Minwoo Roh
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Minwoo Roh has a total of 22 co-authors. Here are the top co-authors by number of shared publications:
- Woo-Kyun Lee
Department of Environmental Science and Ecological Engineering, Korea University, Seoul 02841, Republic of Korea
4 shared publications - Cholho Song
OJEong Resilience Institute (OJERI), Korea University, Seoul 02841, Republic of Korea
2 shared publications - Joon Kim
OJEong Resilience Institute (OJERI), Korea University, Seoul 02841, Republic of Korea
2 shared publications - Hojeong Park
Korea University
1 shared publication - Hyun-Woo Jo
OJEong Resilience Institute (OJERI), Korea University, Seoul 02841, Republic of Korea
1 shared publication - Jeman Lee1 shared publication
Latest Publications
- Open Access
- 17 September 2026
- Published by Elsevier in SSRN Electronic Journal
Spatial heterogeneity of anthropogenic effects on winter wildfire risk in South Korea: A SHAP-based XGBoost analysis with spatial block cross-validation
Winter wildfires in South Korea have intensified markedly over the past decade, yet operational fire danger rating systems rely exclusively on meteorological inputs and cannot detect the anthropogenic ignition conditions that drive the majority of fire events. This study develops an XGBoost-based binary classification model integrating 15 meteorological, topographic, vegetation, and human activity predictors at 100 m resolution, evaluated using quantile-based spatial block cross-validation (5-fold) to ensure spatially honest performance estimation. Firewood boiler kernel density — restricted to households within 100 m of forest boundaries (n = 29,399) — was incorporated as a spatially explicit ignition-source proxy alongside four proximity-based anthropogenic variables. SHapley Additive exPlanations (SHAP) decompose predicted fire risk into per-variable contributions, enabling systematic examination of (i) whether human activity variables explain a distinct predicted risk dimension that co-varies only weakly with meteorological and topographic predictors, (ii) whether the effect of firewood boiler density is spatially uniform or locally concentrated, and (iii) how the full anthropogenic-integrated model compares to a meteorology-topography-forest-type baseline. The full model achieved a mean spatial block-validated AUC of 0.751, compared to 0.696 for the 7-variable baseline (ΔAUC = +0.054; paired t-test p = 0.001). Cross-group SHAP correlations between climate and human activity variables averaged r = 0.126, indicating that anthropogenic predictors explain a distinct predicted risk dimension that co-varies only weakly with meteorological inputs. Firewood boiler density ranked 13th nationally in SHAP importance but rose to 2nd within the highest-density stratum (≥ 2.0 km⁻²), with positive-SHAP points (n = 381) exhibiting a fire occurrence rate of 55.4% — approximately 2.2 times the dataset mean. Positive boiler SHAP contributions were confined exclusively to the ≥ 2.0 km⁻² stratum under low-temperature conditions (TEMP ≤ 7.7°C), consistent with thermally gated peak heating-season operation. These findings demonstrate that anthropogenic ignition proxies explain fire risk dimensions beyond meteorological indicators, and that spatially heterogeneous effects of individual variables necessitate place-based rather than uniform fire prevention strategies.
- 17 September 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessSpatial heterogeneity of anthropogenic effects on winter wildfire risk in South Korea: A SHAP-based XGBoost analysis with spatial block cross-validation
Winter wildfires in South Korea have intensified markedly over the past decade, yet operational fire danger rating systems rely exclusively on meteorological inputs and cannot detect the anthropogenic ignition conditions that drive the majority of fire events. This study develops an XGBoost-based binary classification model integrating 15 meteorological, topographic, vegetation, and human activity predictors at 100 m resolution, evaluated using quantile-based spatial block cross-validation (5-fold) to ensure spatially honest performance estimation. Firewood boiler kernel density — restricted to households within 100 m of forest boundaries (n = 29,399) — was incorporated as a spatially explicit ignition-source proxy alongside four proximity-based anthropogenic variables. SHapley Additive exPlanations (SHAP) decompose predicted fire risk into per-variable contributions, enabling systematic examination of (i) whether human activity variables explain a distinct predicted risk dimension that co-varies only weakly with meteorological and topographic predictors, (ii) whether the effect of firewood boiler density is spatially uniform or locally concentrated, and (iii) how the full anthropogenic-integrated model compares to a meteorology-topography-forest-type baseline. The full model achieved a mean spatial block-validated AUC of 0.751, compared to 0.696 for the 7-variable baseline (ΔAUC = +0.054; paired t-test p = 0.001). Cross-group SHAP correlations between climate and human activity variables averaged r = 0.126, indicating that anthropogenic predictors explain a distinct predicted risk dimension that co-varies only weakly with meteorological inputs. Firewood boiler density ranked 13th nationally in SHAP importance but rose to 2nd within the highest-density stratum (≥ 2.0 km⁻²), with positive-SHAP points (n = 381) exhibiting a fire occurrence rate of 55.4% — approximately 2.2 times the dataset mean. Positive boiler SHAP contributions were confined exclusively to the ≥ 2.0 km⁻² stratum under low-temperature conditions (TEMP ≤ 7.7°C), consistent with thermally gated peak heating-season operation. These findings demonstrate that anthropogenic ignition proxies explain fire risk dimensions beyond meteorological indicators, and that spatially heterogeneous effects of individual variables necessitate place-based rather than uniform fire prevention strategies.
