Page 145 - A bird’s-eye view of recreation - Rogier Pouwels

on a path network are linked to one route. Information from the track logs was used for the order of the data points (Fig. 1). For 10 tracks no routes could be constructed as they contained too few data points. At this stage of the analysis the resulting dataset contained 1553 routes.
Finally, in the fourth step a visual check of the constructed routes was conducted using QGIS. During the check small segments, or ‘dangling nodes’, of the routes were deleted (Fig. 1). These segments originated from snapping a single data point to the nearest path. At crossings this sometimes resulted in allocating the GPS data point to a path the visitor most likely would have crossed instead of followed. Only segments of paths were deleted when the snapped point was within 100 m, as the crow flies, of the main route a visitor had most likely followed. The set of 1553 routes was used to derive rules of thumb. For the random forest model only car parks with 10 or more routes in the database were taken into account, resulting in frequency maps for 36 car parks based on 1521 expected routes.
Appendices
143
A