Characterizing the landscape of movement to identify critical wildlife habitat and corridors.

Landscape planning that ensures the ecological integrity of ecosystems is critical in the face of rapid human-driven habitat conversion and development pressure. Wildlife tracking data provide unique and valuable information on animal distribution and location-specific behaviors that can serve to increase the efficacy of such planning. Given the spatiotemporal complexity inherent to animal movements, the interaction between movement behavior and a location is often oversimplified in commonly applied analyses of tracking data. We analyzed GPS-tracking-derived metrics of intensity of use, structural properties (based on network theory), and properties of the movement path (speed and directionality) with machine learning to define homogeneous spatial movement types. We applied our approach to a long-term tracking data set of over 130 African elephants (Loxodonta africana) in an area under pressure from infrastructure development. We identified 5 unique location-specific movement categories displayed by elephants, generally defined as high, medium, and low use intensity, and 2 types of connectivity corridors associated with fast and slow movements. High-use and slow-movement corridors were associated with similar landscape characteristics associated with productive areas near water, whereas low-use and fast corridors were characterized by areas of low productivity farther from water. By combining information on intensity of use, properties of movement paths, and structural aspects of movement across the landscape, our approach provides an explicit definition of the functional role of areas for movement across the landscape that we term the movescape. This combined, high-resolution information regarding wildlife space use offers mechanistic information that can improve landscape planning.

Caracterización del Paisaje de Movimiento para Identificar Hábitats y Corredores de Fauna Importantes Resumen La planeación de paisajes que asegura la integridad ecológica de los ecosistemas es muy importante de cara a la rápida conversión de hábitats llevada por la acción humana y la presión del desarrollo. Los datos de rastreo de fauna proporcionan información única y valiosa sobre la distribución animal y el comportamiento específico por localidad que puede servir para incrementar la eficiencia de dicha planeación. Dada la complejidad espaciotemporal inherente al movimiento animal, la interacción entre la conducta de movimiento y la ubicación con frecuencia se ve sobre simplificada en los análisis de información de rastreos aplicados comúnmente. Analizamos las medidas derivadas de rastreos por GPS de la intensidad de uso, las propiedades estructurales (basadas en la teoría de redes) y las propiedades de la vía de movimiento (velocidad y direccionalidad) con aprendizaje automatizado para definir los tipos de movimiento espacial homogéneo. Aplicamos nuestra estrategia a un conjunto de datos de rastreo a largo plazo de más de 130 elefantes africanos (Loxodonta africana) en un área bajo presión ocasionada por el desarrollo de infraestructura. Identificamos cinco categorías de movimiento específico por localidad exhibidas por los elefantes, definidas en términos generales como intensidad de uso alta, media y baja. También identificamos dos tipos de corredores de conectividad asociados con movimientos rápidos y lentos. Los corredores de intensidad de uso alta y movimiento lento estuvieron asociados con las características similares de paisaje asociadas a las áreas productivas cercanas a cuerpos de agua, mientras que los corredores de intensidad baja y movimiento rápido estuvieron caracterizados por áreas de baja productividad alejadas de los cuerpos de agua. Con la combinación de la información sobre la intensidad de uso, las propiedades de las vías de movimiento y los aspectos estructurales del movimiento a lo largo del paisaje, nuestra estrategia proporciona una definición explícita del papel funcional que tienen las áreas de movimiento en el paisaje, la cual denominamos paisaje de movimiento (movescape). Esta información combinada y de alta resolución con respecto al uso espacial por la fauna ofrece información mecánica que puede mejorar la planeación del paisaje.

Applying network theory to animal movements to identify properties of landscape space use.

Network (graph) theory is a popular analytical framework to characterize the structure and dynamics among discrete objects and is particularly effective at identifying critical hubs and patterns of connectivity. The identification of such attributes is a fundamental objective of animal movement research, yet network theory has rarely been applied directly to animal relocation data. We develop an approach that allows the analysis of movement data using network theory by defining occupied pixels as nodes and connection among these pixels as edges. We first quantify node-level (local) metrics and graph-level (system) metrics on simulated movement trajectories to assess the ability of these metrics to pull out known properties in movement paths. We then apply our framework to empirical data from African elephants (Loxodonta africana), giant Galapagos tortoises (Chelonoidis spp.), and mule deer (Odocoileous hemionus). Our results indicate that certain node-level metrics, namely degree, weight, and betweenness, perform well in capturing local patterns of space use, such as the definition of core areas and paths used for inter-patch movement. These metrics were generally applicable across data sets, indicating their robustness to assumptions structuring analysis or strategies of movement. Other metrics capture local patterns effectively, but were sensitive to specified graph properties, indicating case specific applications. Our analysis indicates that graph-level metrics are unlikely to outperform other approaches for the categorization of general movement strategies (central place foraging, migration, nomadism). By identifying critical nodes, our approach provides a robust quantitative framework to identify local properties of space use that can be used to evaluate the effect of the loss of specific nodes on range wide connectivity. Our network approach is intuitive, and can be implemented across imperfectly sampled or large-scale data sets efficiently, providing a framework for conservationists to analyze movement data. Functions created for the analyses are available within the R package moveNT.