Research Note: Linking sensory perceptions with landscape elements through a combined approach based on prior knowledge and machine learning

Abstract

Challenges remain when assessing cultural ecosystem services (CES) due to the subjective and intangible features. Landsense ecology highlights the individual perception, providing a novel dimension to the CES assessment. Based on landsense ecology, this study aims to identify and correlate landscape elements and sensory perceptions using social media data. Taking the case of Belgian greenspaces, the factors were identified using custom lexicons through unsupervised and weakly supervised learning methods. The perception correlation network was established to link sensory perceptions with landscape elements. Integrating data from different platforms, we found that social elements were perceived frequently and that the most significant perceptions were vision and touch. Results also revealed that the perceptions of vision, hearing and smell were collectively affected by multiple elements, while the perceptions of taste and touch were strongly related to specific elements. Finally, this study encouraged creating multi-sensory spaces in greenspace management.