- Open Access
- 16 April 2025
- Published by Springer Nature in Landslides
- Vol. 22 (8) , 2809-2827
Machine learning-based rainfall-induced landslide susceptibility model and short-term early warning assessment in South Korea
The extreme rainfall events associated with climate change trigger landslides. Approximately 60% of South Korea comprises mountainous terrain, with steep slopes rendering it particularly prone to landslides. Despite the implementation of early warning systems by the Korea Forest Service (KFS), landslide damage remains substantial, with approximately 2345 hectares affected over the past 5 years, resulting in severe human and economic costs. The current 24-h early warning system, based on Tier 3 administrative division (Town), faces challenges in accurately identifying high-susceptibility landslide areas. Thus, a daily landslide susceptibility model that integrates landslide-associated conditioning factors with meteorological, topographic, and environmental data was designed to assess landslide susceptibility with a spatial resolution of 100 m. Using AutoML, we identified Random Forest as the optimal model for predicting landslide susceptibility. Training the model with landslide data from 2016 to 2022 resulted in an accuracy of 0.93, AUC of 0.98, and F- 1 score of 0.98. A kappa value of 0.85 indicated the effective classification of past landslides using testing data. Location-based validation using 2023 occurrences revealed highly susceptible classifications for 88% of 43 landslides, while spatial scale-based hazard assessment using observed data indicated high hazard for 96% of 607 landslides in Tiers 3 and 4 (Township). Weather forecasting was also found to affect accuracy, with 76% accuracy for forecasts made at 5:00 PM and 41% for forecasts made at 8:00 AM. It was confirmed that further calibration of forecasting data can enhance the performance of the susceptibility model. The designed process thus enhances landslide prevention and preparedness on both local and regional scales, offering a crucial tool for mitigating the impact of landslides in South Korea.
- 16 April 2025
- Published by Springer Nature in Landslides
- Vol. 22 (8) , 2809-2827
Open AccessMachine learning-based rainfall-induced landslide susceptibility model and short-term early warning assessment in South Korea
The extreme rainfall events associated with climate change trigger landslides. Approximately 60% of South Korea comprises mountainous terrain, with steep slopes rendering it particularly prone to landslides. Despite the implementation of early warning systems by the Korea Forest Service (KFS), landslide damage remains substantial, with approximately 2345 hectares affected over the past 5 years, resulting in severe human and economic costs. The current 24-h early warning system, based on Tier 3 administrative division (Town), faces challenges in accurately identifying high-susceptibility landslide areas. Thus, a daily landslide susceptibility model that integrates landslide-associated conditioning factors with meteorological, topographic, and environmental data was designed to assess landslide susceptibility with a spatial resolution of 100 m. Using AutoML, we identified Random Forest as the optimal model for predicting landslide susceptibility. Training the model with landslide data from 2016 to 2022 resulted in an accuracy of 0.93, AUC of 0.98, and F- 1 score of 0.98. A kappa value of 0.85 indicated the effective classification of past landslides using testing data. Location-based validation using 2023 occurrences revealed highly susceptible classifications for 88% of 43 landslides, while spatial scale-based hazard assessment using observed data indicated high hazard for 96% of 607 landslides in Tiers 3 and 4 (Township). Weather forecasting was also found to affect accuracy, with 76% accuracy for forecasts made at 5:00 PM and 41% for forecasts made at 8:00 AM. It was confirmed that further calibration of forecasting data can enhance the performance of the susceptibility model. The designed process thus enhances landslide prevention and preparedness on both local and regional scales, offering a crucial tool for mitigating the impact of landslides in South Korea.
