Work

Early detection of gentrification risk using Landsat ARD and machine learning

Year: 2025

Type: article

Abstract: Mapping gentrification and displacement risk is essential for understanding the social and economic impacts of urban redevelopment. Traditional approaches detect gentrification by tracking shifts in census and survey indicators (income, rent, tenure, demographics). These data are coarse and delayed, so detection is typically post-event rather than early. This study explores a new framework that uses Landsat Analysis Ready Data and New York City construction records to detect parcel-level reconstruction and combine those signals into a tract-level building reconstruction intensity, compared with a citywide baseline to flag tracts at risk of gentrification. Two machine learning models were trained without socioeconomic inputs. The models achieved an overall accuracy of 87% and F1 scores of 0.78 for redevelopment identification, and the agreement with two tract level reference maps is moderate, as expected given scope differences. The findings demonstrate that satellite time series offer citywide, continuous coverage, reducing administrative lags and applying a consistent detection method across neighborhoods. The resulting maps serve as early warning flags for the risk of gentrification. This approach could help planners and policymakers anticipate and reduce negative impacts, including displacement and housing instability, while promoting equitable access to housing and inclusive neighborhood development. The study also highlights the value of combining AI with Earth observation to map urban change. Future work will incorporate socioeconomic and environmental indicators, as well as utilize higher-resolution imagery, to enhance the interpretation and detection of small sites. • Gentrification early detection from remote sensing time series and ML. • The study achieves high accuracy in detecting reconstructions with remote sensing. • Models can generalize reconstruction patterns across spatial and temporal ranges. • Moderate agreement with gentrification maps from two socioeconomic-based studies. • Results are analysis-ready to support early action on gentrification/displacement.

Source: Remote Sensing Applications Society and Environment

DOI: 10.1016/j.rsase.2025.101757

Author: Li Lin

Institution: United Arab Emirates University

Language: English

Topic: Remote Sensing and Land Use

Subfield: Atmospheric Science

Field: Earth and Planetary Sciences

Domain: Physical Sciences

SDGs: Sustainable cities and communities, Reduced inequalities

Open Access status: hybrid

Estimated APC paid: $2,640