Using Time Series Deep Learning Models to Track and Predict Green Space Distribution across Communities in Washington, DC
Despite the various benefits of urban green space, the environmental justice literature points out its inequitable distribution and warns that it will persist. However, descriptions of urban history also point to the changing socioeconomic profile of communities surrounding green space. These two conflicting perspectives raise the following research question: Is spatial inequity a fixed condition, or a dynamic process that evolves with the city itself? This question is particularly pressing in Washington, DC, where historical injustices intersect with neighborhood change. Moving beyond a static diagnosis, this study investigates the temporal dynamics of green space equity to understand how its distribution has changed and to forecast where future inequities may emerge. This study traces green space change from 1990 to 2020 and forecasts to 2050 using time series deep learning models, namely, Gated Recurrent Unit and Convolutional Gated Recurrent Unit. Forecasts from these two deep learning models show a complex pattern. While both project a net citywide increase, the more spatially detailed model reveals average tract-level declines, indicating distributional deterioration. Beyond diagnosis, the study offers policy-ready predictive maps that can operationalize equity by prioritizing corridors and boundaries for protection or restoration, sequencing investments toward vulnerable tracts, and prospectively tracking distributional targets.
Authors
- Minkyu Park (ORCID: https://orcid.org/0009-0005-7983-495X)
- Seongbeom Park (ORCID: https://orcid.org/0000-0002-0759-6826)
- Jaekyung Lee (ORCID: https://orcid.org/0000-0001-5168-6450)
Institutions
- Hongik University (KR)
Publication Details
- Journal
- Journal of Urban Planning and Development
- Published
- 2026-11-19
- DOI
- https://doi.org/10.1061/jupddm.upeng-6891
- Primary Topic
- Urban Green Space and Health
- Type
- article
- Field-Weighted Citation Impact
- 0.00