GIScience 2016 Short Paper Proceedings Context-sensitive spatiotemporal simulation model for movement S. Dodge Department of Geography, Environment, and Society, University of Minnesota Email: sdodge@umn.edu Abstract This paper presents a context-sensitive spatiotemporal model to simulate movement trajectories. The model incorporates both the correlated random walk and time-geography theories to generate a more realistic trajectory of an agent within its environment. 1. Introduction Movement is an essential form of temporal change that is an integral characteristic of dynamic entities (e.g. humans, animals, vehicles, diseases). It is the focus of research in a range of application domains such as transportation, movement ecology, environmental studies, and human health. Movement models help us to better understand the characteristics of movement, enable us to simulate movement and predict its patterns (Dodge 2016). Examples of existing movement models include the random walk and its variations (Codling et al. 2008, Technitis et al. 2015), time-geography (Miller 2005, Song and Miller 2014), and Brownian Bridge (Horne et al. 2007) models. These models either generate trajectories using a set of geometric movement parameters (turn angle, distance), or they identify a visitation probability surface for an agent considering its speed and time budget. Existing models often disregard the characteristics of the environment or the context within which the movement takes place. Simulation of movement in relation to its embedding context is an essential problem that is applied to generate trajectories to fill gaps in low-resolution tracking datasets, or to examine behavioral responses of moving agents to environmental changes. This paper introduces a context-sensitive spatiotemporal simulation model based on a correlated random walk with external biases and is controlled by time-geography constraints of the moving agent. The novelty of the model is that at each step the simulation is driven by behavior and the contextual factors (i.e. environment, geography) that influence the local movement of the agent. As a case study, this research uses GPS observations of a tiger to parameterize the model and to simulate the tiger’s movement between actual GPS observations. 2. Movement Simulation The overall goal is to generate a trajectory (a sequence of spatiotemporal points) from a start location and time !(# $ , & $ , ' $ ) to an end location and time )(# * , & * , ' * ). The simulation uses a correlated random walk from ! with an external bias to move towards ) (i.e. global constraints). The local movement at each step is driven by agent’s behavior and contextual factors. The model specifications are: (1) the maximum movement speed is determined by behavior (e.g. patrolling, hunting, foraging, biking), (2) the global movement path and speed are controlled by the actual time-budget to reach the end-point, and (3) the path is influenced by agent’s local choices based on context (environmental drivers and spatial constraints, -e.g. general movement direction, slope preferences, trail network). The simulation algorithm runs on regular time intervals defined by the user, named step time , to ensure the global movement occurs within the time-budget ('+ = ' * − ' $ ). The maximum speed of the agent (/ 012 ) is determined based on expert knowledge or derived