A cross-regional machine learning transfer prediction framework for heat risk based on environmental characteristics of mature new towns
- 1Department of Smart City Engineering, Hanyang University ERICA Campus, Ansan, Republic of Korea
- 2School of Architecture and Architectural Engineering, Hanyang University ERICA Campus, Ansan, Republic of Korea
Abstract
Heat risk assessment for planned new towns is challenging because actual thermal environments, population activities, and built environment conditions are not yet fully formed. This study examines whether heat risk relationships identified in a mature new town environment can be translated into planning stage variables to support ex-ante risk screening before development is completed. Using Bundang-Pangyo, South Korea, as the mature reference area, composite heat risk was characterized through heat hazard, exposure, and built environment vulnerability. Spatial analysis revealed substantial within city heterogeneity, indicating that heat risk relationships were not spatially uniform across the reference area. Planning stage land use, road, and building development characteristics were then used as common intermediary variables for cross-regional machine learning prediction. The results show that conventional land use and road characteristics captured meaningful variation in heat risk, while the inclusion of additional building development and three-dimensional planning information substantially improved predictive performance. Application to Goyang Changneung and Namyangju Wangsuk produced clearly differentiated planning stage risk profiles, demonstrating that the transferred framework can distinguish spatial variations associated with different planned urban configurations. SHAP analysis indicated that planned cooling balance, planned normalized building coverage ratio, and planned impervious surface ratio were among the more prominent variables associated with the transferred predictions. These findings emphasize the joint importance of ecological cooling capacity and development intensity for planning stage heat risk screening. Methodologically, the study provides a way to translate mature city heat risk knowledge into variables available before construction, extending urban heat risk assessment from post development diagnosis toward planning stage comparative screening.
Related
- Hybrid stacked bidirectional deep learning with attention based air pollution prediction for environmental sustainability
- A geospatial machine-learning framework for regional electrification estimation mapping and trend analysis using demographic and infrastructure features
- An explainable deep learning framework for sub-seasonal to seasonal agricultural drought prediction over India
- A hybrid spatio-temporal deep learning framework for urban travel time prediction and intelligent routing: Integrating recurrent and graph neural networks
- Spatially compounding climate extremes and environmental risk: cross-regional synchrony, spatial buffering loss, and systemic consequences
BibTeX
@article{lee2026crossregional,
title = {{A cross-regional machine learning transfer prediction framework for heat risk based on environmental characteristics of mature new towns}},
author = {Dongyoon Lee and Juncheng Zeng and Hwanyong Kim},
journal = {Frontiers in Environmental Science},
year = {2026},
month = oct,
doi = {10.3389/fenvs.2026.2012398},
url = {https://doi.org/10.3389/fenvs.2026.2012398},
note = {CC BY},
}
RIS
TY - JOUR
TI - A cross-regional machine learning transfer prediction framework for heat risk based on environmental characteristics of mature new towns
AU - Dongyoon Lee
AU - Juncheng Zeng
AU - Hwanyong Kim
PY - 2026
DA - 2026/10/28
JO - Frontiers in Environmental Science
DO - 10.3389/fenvs.2026.2012398
UR - https://doi.org/10.3389/fenvs.2026.2012398
AB - Heat risk assessment for planned new towns is challenging because actual thermal environments, population activities, and built environment conditions are not yet fully formed. This study examines whether heat risk relationships identified in a mature new town environment can be translated into planning stage variables to support ex-ante risk screening before development is completed. Using Bundang-Pangyo, South Korea, as the mature reference area, composite heat risk was characterized through heat hazard, exposure, and built environment vulnerability. Spatial analysis revealed substantial within city heterogeneity, indicating that heat risk relationships were not spatially uniform across the reference area. Planning stage land use, road, and building development characteristics were then used as common intermediary variables for cross-regional machine learning prediction. The results show that conventional land use and road characteristics captured meaningful variation in heat risk, while the inclusion of additional building development and three-dimensional planning information substantially improved predictive performance. Application to Goyang Changneung and Namyangju Wangsuk produced clearly differentiated planning stage risk profiles, demonstrating that the transferred framework can distinguish spatial variations associated with different planned urban configurations. SHAP analysis indicated that planned cooling balance, planned normalized building coverage ratio, and planned impervious surface ratio were among the more prominent variables associated with the transferred predictions. These findings emphasize the joint importance of ecological cooling capacity and development intensity for planning stage heat risk screening. Methodologically, the study provides a way to translate mature city heat risk knowledge into variables available before construction, extending urban heat risk assessment from post development diagnosis toward planning stage comparative screening.
SN - 2296-665X
ER -