Comparing Designated Green Areas and Satellite-Derived Vegetation Greenness in Housing Price Prediction Using Explainable Artificial Intelligence

The study examines how formally designated green areas and satellite-derived vegetation greenness, measured using the Normalized Difference Vegetation Index (NDVI), are associated with apartment prices in Seoul. Using 10,629 apartment transactions from 2022, we compare a semi-log hedonic price model with Random Forest, XGBoost, and LightGBM and apply explainable artificial intelligence techniques to interpret the best-performing model. The results show that a higher proportion of designated green area is generally associated with lower predicted housing prices, whereas higher NDVI values are associated with higher predicted prices, although this positive predictive relationship weakens at higher NDVI levels. These contrasting patterns indicate that planning-based green-area designations and satellite-observed vegetation greenness capture different dimensions of the residential environment. Because NDVI measures vegetation greenness and vigor rather than its maintenance, accessibility, usability, or perceived quality, the findings should not be interpreted as evidence of vegetation quality or residents’ preferences for particular types of green space. The study demonstrates the value of considering planning-based and remotely sensed vegetation measures jointly when examining relationships between urban environmental characteristics and housing prices.