The positive effects of nature on mental health are well-known. In response, cities are “greening” to improve the well-being of their residents. Urban ecosystem services (ES) assessments can enable city planners to spatially model the contributions of natural features to people well-being. To do this, novel ES models must be developed to quantify these contributions. Quantifying the experience of nature is a key component in these new models but it is difficult to quantify. In this project, the potential of computer vision and
social media data to quantify the experience of nature in the urban environment will be explored to support the development of urban ES models.
Humans hold deep connections with nature and research has shown that regular exposure to nature has significant positive effects on our mental health and well-being. At the same time, the world is experiencing rapid rates of urbanisation. Thus, ensuring people have a regular exposure to nature in their daily lives has become an increasing priority for policymakers.
Ecosystem Services (ES) have emerged as a concept to help us better understand, value and manage the contributions of ecosystems to human well-being. In response to the rapid rate of global urbanisation, the focus in ES research has shifted towards the urban environment. However, urban ES modelling is still in its infancy and methods need to be developed to support ES assessments.
Quantifying peoples’ experience of nature has been identified as one key factor in such modelling exercises. This links the presence of natural features to the resulting mental health benefits experienced by people. However, these are difficult to capture on any large scale without the use of extensive and costly survey methods. It is therefore beneficial in many cases to model benefits directly on the natural features present, and the potential exposure to these features using population proximity measures.
Recently, peoples’ use of social media such as Flickr has generated rich and expansive new sources of data on how people experience and interact with nature. At the same time, rapid advances in computer vision and the availability of training data offer new opportunities to quantify people's experience of nature in the urban environment. For example, the Places365 database provides a labeled image dataset with a number of naturerelated scenes and attributes. Consequently, a computer vision algorithm trained on this dataset can give insights into the type of human-nature interactions that may be taking place using the imagery uploaded to social media.
This research sets out to clarify the link between natural features and people's experiences of nature in the urban environment using computer vision and social media data. First, geo-located social media data will be downloaded for an urban study area. Second, naturerelated scene categories and attributes will be extracted using a computer vision algorithm trained on the Places365 database. Third, these results will be related to natural features using regression techniques. The result will generate insights into the type of natural features that generate (different) experiences of nature. This will assist the development of ES models and their use in urban planning for human well-being.
- Download Flickr imagery for an urban environment
- Extract nature-related scene categories and attributes using Places365 database
- Compare to presence of natural features using regression methods
- Bratman, G. N., Anderson, C. B., Berman, M. G., Cochran, B., De Vries, S., Flanders, J., ... & Kahn, P. H. (2019). Nature and mental health: An ecosystem service perspective. Science advances, 5(7), eaax0903.
- Gómez-Baggethun, E., & Barton, D. N. (2013). Classifying and valuing ecosystem services for urban planning. Ecological economics, 86, 235-245.
- Ibrahim, M. R., Haworth, J., & Cheng, T. (2020). Understanding cities with machine eyes: A review of deep computer vision in urban analytics. Cities, 96, 102481.
- Cox, D. T., Shanahan, D. F., Hudson, H. L., Plummer, K. E., Siriwardena, G. M., Fuller, R. A., ... & Gaston, K. J. (2017). Doses of neighborhood nature: the benefits for mental health of living with nature. BioScience, 67(2), 147-155.
- Zhou, B., Lapedriza, A., Khosla, A., Oliva, A., & Torralba, A. (2017). Places: A 10 million image database for scene recognition. IEEE transactions on pattern analysis and machine intelligence, 40(6), 1452-1464.
- Ilieva, R. T., & McPhearson, T. (2018). Social-media data for urban sustainability. Nature Sustainability, 1(10), 553-565.
- Programming experience, or the strong motivation to learn it
- Experience with machine/deep learning, computer vision, or advanced statistics is highly recommended (e.g. Completion of GRS-34806 Deep learning course or equivalent)
Theme(s): Modelling & visualisation; Human – space interaction