UC Irvine

Ecosystems and societies face increasing pressures from global change, including climate and land use changes. Subsequent impacts to ecosystems are compounded through the interactions, dependencies, and feedbacks with human systems (i.e. social-ecological systems). The interdependence of ecosystems and society is manifested through the invaluable contributions that nature provides to people, commonly referred to as ecosystem services. As climate change alters the structure and functioning of ecosystems, there are corresponding impacts on ecosystem services, consequently impacting human well-being. While climate change impacts on ecosystems are well studied, the subsequent impacts on the services they provide to people are less understood, especially in the case of the non-material benefits people derive from nature (i.e. cultural ecosystem services). Thus, we lack a holistic understanding of the implications of climate change on human well-being. My dissertation draws on a multitude of disciplines to unveil and test innovative methods, tools, and data to overcome historical limitations in studying cultural ecosystem services and assess their vulnerability to climate change. Specifically, I focus on using social sensing data (i.e. sensing/data collected from humans and/or their devices) and machine learning to map, model, and value climate change impacts on cultural ecosystem services in a holistic social-ecological fashion and in understudied data-poor regions of the world. Ultimately, this work develops and tests new approaches to accounting for the non-material benefits of nature to people in order to create a more comprehensive understanding of how climate change will impact human well-being.The first chapter of this dissertation systematically reviews the literature to assess how machine learning and big data are being used in ecosystem service research to overcome major gaps in the field. I find that although cultural ecosystem services are understudied across the general literature, studies that are starting to incorporate these tools are able to address historical limitations to better account for the subjectivity, nonlinearity, and complexity of cultural ecosystem services. The second chapter develops and tests new methods, informed by the insights of chapter one, using machine learning (Random Forest) and social sensing data (Flickr) to map recreational ecosystem services across California, model the social-ecological drivers of recreation, and model the future impacts of climate change on recreation. I find that the social sensing data effectively represents recreational ecosystem service patterns across the landscape. This allowed me to use machine learning (i.e. Random Forest) to connect ecosystem service use to social-ecological drivers, showing access to cultural ecosystem service supply and climate play significant roles in driving flows to people. Further, our machine learning model predicts that climate change will exacerbate peak season recreational patterns, with highly popular regions becoming more suitable for cultural ecosystem service use and vice versa. The third chapter expands on this methodology to test its use at a large scale in a historically understudied and data-poor region. Social sensing data from eBird is used to represent the use of cultural ecosystem services across the continent of Africa, Random Forest is used to explore the social-ecological drivers of cultural ecosystem services, and Maximum Entropy Modeling is used to map cultural ecosystem service suitability and predict future suitability. I find that social sensing data from eBird is a globally available and effective proxy for cultural ecosystem service use. Further, I find that climate change will increasingly constrain the flow of cultural ecosystem services to people across Africa, with biodiversity change also playing a large role, and land use change playing a much more moderate role. Overall, I demonstrate how these tools, data, and methods can be utilized to scale and implement the study of ecosystem service flows and future impacts to them across the world. The fourth chapter expands on chapter three by integrating environmental economics methods to value the non-market utility of cultural ecosystem services and future impacts. I use social sensing data from the eBird citizen science project, machine learning (i.e. Maximum Entropy modeling), and econometric methods (i.e. the travel cost method) to map, model, and value current and future birding cultural ecosystem services across a biodiversity hotspot, South Africa. Leveraging social sensing data, I reveal national patterns of birding-related cultural ecosystem servcies and identify the beneficiaries, enabling the valuation of non-market ecosystem service demand. Additionally, through using both the social sensing data and Maximum Entropy modeling, I discern variations in the value of cultural ecosystem services, the social-ecological drivers, and the differential effects of anticipated climate, biodiversity, and land cover changes on domestic and international beneficiaries. This analysis and valuation showcases the potential of emerging data and tools to enhance and scale more holistic social-ecological analyses that account for non-market non-material ecosystem service value, improving the integration of ecosystem services in management and policy for a future shaped by global change.