Investigating cultural ecosystem services of the Caatinga on Flickr

Abstract

Public interest in nature can be promoted through social media by assessing the importance of a species to people and identifying new emblems of conservationist appeal. We aimed to assess the public interest in cultural ecosystem services in the Caatinga (seasonal dry forest). Ecosystem services were categorized based on approximately 1500 photographs posted on Flickr. These photographs were analyzed using manual and deep-learning (DL) approaches. The most observed categories for both approaches were “Enjoyment of the Landscape” (36.8%), “Appreciation of Nature – Animals’’ (25.6%), and “Social Activities” (19.3%). However, we found significant differences between manual and DL classifications owing to the difficulties in classifying categories using the DL model. The findings suggest a low cultural ecosystem service representation on the photo-sharing platform Flickr in the Caatinga region, even after removing 67% of the collected data. This may be attributed to the limited interest in Flickr among the Caatinga residents. Deep learning (DL) techniques hold potential for studying cultural ecosystem services, but their efficacy depends on the algorithm's capacity to discern human-nature interactions and various natural elements. Our observations indicate that increasing the scale of the training and test datasets and incorporating additional categories to account for Caatinga diversity may enhance the results.