University of Edinburgh logo British Antarctic Survey logo National Oceanographic Centre logo University of Leeds logo

PhD Projects

Greenness versus greyness: greenspaces in and around cities for hazard mitigation and resilience

Satellite Image Composite over Quito city, Ecuador and surrounding landscape illustrating greyness of city urban extent versus greeness of the mountains and surrounding area.
Satellite Image Composite over Quito city, Ecuador and surrounding landscape illustrating greyness of city urban extent versus greenness of the mountains and surrounding area. (GoogleEarth)

The overarching aim of this proposal is to increase our resilience to geo- and climatic hazards by assessing the current state and identifying the future potential of greenspaces in and around cities in the Global South to better inform future urban expansion. This will be achieved using Earth Observation satellites comprising the latest spectral information and developing data science techniques to analyse indices for statistical comparisons to apply to this problem.

Background & Motivation
More than half the world’s population live in urban areas, and this is rapidly increasing through time. Much of these new urban spaces are currently being created, so now is the time that we can best inform the development of these areas to better create sustainable cities which are resilient to geo- and climatic hazards. One component of cities that has been highlighted as enriching sustainability is that of greenspaces within and around urban areas that offer many benefits to society in the form of health and wellbeing (Cartier, 2021). Identifying and fully characterising greenspaces has been difficult to achieve so far as it is intensive to do on a detailed city scale. The recent rapid growth in the availability of Earth Observation data provides the timely opportunity to capture the pattern of urban expansion and greenspace change across many cities on a global scale. Fully exploiting Earth Observation data and harnessing its potential is at the core of this project to address the scientific questions that can help with some of the UN Sustainable Development Goals, including sustainable cities, life on land and climate action.

The initial objective is to establish current patterns and rates of the growth of cities as well as the loss and renewal of greenspaces in and around them, with particular focus on cities exposed to natural hazards (geo and climatic) in the vulnerable parts of the world exposed to such disasters. This will be achieved using the extent of vegetation and urban surfaces in and around cities and adopting a decadal time series approach as it changes and expands through time using multiple satellite sensors and the latest deep learning techniques.
This project will look to identify the co-benefits of using such spaces for resilience against geohazards and climate hazards which work on different timescales. For example, greenspaces can reduce the heat island effect and potential for flooding and fires on an annual basis but can also be useful for resilience and disaster management from geohazard shocks in the long term from landslides, earthquakes and volcanic eruptions.

This project will exploit the optical, hyperspectral and radar domains to characterise greenspaces in and around cities, compared to the extent of concrete (greyness) using Earth Observation data, and then generate a quality index of greenspaces (including scrub, grasslands, and forests) that maybe affected from heat and water stresses, as well as possibly identifying particular plant species and their relative vulnerability to stresses. This will highlight priority areas to target for greening projects (scaled for example by the highest heat island effect zones to indicate where it is most required) as well as help identify potential loss of valuable greenspaces from urban sprawl.

In doing so the project will generate new and novel approaches and algorithms that are generally applicable to cities globally, but individual cities will be targeted to develop case studies and assess the validity of generalising the approaches across different land covers and latitudes. Good starting cities to focus on (in which we are currently doing ongoing research into geohazards) are Quito (Ecuador), Bishkek (Kyrgyzstan) and Nairobi (Kenya). Depending on the interests of the candidate, the project could expand along the impact of greenspace on flooding hazards such as the roughness, catchment infiltration and storage areas in flood modelling. Or more focus could be on the Urban Heat Island Effect and the effect of greenspaces have (Wang et al., 2021) using satellite thermal data which could inform policy on where and how much an urban planner would need greenspace allocation to mitigate one degree of warming.

This is timely as there is an expanding fleet of satellites with advanced sensors onboard to provide data with a global reach to tackle some of these challenges. Sentinel-2 has provided ubiquitous and voluminous multispectral data that is establishing a large and ongoing archive that can be exploited for the initial analysis and training for deep learning models. From 2018 there has been a rapid increase in the amount of hyperspectral data (hundreds of spectral bands) from a growing number of satellite-based platforms providing medium resolution data (30 metres) over swath extents 30-60 km wide (Transom et al., 2018). Given the growing amount of data available, for us to be able to harness the potential of these big datasets requires the application and development of processing and advanced analysis techniques, as well as the implementation of pixel and object-based Deep Learning approaches and the fusing of differing datasets (Ren et al., 2020) (such as with hyperspectral, multispectral, optical, and digital elevation data). Furthermore, much higher resolution data (metre scale) from PlanetScope could contribute to the object-based deep learning approaches. In addition to current datasets such as from Sentinel-2 and soon to be launched systems such as EnMAP, the project will also make use of older and legacy datasets from ASTER, Hyperion and Landsat as well as potentially historic spy satellite imagery to establish much longer timeseries of urban greenspace change. Other potential datasets to use are high resolution optical stereo imagery for digital elevation model generation and the exploitation of radar imagery and interferometric data which is used in the assessment of geohazards (Elliott, 2020) and SAR for soil moisture, depending on the direction of the project taken.

The studentship will be held at the University of Leeds, and the candidate will receive extensive training in Earth Observation and Machine Learning Techniques. 

Cartier, K. M. S. (2021), Growing equity in city green space, Eos, 102, doi:10.1029/2021EO158443.
Elliott, J. R. (2020). Earth Observation for the assessment of earthquake hazard, risk and disaster management. Surveys in Geophysics, 1-32. doi:10.1007/s10712-020-09606-4.
Ren, K., Sun, W., Meng, X., Yang, G., & Du, Q. (2020). Fusing China GF-5 Hyperspectral Data with GF-1, GF-2 and Sentinel-2A Multispectral Data: Which Methods Should Be Used?. Remote Sensing, 12(5), 882, doi:10.3390/rs12050882.
Transon, J., d’Andrimont, R., Maugnard, A., & Defourny, P. (2018). Survey of hyperspectral earth observation applications from space in the sentinel-2 context. Remote Sensing, 10(2), 157, doi:10.3390/rs10020157
Wang, X., Dallimer, M., Scott, C. E., Shi, W., & Gao, J. (2021). Tree species richness and diversity predicts the magnitude of urban heat island mitigation effects of greenspaces. Science of The Total Environment, 770, 145211, doi:10.1016/j.scitotenv.2021.145211