Can machine learning unlock a bright green future for desert cities?
Machine learning could help arid cities protect vital green spaces as climate change, urban growth and water scarcity intensify
Sponsored by United Arab Emirates University
Sponsored by
Sponsored by
Arid cities like Al Ain in the United Arab Emirates rely on green spaces to cool their streets, improve air quality and enhance their citizens’ well-being. Encircled by desert, these cities are especially vulnerable to climate change. Water is already scarce. A hotter, drier climate would bring with it serious problems for policymakers who want to maintain green spaces for future generations – especially as the city continues to grow.
At the United Arab Emirates University (UAEU), researchers have developed a framework that applies machine learning to the problem, and it could be used as a screening tool to help planners identify high-suitability areas that may merit protection, assess where vegetation is most vulnerable to climate change and urbanisation, and identify locations for new greening while taking water constraints into account, subject to local field validation.
“The larger objective is not simply to create more green space,” says Naeema Al Hosani, professor and chair of UAEU’s Department of Geography and Urban Sustainability and director of the Desert Environment Research Center. “It is to help cities protect the right areas, select interventions that can endure local conditions, and gain the greatest cooling, ecological and social benefit from every unit of water and land invested.”
The framework was designed to map vegetation suitability under baseline conditions and a projected 2045 high-emissions scenario. The multidisciplinary project brings together UAEU researchers and colleagues from Northwest University in China and the University of Colorado Boulder’s Cooperative Institute for Research in Environmental Sciences. It uses downscaled CMIP6 climate projections from the CHELSA dataset.
The team applied four machine-learning algorithms to 16 environmental, climatic, hydrological and socioeconomic variables. These datasets were carefully standardised. For the 2045 scenario, two bioclimatic indicators – the maximum temperature of the warmest month and precipitation during the driest month – were used to represent extreme heat and aridity.
“Future projections always contain uncertainty, so our aim was not to claim that one exact map will become reality,” says Mona Ramadan, assistant professor in UAEU’s Department of Geography and Urban Sustainability and a researcher at the Desert Environment Research Center. “It was to create a credible stress test that shows where vulnerability could emerge if climate and urban pressures intensify.”
Finding the balance between predictive accuracy and the interpretability of the results was key. So too was creating a framework that delivered outputs that could be easily understood by planners.
The team’s findings highlight the difficulties facing arid cities in a warming world. Maintaining urban green spaces will depend on efficient water management, the surface materials used in development and how urban growth changes the city’s microclimate. Under the study’s 2045 high-emissions scenario, the framework projected a modest decline in the share of land with excellent vegetation suitability, with the remaining areas with excellent suitability projected to become more geographically fragmented. “For planners, the loss of continuity between high-quality green areas can be as important as the total number of hectares lost,” says Al Hosani.
The team hopes to adapt and test the framework in other cities in the region, and expand its functionality beyond telling planners where vegetation might thrive to assess how different interventions may perform under different planning and climate scenarios, including their water and maintenance requirements.