Analysis Ready Data (ARD) for Detecting Urban Gentrification: Current Capabilities and Challenges

Abstract

The growing abundance of remote sensing data has led to increased challenges in accessing, integrating, and efficiently analyzing geospatial information. To address this issue, the concept of Analysis Ready Data (ARD) has emerged, offering a solution that requires minimal user effort and promotes interoperability across time and different datasets by adhering to specific processing requirements. This research focuses on exploring the capabilities and limitations of ARD in detecting changes in urban land use and land cover, with a particular emphasis on urban gentrification. A use case was developed to identify gentrification patterns using remote sensing time series and machine learning algorithms. The training dataset comprised Landsat ARD and Sentinel ARD acquired from USGS and ESA, respectively, spanning the years 2013 to 2022. Initial processing involved extracting pixels within the area of interest based on building footprints from 2013. Quality assurance procedures were then applied to retain only clear and reliable pixels. The resulting data was organized as a space-time cube spanning the years 2013 to 2022. Temporal changes were subsequently cross-referenced with construction permits to track the development of new buildings. These findings were utilized as training data for a machine learning program designed to identify gentrification patterns based on land changes in other regions. The study reveals that ARD significantly streamlines the preprocessing phase, although obtaining accurate time labels for each pixel remains a challenging aspect.