Authors: Oleksandr Karasov*, University of Tartu, Evelyn Uuemaa, University of Tartu, Tiit Tammaru, University of Tartu
Topics: Urban Geography, Remote Sensing, Digital Geographies
Keywords: remote sensing, google street view, socioeconomic, income, sentinel
Session Type: Virtual Poster
Start / End Time: 8:00 AM / 9:15 AM
Room: Virtual 26
Presentation File: Download
Remotely sensed and ground-based data have become the major proxies on the various aspects of socioeconomic status of the neighborhoods. We aimed at combining globally available remote sensing data (Sentinel-2 imagery of 10 m spatial resolution) and Google Street View imagery in an integrated research framework to explain the average gross salary levels in Tallinn (the capital of Estonia) by means of harmonized classification of neighbourhoods from space and ground. We compared the quality of pixel- and object-based supervised neighbourhood classification of Sentinel-2 cloudless mosaics over Tallinn for 2018. We also utilised a pre-trained convolutional 48-layered neural network (Google's InceptionV3) to identify visually similar Google Street View panoramic images over Tallinn for 2018 according to the remote sensing-based classification. Finally, we related the share of particular neighbourhood types as well as some morphological urban characteristics (composition and configuration of neighbourhood types) with supplementary data (Airbnb price per night, number of restaurants, etc) to the average gross salary of Tallinn's dwellers from the annual national survey. Our results suggest that joint use of freely and globally available remote and social sensing data allows a reliable approach to map socioeconomic status of neighbourhoods, potentially across scales.