Integrating social media data and machine learning to analyse scenarios of landscape appreciation

• Published in: Ecosystem services Vol. 55; p. 101422
• Main Authors: Richards, Daniel Rex, Lavorel, Sandra
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01-06-2022; Elsevier
• Subjects: Ecosystem services; Environmental Sciences; Native forest; Restoration; Sensitivity analysis; Spatial mismatch; Ecosystem services; Spatial mismatch; Sensitivity analysis; Native forest; Restoration
• ISSN: 2212-0416; 2212-0416
• DOI: 10.1016/j.ecoser.2022.101422

•Machine learning to model the occurrence of landscape appreciation photographs.•Landscape appreciation model used in sensitivity analysis.•Spatial trade-offs between native forest restoration priorities for carbon and cultural service.
Cultural ecosystem services can be challenging to simulate, leading to their under-representation in future scenario modelling to support decision-making. Here we use the density of landscape appreciation photographs uploaded to social media to parameterise an empirical model of landscape appreciation. We developed the model using over 150,000 photographs uploaded to the website Flickr in Aotearoa New Zealand. The current distribution of landscape appreciation photographs was influenced by a combination of biophysical and socio-ecological factors, including the land cover, altitude and distance from the coastline, and human population densities and accessibility. We used the landscape appreciation model to conduct a sensitivity analysis of the impacts of native forest restoration on agricultural land. The sensitivity analysis identified priority areas where restoration would have a larger positive impact. By comparing with a model of carbon storage gained through native forestation, we highlighted substantial spatial mismatches between these conflicting ecosystem service objectives. Empirical models of landscape appreciation derived from social media data can provide flexible, sensitive simulation tools for assessing how future changes in landscape management may impact indicators of cultural ecosystem service value. Similar models could be developed in any region with sufficient social media data for parameterisation, and could be altered to focus on management actions at smaller spatial scales.