"Modeling Breeding Bird surveys"Erard, AdélieAdélie Erard (MAP5, Université Paris Cité), Raphaël Lachièze-Rey (DYOGENE, INRIA Paris ; MAP5, Université Paris Cité), Romain Lorrillière (CESCO, Muséum national d'Histoire naturelle) A breeding bird survey (BBS) is a systematic method used to monitor and assess bird populations during their breeding season. The primary goal of breeding bird surveys is to gather data on the abundance, distribution, and breeding success of bird species within specific habitats. Moreover, as birds are good indicators of ecosystem health and biodiversity, monitoring breeding bird populations helps assess the overall health and diversity of habitats. Changes in bird populations may reflect broader environmental changes, such as habitat loss, degradation, or climate change. The French BBS database gives us information on how many birds of each species were seen each year at each location and a precise description of the habitat of this location. To extend environment's descriptions we added data on climate, pesticide use and land use. Our goal is to predict in which conditions bird populations disappear and what are the explanatory variable. We thus construct a model explaining changes in the dynamic of bird populations. The first thing we do is to use random forest algorithm on our database to find which environmental variable are influent on bird populations dynamics. We found that most of changes in abundances are explained by environmental changes. We will thus suppose that changes in abundance are only due to environmental condition and will model it by a Cox process. Briefly, Cox processes are Poisson process but with a random intensity modeling changes in environmental conditions. We then want to find intensity of the process in each environment and estimate the intensity process. |
« back