- Open Access
- 18 March 2025
- Published by Copernicus GmbH
Spatial Prediction of the Future Landslide Susceptibility under the SSP Scenario Using Machine learning Algorithms
The increase in localized heavy rainfall and intense storms due to climate change has led to a continuous rise in landslide damages in South Korea, including slope failures and debris flows. While post-landslide recovery and damage site assessments are crucial, it is equally important to develop proactive and systematic landslide adaptation strategies to predict and prepare for landslides in advance. This study aims to develop an interpretable machine learning-based landslide susceptibility model and analyze landslide-prone areas under future climate change scenarios. Through this approach, it seeks to clearly identify the impact of forest management factors on landslides and establish effective adaptation strategies tailored to climate change scenarios. A dataset comprising 6,517 recorded landslide events from 2011 to 2024 was utilized. Various external and internal conditioning factors were obtained and constructed with a resolution of 100 meters. Climate scenario analysis employed SSP 1-2.6, SSP 2-4.5, and SSP 5-8.5, with extreme climate factors including rainfall intensity, the number of heavy rain days, daily rainfall, and 5day cumulative rainfall. Notably, changes in stand age class, DBH class, and growing stock under future forest management scenarios were calculated and integrated into the landslide model, enabling an evaluation of how management strategies affect landslide susceptibility. Results were validated by comparing past actual occurrence data. The SSP 5-8.5 scenario indicates a significant increase in landslide occurrences. These findings provide valuable insights into the effects of climate change on landslide susceptibility in South Korea and examine the potential of future landslide management strategies to mitigate associated susceptibility.
- 18 March 2025
- Published by Copernicus GmbH
Open AccessSpatial Prediction of the Future Landslide Susceptibility under the SSP Scenario Using Machine learning Algorithms
The increase in localized heavy rainfall and intense storms due to climate change has led to a continuous rise in landslide damages in South Korea, including slope failures and debris flows. While post-landslide recovery and damage site assessments are crucial, it is equally important to develop proactive and systematic landslide adaptation strategies to predict and prepare for landslides in advance. This study aims to develop an interpretable machine learning-based landslide susceptibility model and analyze landslide-prone areas under future climate change scenarios. Through this approach, it seeks to clearly identify the impact of forest management factors on landslides and establish effective adaptation strategies tailored to climate change scenarios. A dataset comprising 6,517 recorded landslide events from 2011 to 2024 was utilized. Various external and internal conditioning factors were obtained and constructed with a resolution of 100 meters. Climate scenario analysis employed SSP 1-2.6, SSP 2-4.5, and SSP 5-8.5, with extreme climate factors including rainfall intensity, the number of heavy rain days, daily rainfall, and 5day cumulative rainfall. Notably, changes in stand age class, DBH class, and growing stock under future forest management scenarios were calculated and integrated into the landslide model, enabling an evaluation of how management strategies affect landslide susceptibility. Results were validated by comparing past actual occurrence data. The SSP 5-8.5 scenario indicates a significant increase in landslide occurrences. These findings provide valuable insights into the effects of climate change on landslide susceptibility in South Korea and examine the potential of future landslide management strategies to mitigate associated susceptibility.
- 31 August 2023
- Published by The Korean Society of Climate Change Research in Journal of Climate Change Research
- Vol. 14 (4) , 385-393
Prediction of forest fire risk according to climate change in Bhutan using a shared socioeconomic pathways (SSP) scenario and random forest
Forest fires destroy millions of forest acres globally, damaging ecosystem services, emitting carbon into the atmosphere, and causing billions of dollars in socio-economic losses, including loss of life. Bhutan, despite its small size, is grappling with forest fires compounded by climate extremes, a remote location, and a lack of preparedness. Under such conditions, it is important to develop policy, technology, and action to respond to forest fires based on scientific research. This study used the Random Forest model (RF) and Shared Socioeconomic Pathways (SSP) approach to create a model for predicting forest fires in Bhutan by considering climate data, spatial data, and forest fire occurrence factors. The RF algorithm was used to analyze the correlations between various input data and predict the change in future forest fire risk in Bhutan. Using SSP scenarios that considered changes due to future population and economic growth, it was possible to predict the risk of forest fires accurately and spatially in Bhutan. The study found that the average forest fire risk was highest during the months of October to January, which are also the dry seasons with low precipitation. However, it also showed high variability across climate change scenarios. SSP scenarios also confirmed the possibility of a significant increase in forest fire risk in the future. The results of this study can be used to support specific policy decisions for forest fire prevention and response in Bhutan and to contribute to the development of forest fire prediction technology in the light of the changing global climate.
