Lasso penalisation identifies consistent trends over time in landscape and climate factors influencing the wintering distribution of the Eurasian Curlew (Numenius arquata)

Abstract

Migratory birds are particularly susceptible to climate change and habitat loss due to their reliance on a global network of ecosystems, with waders and seabirds undergoing significant population decline. The Eurasian Curlew (Numenius arquata) is classified as Near Threatened on the IUCN Red List of Threatened Species, but breeding populations in Great Britain and the island of Ireland have declined drastically, with the species on the brink of local extinction. We present a set of models of the distribution of Curlew sightings between November and February in the Great Britain and the island of Ireland over the period 2003 to 2019. Using a model selection process (cross-validated lasso regression), we reduce the fairly large set of CORINE satellite land cover classes to a much smaller set of explanatory variables which we combine with environmental variables and fit binomial Generalized Linear Models to Curlew observation records. This enables us to build up a detailed picture of where and when Curlew are sighted between November and February over the 17 years of the study. Reproducibly, from November to January between 2003 and 2019, the coastal land cover classes, Estuaries, Intertidal Flats, Salt Marshes and Port Areas, feature prominently in the sets of explanatory variables selected by the lasso regression. Moreover, this study represents the first regional scale analysis on the impact of landscape and climate features on wintering curlew distribution, identifying the importance of landscape factors that warrant further research, such as the importance of artificial structures and the importance of February within the migration of the Curlew.