Geunhan Kim

Urban heat mitigation requires knowledge of land surface temperature (LST) responses to land-cover conversion, including the location and magnitude of change. SHapley Additive exPlanations (SHAP) cannot estimate such conversion effects, and existing effect estimators may violate the constraint that land-cover areas within a bounded space sum to a constant. This study proposed a composition-preserving accumulated local effects (ALE) approach, wherein an increase in one land cover is offset by an equal decrease in another, to preserve the total area within each ring. Using 2019 summer daytime Landsat 8 LST, level-3 land cover, and terrain data for Seoul, an XGBoost model with multi-ring buffers (50-m intervals to 500 m) achieved R² values of 0.956 without and 0.965 with vegetation indices. Converting 2,000 m² of single-detached residential land to other grasslands reduced LST by 0.64 °C within 50 m, compared with 0.07 °C at 100–150 m, with the area required for 0.3 °C of cooling increasing from 1,018 to 6,085 m². Area accounted for 59.2% of the variation in greening effect, with 93.7% concentrated within 100 m. The band with the greatest SHAP contribution (50–100 m) differed from that with the largest conversion effect (≤50 m), indicating that the distances explaining current LST patterns may differ from those at which conversion is most effective. This framework distinguishes variable contribution from conversion effects and supports the prioritization of greening location, cover type, and target area in space-constrained cities.

Land surface temperature (LST) is a representative indicator for predicting urban heat island. The single-buffer approach, which has been used to analyze the relationship between LST and explanatory variables, cannot capture the effects of both the distance of independent variables and the area of land cover included within the buffer zone. Therefore, we aimed to investigate the relationship between LST and natural and artificial land cover in urban regions, using a multi-ring buffer model to refine independent variables in 50-m increments, by land cover type. Four tree-based machine learning models—Random Forest, LightGBM, CatBoost, and XGBoost—were compared. XGBoost exhibited the highest performance, with R² range of 0.88–0.98 across 50–500-m intervals. LST prediction analysis was effective up to 350 m. Further, the significant spatial range of variables shifted with 50-m intervals, enabling a more granular analysis of the significant spatial range of independent variables. The restoration simulation showed that greening efforts on the Western Arterial Road and Jemulpo Tunnel reduced LST within an approximate range of 150 m. These findings provide crucial scientific evidence for determining the placement of urban parks and formulating urban planning and management strategies aimed at optimizing LST reduction in urban environments.

In urban areas with poor plant growth environments, selecting trees for planting requires the consideration of the urban environment as well as the local conditions surrounding the proposed planting site. To derive the key factors for the optimal selection of urban trees for planting, this study aimed to employ the extreme gradient boosting machine learning method and Shapley additive explanations to create an eXplainable AI model. This approach enabled the analysis of trees planted within a similar period but exhibiting different vitalities using the normalized difference vegetation index and facilitated the identification of relationships between trees and a range of urban environments and local conditions. The explanatory power of the model was 0.454–0.605 based on the R ² of different tree species. The spatial distance that should be considered when planting appropriate urban trees varied from 10 to 40 m. Notably, the priority of the influencing factors varied among tree species. Considering various factors, a map suggesting the best trees for planting within the experimental area was generated. Utilizing the results of these analyses, the trees selected are expected to grow well in areas with the given conditions when planted in urban spaces, ultimately providing a scientific basis for urban tree planting and management policies.

The recognition of the various functions of urban greenery and urban trees, including their greenhouse gas (GHG) absorption capacity, is expanding. However, studies on trees in urban areas are relatively limited compared with those in natural or non-urban areas, and the scope of research is limited to major tree species or specific stress-related factors. Here, we identified the main trees that can be planted in urban conditions in South Korea based on comprehensive literature review and expert opinions. We considered not only the inherent carbon storage abilities of trees, but also their ability to adapt and grow in urban conditions. First, based on the review, we selected 24 tree species portraying the highest preference and frequency of planting in urban areas, from a total of 166 forest tree species planted or recommended for planting in South Korea. The ecological suitability and economic feasibility of the selected tree species were analyzed based on expert opinions and the fuzzy multiple-criteria decision-making (FMCDM) model. This analysis led to the prioritization of the tree species, identifying those with the highest potential for urban planting. Some of the suitable tree species were Zelkova serrata, Ginkgo biloba, Chionanthus retusus, and Prunus yedoensis Matsum. To establish a systematic foundation for sustainable urban development and carbon neutrality, it is necessary to prioritize the development of carbon emission factors (CEFs) for tree species among the high-priority candidates currently lacking such data (e.g., Styphnolobium japonicum, Cornus kousa, Chamaecyparis obtusa, etc). Furthermore, the creation of an integrated planting model comprehensively considering various functions is essential. Our results can contribute to urban planning, design, and policymaking, especially with respect to the development of efficient carbon sinks.

Sung-Hyun Gong, Hyung-Sup Jung, Geun-Han Kim, Geun-Hyouk Han, Il-Hoon Choi, Jin-Sung Hong. Korean J. Remote Sens. 2025;41:461-74. https://doi.org/10.7780/kjrs.2025.41.2.2.6

Air pollution in East Asia presents critical environmental and health challenges, particularly in industrial regions affected by domestic and cross-border emissions. This study developed a GEO AI dataset specifically for industrial park segmentation, integrating Sentinel-2 satellite imagery, Geostationary Environment Monitoring Spectrometer (GEMS) geostationary satellite data, and Air Quality Monitoring Network data. Optimized for semantic segmentation tasks with labeled data specifically for industrial park classification, this dataset serves as a foundational asset for the precise identification and spatial tracking of major air pollution sources. We validated the dataset’s applicability using a modified U-Net model, achieving a mean intersection over union of 0.8146 and pixel accuracy of 0.9608, thereby demonstrating its potential as a tool for detecting and monitoring pollutant sources in industrial areas. With future expansion through additional temporal data and diverse pollutant measurements, this dataset is anticipated to support regional air quality monitoring efforts and inform strategies for pollution control across East Asia.