Page 144 - A bird’s-eye view of recreation - Rogier Pouwels

A bird's-eye view of recreation
Appendix 2 Data handling of GPS tracks (Chapter 2)
The data points of each GPS track were used to construct the expected route of a visitor or visitor group in four steps. The four steps are illustrated for one track in Figure 1. This track was chosen as it illustrates all the potential problems we encountered in constructing routes from single data points. The first step in the preparation of the dataset was the removal of outliers and data points that are considered redundant for further analyses. Outliers are data points that are located at large distances from the rest of the data points on a specific track. We found two types of outliers: outliers caused by researchers switching the GPS device on and off before arriving at a car park without resetting the device, and outliers due to errors in the communication between the GPS device and satellites (Piedallu and Gégout 2005). A visual check revealed that for some tracks two consecutive data points were outliers. To select these consecutive errors we calculated the average distance to the three previous data points and to the three following data points. We used the rule that one of the average distances had to exceed 500 m and the other at least 250 m to be considered an outlier. The dataset also contains clusters of single data points at the start of a visit and at the end of a visit, due to the handling time between researchers and visitors, and at locations where visitors probably had a short stop. These clusters of data points contain many data points that may be considered redundant for determining the route followed. To decrease preparation time single data points within 5 m of one another were reduced to one data point for further analysis (Fig. 1). The removal of outliers and redundant points resulted in a 5% reduction in the number of single data points.
In the second step, single data points were assigned to the path network using the snapping method from the ArcGIS Toolbox (http://pro.arcgis.com/en/pro-app/tool- reference/editing/snap.htm). We used the snapping rule to assign data points to the nearest path within a distance of 50 m. Single data points that are further away from the path network were excluded for further analyses (Fig. 1). This preparation step resulted in a 1% reduction of the single data points.
The third step was the construction of the routes. Many tracks missed single data points for small parts of the route followed. To fill these gaps a travelling salesman route algorithm was used in QGIS Desktop (v2.14.12) with GRASS (v7.2.0) (https:// grass.osgeo.org/grass70/manuals/v.net.path.html). This algorithm constructs routes based on the order of data points. The shortest route between different data points
142