Machine Learning-Based Comparative Analysis of Human and Economic Losses from Cold-Climate Mountain Floods: Case Studies of Zarindasht–Firouzkouh (Iran, 2022) and Houghton, Michigan (2018)
Fuente:
Zenodo
Gespeichert in:
| 1. Verfasser: | |
|---|---|
| Format: | Recurso digital |
| Sprache: | Englisch |
| Veröffentlicht: |
Zenodo
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866901899595218944 |
|---|---|
| author | Sohrabi, Mohammad |
| author_facet | Sohrabi, Mohammad |
| contents | <p><span>Core Problem:Cold-climate mountain floods pose increasing threats to human life and economic stability, particularly in regions affected by short-duration monsoonal rainfall. In July 2022, the Zarindasht–Firouzkouh flood in northern Iran caused significant destruction, including 15 fatalities and severe infrastructural losses. Similarly, the Houghton, Michigan flood (2018) inflicted major damage due to extreme rainfall and runoff on steep terrain. Despite different socio-economic contexts, both disasters reveal gaps in early-warning capacity and adaptive planning. Understanding their similarities and differences is crucial for improving resilience in mountainous environments<span dir="RTL" lang="AR-SA">.</span>Approach<span dir="RTL" lang="AR-SA">:</span>This study applies a machine learning–based comparative framework integrating GIS, remote sensing, and socio-economic datasets. Predictive algorithms—Random Forest (RF), Gradient Boosting (GBM), and XGBoost—were trained on hydrological, topographical, and land-use variables to estimate flood severity and potential losses. Spatial and temporal data from Sentinel-2, DEMs, and disaster databases were analyzed to identify critical risk drivers in both regions<span dir="RTL" lang="AR-SA">.</span>Key Findings:Model performance achieved high predictive accuracy (R² = 0.91; RMSE = 0.23). The most influential predictors included precipitation anomaly, slope, vegetation cover, and population density. While physical exposure was higher in Houghton, social vulnerability was substantially greater in Zarindasht–Firouzkouh due to limited preparedness, infrastructure fragility, and lack of insurance coverage.Impact:This comparative analysis highlights the need for AI-enabled disaster governance to enhance adaptive capacity in mountain communities. By merging predictive analytics with socio-hydrological data, the framework supports data-driven policy design and cross-country learning for sustainable flood resilience in cold-climate regions.</span></p> |
| format | Recurso digital |
| id | zenodo_https___doi_org_10_5281_zenodo_17619207 |
| institution | Zenodo |
| language | eng |
| publishDate | 2025 |
| publisher | Zenodo |
| record_format | zenodo |
| spellingShingle | Machine Learning-Based Comparative Analysis of Human and Economic Losses from Cold-Climate Mountain Floods: Case Studies of Zarindasht–Firouzkouh (Iran, 2022) and Houghton, Michigan (2018) Sohrabi, Mohammad Hoghton Michigan Firozkoh Floods Human Vulnerability Machine Learning Machin lerning <p><span>Core Problem:Cold-climate mountain floods pose increasing threats to human life and economic stability, particularly in regions affected by short-duration monsoonal rainfall. In July 2022, the Zarindasht–Firouzkouh flood in northern Iran caused significant destruction, including 15 fatalities and severe infrastructural losses. Similarly, the Houghton, Michigan flood (2018) inflicted major damage due to extreme rainfall and runoff on steep terrain. Despite different socio-economic contexts, both disasters reveal gaps in early-warning capacity and adaptive planning. Understanding their similarities and differences is crucial for improving resilience in mountainous environments<span dir="RTL" lang="AR-SA">.</span>Approach<span dir="RTL" lang="AR-SA">:</span>This study applies a machine learning–based comparative framework integrating GIS, remote sensing, and socio-economic datasets. Predictive algorithms—Random Forest (RF), Gradient Boosting (GBM), and XGBoost—were trained on hydrological, topographical, and land-use variables to estimate flood severity and potential losses. Spatial and temporal data from Sentinel-2, DEMs, and disaster databases were analyzed to identify critical risk drivers in both regions<span dir="RTL" lang="AR-SA">.</span>Key Findings:Model performance achieved high predictive accuracy (R² = 0.91; RMSE = 0.23). The most influential predictors included precipitation anomaly, slope, vegetation cover, and population density. While physical exposure was higher in Houghton, social vulnerability was substantially greater in Zarindasht–Firouzkouh due to limited preparedness, infrastructure fragility, and lack of insurance coverage.Impact:This comparative analysis highlights the need for AI-enabled disaster governance to enhance adaptive capacity in mountain communities. By merging predictive analytics with socio-hydrological data, the framework supports data-driven policy design and cross-country learning for sustainable flood resilience in cold-climate regions.</span></p> |
| title | Machine Learning-Based Comparative Analysis of Human and Economic Losses from Cold-Climate Mountain Floods: Case Studies of Zarindasht–Firouzkouh (Iran, 2022) and Houghton, Michigan (2018) |
| topic | Hoghton Michigan Firozkoh Floods Human Vulnerability Machine Learning Machin lerning |
| url | https://doi.org/10.5281/zenodo.17619207 |