- 31 August 2023
- Published by The Korean Society of Climate Change Research in Journal of Climate Change Research
- Vol. 14 (4) , 385-393
Prediction of forest fire risk according to climate change in Bhutan using a shared socioeconomic pathways (SSP) scenario and random forest
Forest fires destroy millions of forest acres globally, damaging ecosystem services, emitting carbon into the atmosphere, and causing billions of dollars in socio-economic losses, including loss of life. Bhutan, despite its small size, is grappling with forest fires compounded by climate extremes, a remote location, and a lack of preparedness. Under such conditions, it is important to develop policy, technology, and action to respond to forest fires based on scientific research. This study used the Random Forest model (RF) and Shared Socioeconomic Pathways (SSP) approach to create a model for predicting forest fires in Bhutan by considering climate data, spatial data, and forest fire occurrence factors. The RF algorithm was used to analyze the correlations between various input data and predict the change in future forest fire risk in Bhutan. Using SSP scenarios that considered changes due to future population and economic growth, it was possible to predict the risk of forest fires accurately and spatially in Bhutan. The study found that the average forest fire risk was highest during the months of October to January, which are also the dry seasons with low precipitation. However, it also showed high variability across climate change scenarios. SSP scenarios also confirmed the possibility of a significant increase in forest fire risk in the future. The results of this study can be used to support specific policy decisions for forest fire prevention and response in Bhutan and to contribute to the development of forest fire prediction technology in the light of the changing global climate.
- 31 October 2022
- Published by The Korean Society of Climate Change Research in Journal of Climate Change Research
- Vol. 13 (5) , 545-555
Current State of the A/R CDM Project in Goseong-gun, Gangwon Province, South Korea for Transfer to SDM and Ongoing Challenges
In this study, A/R CDM project registered with the UNFCCC in 2012 was reviewed for transfer to SDM. Additionally, economic analysis was performed. The amount of CO₂ stock for 2021 was 44.15 tCO₂/yr at the project site. As a result of analysis, the total stock through field survey was higher than PDD. In the case of tree species, the stock of Pinus koraiensis was slightly lower than others. It means that growth of Pinus koraiensis is slow in the site. On the other hand, as Pinus koraiensis occupies the largest area of the project, it is necessary to improve the stock by preparing an optimal management plans. On economic analysis of the entire business period for securing carbon absorption reductions, the B/C ratio was less than 1, indicating that there was no business feasibility. However, considering the social environment contribution, improving image of the state and companies participating in social contribution activities such as afforestation, and educational effects, it can obtain investment feasibility of project, continuously. In order to maintain the project, transparency must be achieved through comparison with PDD. It was possible to confirm the difference between the planting period, the area, and the predicted amount of carbon absorption reduction. It appears possible to achieve transparency through PRC along with the SDM transition.
- 31 October 2022
- Published by The Korean Society of Climate Change Research in Journal of Climate Change Research
- Vol. 13 (5) , 545-555
Current State of the A/R CDM Project in Goseong-gun, Gangwon Province, South Korea for Transfer to SDM and Ongoing Challenges
In this study, A/R CDM project registered with the UNFCCC in 2012 was reviewed for transfer to SDM. Additionally, economic analysis was performed. The amount of CO₂ stock for 2021 was 44.15 tCO₂/yr at the project site. As a result of analysis, the total stock through field survey was higher than PDD. In the case of tree species, the stock of Pinus koraiensis was slightly lower than others. It means that growth of Pinus koraiensis is slow in the site. On the other hand, as Pinus koraiensis occupies the largest area of the project, it is necessary to improve the stock by preparing an optimal management plans. On economic analysis of the entire business period for securing carbon absorption reductions, the B/C ratio was less than 1, indicating that there was no business feasibility. However, considering the social environment contribution, improving image of the state and companies participating in social contribution activities such as afforestation, and educational effects, it can obtain investment feasibility of project, continuously. In order to maintain the project, transparency must be achieved through comparison with PDD. It was possible to confirm the difference between the planting period, the area, and the predicted amount of carbon absorption reduction. It appears possible to achieve transparency through PRC along with the SDM transition.
