Towards Behavior Dependent Species Distribution Modeling
(8.-12. may 2017)Abstract
Movement is a crucial element in animal ecology. It offers important information on individual and collective behavior and providing the basis for efficient conservation planning. In order to fully understand movement one needs to consider the underlying environmental conditions that guide it. Its perception by the animal influences its decision process and, consequently, its movement. In this context, remote sensing offers a unique opportunity. It supports the monitoring of surface conditions providing a unique overview on natural and human driven processes on multiple spatial and temporal scales as well as on their impact on animal movement. However, connecting movement and remote sensing data is not an easy task. Movement occurs on a small temporal scales (e.g. minutes, hours) and varying spatial scales (e.g. meters to kilometers) making a direct assimilation of movement and remote sensing near impossible when considering the spatial and temporal constraints of both datasets. Additionally, while movement data provides information on the presence of an individual describing true absences (i.e. low suitability) for species distribution modeling is a challenge. Existing models such as MaxEnt attempt to tackle this issue by performing random background sampling. However, this assumes very specific environmental requirements from the animal and disregards the element of choice as well as the underlying complexity of the landscape. Within this presentation we discuss the impact of behavior dependent sample selection strategies from movement data on the accuracy of species distribution modeling. In this analysis we used White Stork (ciconia ciconia) tracking data from 13 individuals collected during the summer of 2013 within Germany.
