Background/Question/Methods Geospatial movements of fish have major implications for fisheries stock assessment, identification and conservation of essential fish habitat, and spatial management measures including marine protected areas. The development of realistic fish movement models has thus become a key area of research in fisheries ecology. Fish may display a range of scale-dependent directed and undirected movement behaviours that are influenced by internal physiological states and external triggers. Modelling options range from simple random walk and advection models to complex, mechanistic models with multiple state-dependent behaviours. While the conceptual and theoretical basis of movement ecology is developing rapidly, movement data on wild fish in natural environments remain limited. Active tracking data are limited in spatial extent and time recorded. Passive tracking data, whilst able to cover a larger area, record only occasional spot samples of a path except in very dense telemetry arrays. In either case, tracking designs may be insufficient to resolve movements of interest. The question thus arises how best to model movement given limited telemetry data.

Results/Conclusions We use a simulation approach to evaluate the precision and bias of movement parameter estimation for statistical and biologically mechanistic models of different complexity, for alternative telemetry designs. Use of complex models in the analysis of telemetry data of low to moderate spatial and temporal cover and resolution gives rise to low precision and in some cases substantial bias in movement parameter estimates. Relatively simple statistical models outperform complex mechanistic models in terms of predictive capacity even for movement paths generated by complex underlying models. We derive quantitative design criteria that can be used to match movement model complexity and telemetry configurations for a commonly used ultrasonic tracking method.