- 31 October 2020
- Published by The Korean Society of Climate Change Research in Journal of Climate Change Research
- Vol. 11 (5-1) , 329-342
Status of Spatial Data Construction for Climate Change Mitigation and Adaptation in Central Asia
The importance of spatial data is rising as a means to reduce vulnerabilities and risks associated with climate change. Considering the recognition of its importance, many developing countries are struggling with capacity building for spatial data construction. The key objective of this study is to understand the current status and challenges of spatial data construction in Central Asian countries with a focus on enhancing the use of spatial data in Central Asian countries through international cooperation. The target countries are Azerbaijan, Kazakhstan, Kyrgyzstan, Tajikistan, Turkmenistan and Uzbekistan. This research used the official data provided on the websites of cartography agencies in Central Asia countries as well as data presented at regional workshops on the integration of statistical and spatial data in Central Asia, which was conducted by the United Nations Economic and Social Commission for Asia and the Pacific (UNESCAP). The collected data are classified into three types: fundamental maps, thematic maps and NSDI (National Spatial Data Infrastructure). Through this classification, analyses of the resolution level and the data-sharing system were conducted. The results show that each country exhibits differences in spatial information capabilities. In the case of Kazakhstan and Uzbekistan, satellite data produced by the aerospace agency are actively used for land management, focusing on the ecosystem. However, other countries mainly rely on statistical data to manage the land, and it is found that spatial data are not actively utilized. In addition, Kazakhstan and Uzbekistan have implemented the NSDI over the past years, which facilitates the sharing of spatial data. However, the rest of the countries in Central Asia show limitations in terms of sharing geospatial data, and additionally lack data infrastructure. This paper reveals the differences in national competency when using and managing geospatial data among Central Asian countries. The results of this analysis will contribute to building future policies and actions, acting as a reference for improving the national capacity, relevant infrastructure, and regional-international cooperation.
- 31 October 2020
- Published by The Korean Society of Climate Change Research in Journal of Climate Change Research
- Vol. 11 (5-1) , 329-342
Status of Spatial Data Construction for Climate Change Mitigation and Adaptation in Central Asia
The importance of spatial data is rising as a means to reduce vulnerabilities and risks associated with climate change. Considering the recognition of its importance, many developing countries are struggling with capacity building for spatial data construction. The key objective of this study is to understand the current status and challenges of spatial data construction in Central Asian countries with a focus on enhancing the use of spatial data in Central Asian countries through international cooperation. The target countries are Azerbaijan, Kazakhstan, Kyrgyzstan, Tajikistan, Turkmenistan and Uzbekistan. This research used the official data provided on the websites of cartography agencies in Central Asia countries as well as data presented at regional workshops on the integration of statistical and spatial data in Central Asia, which was conducted by the United Nations Economic and Social Commission for Asia and the Pacific (UNESCAP). The collected data are classified into three types: fundamental maps, thematic maps and NSDI (National Spatial Data Infrastructure). Through this classification, analyses of the resolution level and the data-sharing system were conducted. The results show that each country exhibits differences in spatial information capabilities. In the case of Kazakhstan and Uzbekistan, satellite data produced by the aerospace agency are actively used for land management, focusing on the ecosystem. However, other countries mainly rely on statistical data to manage the land, and it is found that spatial data are not actively utilized. In addition, Kazakhstan and Uzbekistan have implemented the NSDI over the past years, which facilitates the sharing of spatial data. However, the rest of the countries in Central Asia show limitations in terms of sharing geospatial data, and additionally lack data infrastructure. This paper reveals the differences in national competency when using and managing geospatial data among Central Asian countries. The results of this analysis will contribute to building future policies and actions, acting as a reference for improving the national capacity, relevant infrastructure, and regional-international cooperation.