This paper presents a geographic information system (GIS)-based methodology for the automated planning of walking routes in areas with sparse road networks. This approach overcomes the limitations of traditional routing methods that rely solely on either vector road graphs or raster-based cost surfaces by combining both types of representations into a single flexible model. This combined model supports routing on both roads and off-road terrain, taking into account land cover, slope, and movement direction. A hierarchy of network edges prioritizes more passable segments, such as paved roads, while still allowing transitions to natural surfaces where necessary. The methodology has been implemented in ArcGIS Pro and tested using field data from the Satino educational station in the Kaluga region. Three types of regular grid-based network models (square with queen and knight adjacency, and hexagonal) were compared, and routing was simulated under different weather conditions using variable edge weights. The results demonstrate that the square grid with queen adjacency provides the best balance between accuracy and computational efficiency. Compared to the traditional least-cost path (LCP) method, the proposed approach reduces raster-induced distortions and allows for greater adaptability to user preferences. This methodology has practical applications in tourism, environmental research, and emergency route planning, particularly in regions with limited formal transportation infrastructure.