Page 26 - A bird’s-eye view of recreation - Rogier Pouwels

A bird's-eye view of recreation
2.3 Methods
Our method consists of six main steps (Fig. 2). First we collected information on visitor distribution using GPS devices. In this step the monitoring data from the GPS devices was prepared for further analyses. Second, we selected explanatory variables that describe the landscape and environment of the New Forest. In the third step we performed an exploratory data analysis to better understand the relationships between the different explanatory variables and characteristics of the routes visitors had followed. In the fourth step we developed a random forest model (Breiman 2001) to estimate the importance of the variables and their interaction in explaining the spatial variation in visitor densities. In the fifth step we used this model as a tool to predict visitor density distribution for the whole area. We illustrate the possible applications of the model by using it to assess the impact of three potential management interventions on the Nightjar population. In the sixth step we derived rules of thumb based on the results of the previous steps. The steps are explained in the next six sections.
2.3.1 Data collection and preparation
The monitoring data with GPS devices were collected on 80 mostly consecutive days during spring and summer in 2004 as part of the PROGRESS research project (Gallagher et al. 2007). Visitors arriving at car parks were asked to participate in the monitoring project. The GPS devices were stored in a plastic carrying case that could be clipped onto the rucksack or jacket of visitors who participated in the survey. Participants were instructed to keep the device with the built-in antenna upward and at an approximate height of 1.5m. Two models of GPS devices were used, the Garmin eTrex and Garmin eTrex Venture. The devices were set to collect single data points at a variable rate to create an optimum representation of the track. After participants returned, the single data points were stored in a database using the Garmin transfer protocol. Additional information regarding the group size, number of dogs and use of a leash for the dog was added to the monitoring data by unique ID.
In total 1563 GPS tracks with 110 505 single data points were collected at 41 car parks. The car parks were selected by the managers of the area. The monitoring frequency differed between car parks; some were monitored over 10 days, while others were only monitored once (Appendix 1). As the number of GPS devices was limited, the proportion of visitors monitored at highly used car parks was probably lower than at less used car parks. As these proportions are unknown the dataset does not represent
24