Ever increasing expansion of wireless and mobile technologies, electronic microprocessors and various communication tools have led to progress the Context-aware computing domain. Geospatial Information System (GIS) is one of the primary systems which utilize such technologies. The main challenge in context-aware GISs is the modeling of relevancy between the user and his/her related contexts. Space and time are two dominant factors to detect the relevant contexts and the other types of relevancies are dependent on spatial and temporal relevancies (spatio-temporal relevancy). Therefore, this research is concentrated on modeling spatio-temporal relevancy in context-aware GISs. It seems that there is no report on modeling all types of spatial relevancies (such as topological, metric and directional relationships) regarding the movement and cognitive characteristics of the moving user in urban context-aware GISs. Regarding this framework, the main contribution of the paper is that the proposed model applies customized Fuzzy Multi Interval Algebra and Voronoi-based Continuous Range Query to introduce spatio-temporal relevant contexts according to their arrangement in space based on the position, cognition, velocity and direction of the user in fuzzy approach. The customization process is undertaken based on the Comprehensive Calculus principles reducing 169 Allen’s relations to 25 spatio-temporal relations. In this research, the user is a tourist who is supposed to be guided from a hotel (his/her origin), to a defined destination. The context-aware system guides him/her according to spatio-temporal relevant contexts. The assessment of the implemented model is done regarding to three parameters including accuracy, time performance and users satisfaction. Tehran Districts 3, 6 and 11 are selected as the study area. In order to test the accuracy parameters, the designed software is run in three different routes with two different velocities and three time durations in 100 iterations. Then the accuracy of the detection of the related contexts is tested by means of a binomial distribution with one-sided 95% confidence level, precision and recall factors. The results of implementation and evolution demonstrated the efficiency of the proposed model in an urban context-aware wayfinding system.