) and not published for the SNP populations. The measure of body mass presented here is the predicted body mass at 1st of June (model “log(body mass) ~ date of capture” - the model was run separately for each population).
Table S2. Sample sizes (number of females)
Age at capture / 0 / 1 / 2 / 3 / 4 / 5 / 6 / 7 / 8 / 9 / 10 / 11 / 12 / 13 / 14 / 15 / 16 / 17 / 23Les Bauges (Armènaz)* / 0 / 54 / 16 / 27 / 38 / 25 / 17 / 18 / 15 / 12 / 8 / 2 / 6 / 0 / 0 / 0 / 0 / 0 / 0
SNP (FUO + TRU) / 10 / 13 / 3 / 8 / 12 / 9 / 10 / 10 / 7 / 11 / 6 / 7 / 5 / 8 / 3 / 1 / 4 / 1 / 1
SNP (Il Fuorn) / 4 / 10 / 1 / 5 / 11 / 6 / 3 / 8 / 4 / 7 / 5 / 7 / 5 / 4 / 2 / 1 / 4 / 1 / 1
SNP (Val Trupchun) / 6 / 3 / 2 / 3 / 1 / 3 / 7 / 2 / 3 / 4 / 1 / 0 / 0 / 4 / 1 / 0 / 0 / 0 / 0
Age at capture/observations / 0 / 1 / 2 / 3 / 4 / 5 / 6 / 7 / 8 / 9 / 10 / 11 / 12 / 13 / 14 / 15 / 16 / 17 / 18 / 19 / 20 / 21 / 22 / 23 / 24 / 26
Les Bauges (Armènaz) / 0 / 54 / 57 / 83 / 104 / 106 / 101 / 97 / 99 / 79 / 78 / 57 / 50 / 34 / 21 / 10 / 7 / 2 / 2 / 1 / 2 / 0 / 0 / 0 / 0 / 0
SNP (FUO + TRU) / 10 / 19 / 16 / 20 / 30 / 36 / 43 / 46 / 47 / 57 / 56 / 51 / 52 / 56 / 47 / 31 / 28 / 20 / 14 / 11 / 7 / 2 / 1 / 1 / 1 / 1
SNP (Il Fuorn) / 4 / 11 / 9 / 12 / 23 / 27 / 28 / 34 / 34 / 40 / 40 / 38 / 48 / 41 / 38 / 25 / 25 / 20 / 14 / 10 / 7 / 2 / 1 / 1 / 1 / 1
SNP (Val Trupchun) / 6 / 8 / 7 / 8 / 7 / 9 / 15 / 12 / 13 / 17 / 16 / 13 / 4 / 15 / 9 / 6 / 3 / 0 / 0 / 1 / 0 / 0 / 0 / 0 / 0 / 0
*For the Bauges population, some animals died from non-natural causes (hunting: n = 27, accidents at capture: n = 2) or were marked with temporary collars (GPS collars with drop-off device for other studies, n = 10). These animals were right-censored in the analyses. The GWR is a protected area where only guided hunt occurs. We are therefore confident that all marked animals that were shot were reported.
Supplementary Methods
Using the program U-care (version 2.3.2) we tested the goodness-of-fit of the Cormack-Jolly-Seber model to the data (Choquet et al. 2009). It satisfactorily fitted the data (Bauges: chi2 = 69.23, P = 0.305, SNP: chi2 = 88.02, P = 0.361). There was no transience (Bauges: Z = -1.76, P = 0.961, SNP: Z = -0.02, P = 0.508), but a strong trap-happiness (Bauges: Z = -5.85, P < 0.001, SNP: Z = -6.05, P < 0.001). This means that successive resighting events are positively correlated [see examples of trap-dependence latosensu, i.e. when animals are not physically recaptured, in (Pradel and Sanz-Aguilar 2012)].
To deal with the trap-happiness effect, we compared two types of models: (1) models describing a change of state of the animals after observation (immediate trap-dependence), or (2) models describing an intrinsic individual heterogeneity in behaviour (mixture models)(Pradel 2009, Cubaynes et al. 2010). In both cases, the resighting probability depends on the state of the animal. For the immediate trap-dependence model the states are: “seen the previous year” and “unseen the previous year”. For the mixture models, the states are: “animals with a high resighting probability” and “animals with a low resighting probability” and they correspond to hidden states (they are not explicit in the dataset). Mixture models can allow transitions between the states or not depending whether the individual heterogeneity is considered to be a fixed individual characteristic. If transitions are allowed, we can estimate the probability to change from one state to the other state (called the transition probability). The probability to be in a given state (here, in the high or low resighting probability states) at first capture is called initial state probability and was considered constant.
