Escape migration decisions in Eurasian Woodcocks: insights from survival analyses using large-scale recovery data

Guillaume Péron, Yves Ferrand, François Gossmann, Claudine Bastat, Michel Guénézan, Olivier Gimenez

ESM 1

Additional elements regarding meteorological data

Data were obtained from nine meteorological stations (Fig. S1). For each station we computed:

-The number of days with lower than zero average temperature (number of freezing days)

-The average temperature over the five months of winter (October – February)

-The minimum recorded temperature over the five months of winter

-The total amount of precipitation over the five months of winter

This produced 9  4 = 36 time series thatwere correlated to each other and were not representative of the range-wide variation in weather conditions. Principal Component Analyses (PCAs) produce linear combinations of the time series that are uncorrelated to each other. These combinations are termed principalcomponents and are ranked in order of decreasing percentage of the variance they capture. The time variation in the first principalcomponent of the overall (36 time series) PCA is displayedin Fig. S2.

We chose to perform PCAs on the location-specific time series, rather than simpler statistics like mean or median, because: (i) PCAs capture common variation in the time series even if this variation occur in opposite directions (e.g., cold weather might induce more rain in the Mediterranean zone, but less in the Continental zone). (ii) PCAs make it possible to disentangle sources of variation (e.g., altitude, latitude, climatic zone) by potentially producing different principalcomponents for each source of variation.

In order to determine which of the four weather variablesimpacted most on the demographic parameters of woodcocks, we performed four separate PCAs each involving 9 time series of the same weather variable. The first principalcomponents of these four PCAs were respectively noted N<0 for the number of freezing days, Θav for the average winter temperature, Θmin for the minimum recorded winter temperature, and Ptot for the total amount of precipitation. The correlations between these four weather indexes are givenin Table S1. In addition, we also examined whether these indexes varied linearly over time (steady improvement or deterioration of weather conditions; line labelled ‘T’ in Table S1).

Given these correlations, we chose to fit only capture-recapture models with the mostbiologically meaningful variables, that is N<0 which supposedly correlated to the availability of the main food source and the level of energy expenditure, and Ptot which supposedly correlated to the availability of the main food source. Since these variables did not exhibit any temporal trend, we also fitted capture-recapture models with a temporal trend in migration and survival probability.

Fig.S1: Locations of the nine Météo France stations where the meteorological data were collected.

Source: Google map.

Fig.S2: Temporal variation in the principal component of the PCA including all 36 time series. This componentexplained 45% of the variance in the meteorological data. Positive values indicate mild winters. Red, orange and green arrows indicate the occurrence of cold spells of respectively high, moderate and weak intensity.

Table S1: Squared correlation coefficients (upper triangle) and F-tests p-values (lower triangle) between weather indexes and between weatherindexes and time. Significant regressions are indicated in bold.

N<0 / Θav / Θmin / Ptot / T
N<0 / number of freezing days / 0.51 / 0.68 / 0.26 / 0.03
Θav / average temperature / <10-3 / 0.54 / 0.11 / 0.006
Θmin / minimum recorded temperature / <10-3 / <10-3 / 0.26 / 0.002
Ptot / total amount of precipitation / 0.02 / 0.15 / 0.02 / 0.02
T / linear temporal trend / 0.46 / 0.75 / 0.83 / 0.53

Fig. S3: Correlation circles for the two weather indexes we used in our CMR analysis. Left panel: number of freezing days. Right panel: total winter precipitation. Percentages indicate the fraction of variance in the station-specific data that was captured by the first (y-axis) and second (x-axis) components of the PCAs. Labels indicate the name of the Meteo France station. The arrows indicate the magnitude and direction of the correlations between station-specific data and the axes. The weather indexes N<0 and Ptot correspond to the y-axes of the left and right panels, respectively.