A tag is worth a thousand pictures: A framework for an empirically grounded typology of relational values through social media

Environmental values depend on social-ecological interactions and, in turn, influence the production of the underlying biophysical ecosystems. Understanding the nuanced nature of the values that humans ascribe to the environment is thus a key frontier for environmental science and planning. The development of many of these values depends on social-ecological interactions, such as outdoor recreation, landscape aesthetic appreciation or educational experiences with and within nature that can be articulated through the framework of cultural ecosystem services (CES). However, the non-material and intangible nature of CES has challenged previous at-tempts to assess the multiple and subjective values that people attach to them. In particular, this study focuses on assessing relational values ascribed to CES, here defined as values resonating with core principles of justice, reciprocity, care, and responsibility towards humans and more-than-humans. Building on emerging approaches for inferring relational CES values through social media (SM) images, this research explores the additional po-tential of a combined analysis of both the visual and textual content of SM data. To do so, we developed an inductive, empirically grounded coding protocol as well as a values typology that could be iteratively tested and verified by three different researchers to improve the consistency and replicability of the assessment. As a case study, we collected images and texts shared on the photo-sharing platform Flickr between 2004 and 2017 that were geotagged within the peri-urban park of Collserola, at the outskirts of Barcelona, Spain. Results reveal a wide spectrum of nine CES values within the park boundaries that show positive and negative correlations among each other, providing useful information for landscape planning and management. Moreover, the study high-lights the need for spatial, temporal and demographic analysis, as well as for supervised machine learning techniques to further leverage SM data into contextual and just decision-making and planning.