The best model to account for this trap-happiness effect was a two-state mixture model with transitions between the states for both the Bauges(as in Bleu et al. 2014) and SNP datasets (see details in Table S3). This implies that animals are separated in two groups (high and low re-sighting probabilities groups) and that animals can change groups (e.g. some animals temporarily changing their home range).
After selection of the models to deal with trap-happiness, we investigated the effect of time on p (re-sighting probability). In general, p varies with years but a full time-dependent model requires numerous parameters. It may be possible to model this effect of the year of observation with less parameters if we model it with the appropriate covariate. We checked whether the effort of observation, defined as log(number of different days with at least one observation of a marked animal), can describe accurately this year effect. Thus, the effort of observation was added as a linear covariate in the model instead of the year. Post-hoc observations revealed that this covariate was not a good predictor of p for some years (p is higher than expected given the effort of observation). Thus, we also tested models with some years that are not dependent on the effort of observation (see the best models in Table S4).
Table S3. Results of models dealing with trap-happiness. The models used in the subsequent analyses are in bold.
k / Deviance / AICcA. Female chamois in Les Bauges
Mixture model that allows transitions between the states (two probabilities depending on the way of the transition) / 30 / 1623.32 / 1685.09
Mixture model that allows transitions between the states (one transition probability) / 29 / 1631.23 / 1690.88
Mixture model / 28 / 1640.48 / 1698.03
Immediate trap-dependence / 27 / 1647.96 / 1703.39
B. Female chamois in the Swiss National Park (SNP)
Mixture model that allows transitions between the states (two probabilities depending on the way of the transition) / 27 / 987.05 / 1043.29
Mixture model that allows transitions between the states (one transition probability) / 26 / 998.40 / 1052.48
Mixture model / 25 / 1039.99 / 1091.91
Immediate trap-dependence / 24 / 1044.48 / 1094.25
Table S4. Results of models dealing with resighting probabilities (p).The models used in the subsequent analyses are in bold.
k / Deviance / AICcA. Chamois in les Bauges
Effect of observation effort except for years 1992-1996, years 1997-1998 and years 2011-2012 / 13 / 1645.39 / 1671.73
Effect of time (year) / 30 / 1623.32 / 1685.09
Effect of observation effort except for years 1992-1996 and years 1997-1998 / 12 / 1663.50 / 1687.79
Effect of observation effort except for years 1992-1996 / 11 / 1674.30 / 1696.55
Effect of observation effort / 10 / 1677.26 / 1697.47
Constant p / 9 / 1742.14 / 1760.31
B. Chamois in SNP
Effect of observation effort except for year 1996 and year 2002 / 13 / 1008.44 / 1034.97
Effect of observation effort / 11 / 1019.78 / 1042.17
Effect of observation effort except for year 1996 / 12 / 1017.84 / 1042.30
Effect of time (year) / 27 / 987.05 / 1043.29
Constant p / 10 / 1054.25 / 1074.57
Supplementary Results
Table S5. Estimates of survival of some of the models presented in Table 1. The survival estimates are back-transformed unless otherwise noted.
A. Bauges population / Age-class / EstimatesModel M4 / 1-7 / 0.96 (CI = 0.94, 0.98)
8-12 / 0.87 (CI = 0.83, 0.90)
>12 / 0.71 (CI = 0.62, 0.79)
Model M5 / 1 / 0.97 (CI = 0.80, 1.00)
2-7 / 0.96 (CI = 0.94, 0.98)
8-12 / 0.87 (CI = 0.83, 0.90)
>12 / 0.71 (CI = 0.62, 0.79)
Model M12 / 1 / 0.98 (CI = 0.97, 0.99)
1 (logit scale) / 3.88 (CI= 3.32, 4.44)
Senescence rate / -0.21 (CI = -0.27, -0.16)
Model M14 / 1 / 0.96 (CI = 0.81, 0.99)
2 / 0.98 (CI = 0.96, 0.99)
2 (logit scale) / 3.71 (CI = 3.17, 4.24)
Senescence rate / -0.22 (CI = -0.28, -0.16)
B. SNP population / Age-class / Estimates
Model M5 / 0-1 / 0.75 (CI = 0.56, 0.88)
2-7 / 0.97 (CI = 0.91, 0.99)
8-12 / 0.98 (CI = 0.90, 1.00)
>12 / 0.80 (CI = 0.74, 0.85)
Model M6 / 0 / 0.73 (CI = 0.37, 0.93)
1 / 0.76 (CI = 0.52, 0.90)
2-7 / 0.97 (CI = 0.91, 0.99)
8-12 / 0.98 (CI = 0.90, 1.00)
>12 / 0.80 (CI = 0.74, 0.85)
Model M9 / 0-1 / 0.75 (CI = 0.56, 0.88)
2-12 / 0.97 (CI = 0.93, 0.99)
>12 / 0.80 (CI = 0.74, 0.85)
Model M20 / 0-1 / 0.75 (CI = 0.56, 0.88)
2-12 / 0.97 (CI = 0.93, 0.99)
13 / 0.84 (CI = 0.76, 0.90)
13 (logit scale) / 1.68 (CI = 1.13, 2.24)
Senescence rate / -0.09 (CI = -0.22, 0.04)
Model M14 / 0-1 / 0.75 (CI = 0.56, 0.88)
2 / 0.99 (CI = 0.96, 1.00)
2 (logit scale) / 4.40 (CI = 3.30, 5.49)
Senescence rate / -0.19 (CI = -0.28, -0.12)
Table S6. Model selection to describe the effect of site (Val Trupchun and Il Fuorn) on age-specific survival of female chamois in the Swiss National Park (SNP).
