SAR (Search and Rescue) operation is a complex process, often carried out in the absence of resources and time. Every single minute matters, as it puts the lost person in more danger. Therefore, it is really crucial to plan and coordinate SAR operation effectively. Because the search area is often very extensive, any leads about where to look first are invaluable. This can be achieved by modelling lost person's behaviour based on the data from past operations. Generated results present probabilities of finding a subject in different segments of the search area, which might benefit the planning and the execution phases of the operation. The authors evaluate one of the commonly used modelling methods and propose several ways to improve it, together with some preliminary evaluation results and an already implemented system, which incorporates the described methodology.