Classifying and Mapping Cultural Ecosystem Services Using Artificial Intelligence and Social Media Data.

Area of Science:

• Ecosystem Services
• Computational Ecology
• Social Media Analysis

Background:

• Quantifying cultural ecosystem services (CES) is challenging due to their intangible nature.
• Social media data, like geo-tagged photos, offers potential for mapping ecosystem use and appreciation.
• Manual analysis of large photo datasets is time-consuming, limiting scalability.

Purpose of the Study:

• To develop and evaluate a deep learning model for automated classification of natural and human elements in social media photographs relevant to CES.
• To assess the model's accuracy in classifying landscape and nature appreciation from Flickr images along the Lithuanian coast.
• To demonstrate the utility of automated analysis for large-scale CES quantification and mapping.

Main Methods:

• Utilized a convolutional neural network (CNN) architecture for image classification.
• Analyzed over 29,000 geo-tagged Flickr photographs from the Lithuanian coast.
• Employed hierarchical clustering to group photographs and assessed classification accuracy against manual methods.

Main Results:

• The deep learning model accurately classified photographs, with 37% related to landscape appreciation and 28% to nature appreciation.
• Identified key clusters in urban coastal areas and specific natural attractions with high vegetation and animal cover.
• The automated approach saved an estimated 100 person-kilometers of manual classification work.

Conclusions:

• Automated analysis of crowdsourced social media data using deep learning provides an efficient tool for quantifying and mapping CES.
• This methodology enables large-scale assessment of CES, supporting informed urban planning and nature reserve management.
• The study highlights the potential of digital photography analysis to expand research capabilities in ecosystem service assessment.