The effect of site was tested on the best model (lowest AICc) describing age-specific survival (see model M9 in Table 1). This best model has 3 age-classes: 0-1, 2-12 and >12 years old. We proceeded in 2 steps: first we tested the effect of site on the resighting (p), transition and initial state probabilities; second we tested the effect of site on the survival probabilities from the best model selected in the first step (model in bold). In the first step, we modelled the effect of site as an additive effect. The number of parameters of each model is k. The models used in the subsequent analyses are in bold.
Model / k / Deviance / AICcStep 1
Effect of site on transition / 12 / 1005.94 / 1030.40
No effect of site (model M9 in Table 1) / 11 / 1008.59 / 1030.97
Effect of site on initial state / 12 / 1007.75 / 1032.21
Effect of site on p and transition / 13 / 1005.68 / 1032.21
Effect of site on p / 12 / 1007.80 / 1032.26
Step 2
Same effect of site on all age-classes / 12 / 1005.31 / 1029.76
Effect of site on age-class 2-12 / 12 / 1005.69 / 1030.15
Effect of site on age-class >12 / 12 / 1005.89 / 1030.35
No effect of site (model M9 in Table 1) / 11 / 1008.59 / 1030.97
Different effect of site on each age-class / 14 / 1003.07 / 1031.68
Effect of site on age-class 0-1 / 12 / 1008.31 / 1032.76
Figure S1.Age-specificsurvivaloffemalechamoisinthetwopopulationsoftheSwissNationalPark(Val TrupchunandIlFuorn).Theestimatesofthemodelwithanadditiveeffectofsiteareshown (i.e.,theeffect of siteisthesameforeachageclassonthelogitscale). This model is presented in Table S6 (AICc = 1029.76).
Figure S2. Age- and year-specific survival of female chamois in the Bauges population.This figure represents the estimates from the Bauges population of a model with an effect of year and site on survival.
Figure S3.Age- and year-specific survival of female chamois in the Swiss National Park (SNP) population. This figure represents the estimates from the SNP populations of a model with an effect of year and site on survival
Supplementary References
Bleu, J., A. Loison, and C. Toïgo. 2014. Is there a trade-off between horn growth and survival in adult female chamois? Biological Journal of the Linnean Society 113:516–521.
Choquet, R., J.-D. Lebreton, O. Gimenez, A.-M. Reboulet, and R. Pradel. 2009. U-CARE: Utilities for performing goodness of fit tests and manipulating CApture-REcapture data. Ecography 32:1071–1074.
Cubaynes, S., R. Pradel, R. Choquet, C. Duchamp, J.-M. Gaillard, J.-D. Lebreton, E. Marboutin, C. Miquel, A.-M. Reboulet, C. Poillot, P. Taberlet, and O. Gimenez. 2010. Importance of accounting for detection heterogeneity when estimating abundance: the case of French wolves. Conservation Biology 24:621–626.
Pradel, R. 2009. The stakes of capture-recapture models with state uncertainty. Pages 781–795 in D. L. Thomson, E. G. Cooch, and M. J. Conroy, editors. Modeling Demographic Processes In Marked Populations. Springer US.
Pradel, R., and A. Sanz-Aguilar. 2012. Modeling trap-awareness and related phenomena in capture-recapture studies. PLoS ONE 7:e32666.