62 Figure 17. Evolution of nest detection over time. A: Empirical data of detection of tracks according to their age for 3 seasons. B: Model of probability of detection based on logistic function. We fitted the ‘GLS’ 2017-2018 to 2022-2023 time series with Peak, LengthB, LengthE, Pmin and Max nesting phenology parameters using the R package phenology (Girondot 2023b). Flat parameter was not retained as it was fitted always close to 0. A Poisson distribution was preferred to a Negative Binomial since the mean number of nests per night is very low (Girondot 2017) and if NB was used, k fitted value was very large (>1000). We also tested different combinations of parameters to fit the more accurate phenological model and implement the possible seasonal for peak (P) and/or for LengthB and Length E (LBLE), based on AICc (Akaike 1974). The selected model has a different seasonal peak date but a similar nesting season length for all 6-time series (Tableau 4).
RkJQdWJsaXNoZXIy MjE1NDI=