Assessing Cultural Ecosystem Services of Riverscapes Using Artificial Intelligence and Photographs

The application of artificial intelligence is gaining traction in the study of cultural ecosystem services (CESs). In response to this trend, we have developed an approach based on photo categorisation and semantic segmentation to assess CESs, including aesthetic values in riverscapes and river elements (large wood, erosion banks and river bars). Semantic segmentation allows the precise identification of photos, including quantification of features coverage. To validate our approach, we chose the riverscape of the Odra River in the Czech Republic, a smaller but significant site protected under the Ramsar Convention and Natura 2000. Our goal was to understand the content of the photographs and to delineate the categories of CESs present in this region. We demonstrated the efficacy of our approach and found that our manual categorisation of publicly available photos from Google Maps and Flickr revealed a consistent trend: visitors predominantly took photos from the river and pond and urban categories. CESs presented in the area include outdoor recreation, bird watching, education and hiking. Semantic segmentation analysis revealed that the photographs primarily depicted water, vegetation types and sky, while the river elements in the photographs were of minimal size. Our results highlight the need for improved visitor education about riverscapes, including river elements defined as important hydromorphological parts of the river. Our method does not require extensive programming knowledge, making it suitable for river managers. This approach allows for better monitoring of important riverscape features, which may contribute to more effective river protection.