Assessing forest recreational potential from social media data and remote sensing technologies data

Conventionally, forest management plans have focused on ensuring a continuous provision of wood. In recent years, political agendas worldwide have recognized the importance of forests' cultural ecosystem services, such as recreation. However, the inclusion of such values in management plans is challenging, and forest managers require novel methodologies and indicators to characterize forest recreation demand and provision. To this end, in this study, we combine remote sensing technologies and crowdsourced social media data to map and value the forest recreational potential of BC's provincial parks system. We trained and deployed convolutional neural networks to automatically classify the content of over 60,000 Flickr images, we then created a random forest model to identify the variables that influence the visitors' choice of recreational activity. These models allowed us to map the most likely recreational activities to occur in BC's provincial parks and perform a spatially explicit assessment of the consumer surplus that these activities generate with a benefit transfer approach. Our findings suggest that the most influential variables in determining the choice of forest recreational activities are topo-graphic, while anthropogenic impacts and forest biometrics variables have less effect. Furthermore, the outcomes of our study support the proposition that the integration of social media and remote sensing technologies allow, in the future, park managers to tailor the management of recreational services to forest visitors' needs.