Inter-habitat migration can have important effects on food-web structure and equilibrium dynamics. For example, migration is essential for the success of conservation biological control when predators must move from refuge to crop habitats. However, the study of such movements can be hampered by difficulty in observing individual organisms when they are small or move through dense vegetation or other complex-structured habitats. In this study we use a combination of density measures and simulation modeling to confirm expected inter-habitat movements of the wolf spider Pardosa littoralis in a two-habitat, cordgrass system. This spider is a common predator on mid-Atlantic marshes where it moves from its winter refuge on the upland marsh (Spartina patens=SP) to its developmental habitat at lower elevations (Spartina alterniflora=SA). Both cordgrass habitats have litter layers that preclude direct observation. Following severe winters, a gradient of declining Pardosa density develops with increasing distance from SP in SA. We evaluated several possible mechanisms underlying the density gradient of spiders, including differential survivorship, random diffusion, ballooning, and density-dependent movement. Individual-based computer models based on each of these mechanisms were tested against observed density measurements. Tests confirmed that only movement-based models provided a satisfactory fit to field data, with the best fit resulting from a combination of ballooning and conspecific avoidance. Thus, simulation modeling confirms the inter-habitat migration of this important intraguild predator, and provides a potentially important tool for assessing the inter-habitat movements of predators and other organisms in natural and managed systems.