Cultural ecosystem services (CES) are an important benefit that habitats provide, particularly in the fragments of natural ecosystems that remain in urban areas. To manage CES we need to understand what people use habitats for, and where different activities take place. It is challenging to assess CES provision, as surveys and interviews are time consuming and can be expensive. Social media data, particularly geo-tagged photographs, are spatially explicit and contain visual information that can be used to infer cultural use. Indicators of CES derived from social media make use of existing data so may contribute useful information for rapid, cost-effective assessments of CES. In this study we develop an indicator of CES usage that is derived from photographs from an image-sharing website, at two different scales. First, we compare four small (<150ha) urban mangrove sites in Singapore, using photograph content to classify sites according to predominant cultural use. Second, the spatial distribution of different CES was modelled within one site using MaxEnt. A resampling simulation was conducted to identify the sensitivity of the photograph classification to the number of photographs classified. Photographs of social recreation, organisms and landscapes occurred most commonly. The proportional occurrence of photograph types differed between sites depending on their characteristics. Within one site, the probability of occurrence of social recreation photographs was highest around built focal points, while photographs of organisms were more likely in the mangrove and terrestrial habitats. Classifying more than 50–70 photographs (which would take approximately 30minutes) gave only small increases in categorisation accuracy. This tool for CES assessment rapidly provided information that would be useful for managing Singapore's mangroves. The approach could be widely applied to assess CES provision across a range of habitats and settings, helping CES to become more commonly considered in ecosystem service evaluations.
Read full abstract