Speaker Bio & Abstract
BiographyAlexandre works as a Data Scientist for the OECD Centre for entrepreneurship, SMEs, regions and cities. He is in charge of implementing the smart data strategy on geospatial data within the organisation, more particularly Earth Observation data, and to contribute to the activities of the OECD Laboratory for Geospatial Analysis. After finishing his master's degree, he worked as an energy data scientist for the US R&D centre of Electricit de France (EDF) in the Silicon Valley. He then joined the OECD early 2021. Alexandre holds a Master's degree in Science and Executive Engineering from Mines ParisTech-PSL, as well as a Master of Science in Mathematics from ENS Paris-Saclay.AbstractMonitoring land use in OECD cities using satellite imagery and deep learningAcross OECD countries, new cities emerge, and established cities expand their physical footprint on land. Whether the land footprints of cities are compact or become more spread out will have a lasting influence on how urban challenges of mobility, carbon emissions, housing affordability, the cost of infrastructure provision, the ease of peoples access to services or business relations, and many more. This study lays a methodological basis for the monitoring of land use in OECD cities, on an internationally consistent basis. An advanced form of machine learning, namely a deep-learning model called U-Net, is applied to classify land use in EC-ESA Sentinel satellite imagery. This approach is complementary to conventional statistical data on land use, as large surfaces of land can be monitored efficiently and on a near real-time basis. As a case application, several economically relevant types of urban land use are mapped and analysed for 687 European metropolitan areas in 2021. Further analyses explore the speed and spatial shape of recent urban expansion as well as the models potential for monitoring OECD countries globally.