Revealing the differences of urban parks’ services to human wellbeing based upon social media data

Abstract Urban green spaces provide a range of services to urban residents; however, how these distinct spaces provide different services, if any, to human wellbeing has been seldom addressed. Using large volumes of social media data considering broad user groups and publicly available content, we developed a method to transform unstructured online comments into a structured assessment of nine categories of ecosystem services. An ‘ecosystem services lexicon’ was created based on 6853 words under twenty-eight subcategories of parks' services to human wellbeing using the word2vec model. The application of the ecosystem service lexicon to urban parks in Beijing revealed that all ecosystem services were perceivable to urban park users; however, the perception frequency of the ecosystem services varied across different parks. Additionally, the perceived ecosystem services were bundled together; four types of bundles were identified with varied dominant services. Technically, this study offered a novel technical procedure that can transfer unstructured free comments into a structured assessment of urban parks' perceived services to human wellbeing. Theoretically, the study revealed small-scale ecosystem service bundles from users' opinions and called for a further cross-scale understanding of ecosystem service bundles. Practically, the study findings can help inform evidence-based park policies, planning, and management.