Al-Mulla et al. (2025) AI Driven Impact Assessment of Shaheen Tropical Cyclone Using Very High-Resolution Satellite Data

Identification

• Journal: Earth Systems and Environment
• Year: 2025
• Date: 2025-11-20
• Authors: Yaseen Al-Mulla, Mohammed Al–Muqaimi, Ahsan Ali, Farid Melgani, Krishna Parimi, Talal Al-Wahaibi
• DOI: 10.1007/s41748-025-00938-y

Research Groups

• Remote Sensing and GIS Research Center, Sultan Qaboos University, Oman
• Department of Soils, Water and Agricultural Engineering, Sultan Qaboos University, Oman
• Department of Information Engineering and Computer Science, University of Trento, Italy

Short Summary

This study quantifies the impact of Tropical Cyclone Shaheen (STC) in Oman using very high-resolution satellite imagery and multiple deep learning models, revealing significant damage to vegetation and buildings, alongside drastic increases in water bodies. It provides detailed, object-specific damage assessments and identifies flood-prone zones to support climate resilience strategies.

Objective

• To train and apply multiple deep learning models using very high-resolution satellite imagery to assess the impact of Tropical Cyclone Shaheen (STC) landfall in Oman.
• To conduct a pre-impact and post-impact assessment of STC on vegetation, buildings, and water bodies in the affected wilayats of Al-Khabourah and Al-Suwaiq.

Study Configuration

• Spatial Scale: Two wilayats (Al-Khabourah and Al-Suwaiq) in the North Al-Batinah Governorate, Oman, analyzed using very high-resolution satellite imagery (41 cm and 50 cm).
• Temporal Scale: Pre-impact imagery from 2021 (e.g., May 11, 2021) and post-impact imagery acquired 3-8 days after the STC landfall (October 7-12, 2021). Rainfall data covered October 2-4, 2021.

Methodology and Data

• Models used:
  • Deep Learning: Single Shot Detector (SSD), Mask R-CNN, YOLOv3 (You Only Look Once v3)
  • Geospatial Analysis: ArcGIS Pro (Version 3.2), ENVI 5.0
  • Indices: Normalized Difference Water Index (NDWI), Normalized Difference Vegetation Index (NDVI), Terrain Ruggedness Index (R)
  • Statistical Analysis: Hot Spot/Cold Spot Analysis (Getis-Ord Gi*), Confusion Matrix, Support Vector Machine (for accuracy assessment)
• Data sources:
  • Satellite Imagery: KOMPSAT-3 (50 cm resolution), GeoEye-1 (41 cm resolution)
  • Ancillary Data: Digital Elevation Models (DEMs) derived from Google Earth Pro, rainfall data (Ministry of Agriculture, Fisheries and Water Resources, Oman), Google Earth Pro archives (for ground truthing), MODIS Terra satellite (for STC track imagery).

Main Results

• The deep learning model achieved an overall accuracy of 94% with a kappa value of 0.92 for land cover classification.
• In Al-Khabourah, 85.8 hectares (51.3%) of dense vegetation and 153.2 hectares (18.9%) of sparse vegetation were destroyed. Water surface areas increased by 496.82 hectares.
• In Al-Suwaiq, 3160 hectares (55.5%) of shrubs and grasses, 1713.9 hectares (67.2%) of dense vegetation, and 609.8 hectares (16.6%) of sparse vegetation were destroyed. Water surface areas increased by 180.8 hectares.
• Building damage in Al-Khabourah affected 42.73 hectares (32%) of structures, with 6.4 hectares (67.5%) of fishermen’s coastal structures destroyed.
• Al-Khabourah experienced a maximum cumulative rainfall of 369 mm, followed by Al-Suwaiq with 300 mm.
• Digital Elevation Models and Terrain Ruggedness Index maps, combined with hotspot analysis, identified flood-prone zones, particularly in coastal areas, with 7.1 square kilometers in Al-Khabourah and 0.61 square kilometers in Al-Suwaiq identified as high-risk hotspots (99% confidence).

Contributions

• This study uniquely applies multiple deep learning models (SSD, Mask R-CNN, YOLOv3) with very high-resolution satellite imagery (41 cm) for a detailed, object-specific impact assessment of a tropical cyclone in the Arabian Peninsula.
• It provides a comprehensive quantification of damage to vegetation, buildings, and changes in water bodies, offering a more granular analysis than previous studies using lower-resolution data.
• The integration of AI with terrain analysis (DEM, Ruggedness Index, Hotspot analysis) effectively identifies and maps vulnerable flood-prone zones, enhancing predictive capabilities for disaster management.
• The findings offer valuable insights for government agencies and landowners to develop more effective pre-impact and post-impact planning, climate resilience strategies, and to anticipate similar tropical cyclones in unexpected regions.

Funding

• His Majesty’s strategic fund at Sultan Qaboos University.
• SQU grant with the code SR/DVC/GISC/20/01.

Citation

@article{AlMulla2025AI,
  author = {Al-Mulla, Yaseen and Al–Muqaimi, Mohammed and Ali, Ahsan and Melgani, Farid and Parimi, Krishna and Al-Wahaibi, Talal},
  title = {AI Driven Impact Assessment of Shaheen Tropical Cyclone Using Very High-Resolution Satellite Data},
  journal = {Earth Systems and Environment},
  year = {2025},
  doi = {10.1007/s41748-025-00938-y},
  url = {https://doi.org/10.1007/s41748-025-00938-y}
}