A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Assessment of surface water detection using Sentinel-1 SAR data: Case study Vojvodina province
This study assesses the utility of Sentinel-1 Synthetic Aperture Radar (SAR) data for surface water detection in Vojvodina province. Using multi-temporal SAR imagery from January 2022 to April 2024, machine learning classifiers including Random Forest, KDTree KNN, and Maximum Likelihood were employed to classify water bodies and non-water areas. Polarized indices derived from Sentinel-1 data, such as the Polarized Ratio, Normalized Difference Polarized Index, and Dual-Polarized Water Index, were utilized to enhance water body detection. Despite challenges in accurately identifying narrow canals, the study achieves a notable overall accuracy of 92.68% with Random Forest, 92.08% with KDTree KNN, and 91.58% with Maximum Likelihood for water classification. Producer accuracy for the water class ranges from 87.75% to 89.65%, while User's accuracy exceeds 96.50% across all classifiers. The calculated Cohen's Kappa coefficients of 0.83 to 0.85 indicate substantial agreement between predicted and reference data, underscoring the effectiveness of Sentinel-1 SAR data in surface water detection. However, spatial resolution limitations present ongoing challenges, particularly in accurately delineating narrow water features like canals. Future research directions include refining algorithms to enhance classification accuracy and addressing these challenges in diverse environmental contexts.
Assessment of surface water detection using Sentinel-1 SAR data: Case study Vojvodina province
This study assesses the utility of Sentinel-1 Synthetic Aperture Radar (SAR) data for surface water detection in Vojvodina province. Using multi-temporal SAR imagery from January 2022 to April 2024, machine learning classifiers including Random Forest, KDTree KNN, and Maximum Likelihood were employed to classify water bodies and non-water areas. Polarized indices derived from Sentinel-1 data, such as the Polarized Ratio, Normalized Difference Polarized Index, and Dual-Polarized Water Index, were utilized to enhance water body detection. Despite challenges in accurately identifying narrow canals, the study achieves a notable overall accuracy of 92.68% with Random Forest, 92.08% with KDTree KNN, and 91.58% with Maximum Likelihood for water classification. Producer accuracy for the water class ranges from 87.75% to 89.65%, while User's accuracy exceeds 96.50% across all classifiers. The calculated Cohen's Kappa coefficients of 0.83 to 0.85 indicate substantial agreement between predicted and reference data, underscoring the effectiveness of Sentinel-1 SAR data in surface water detection. However, spatial resolution limitations present ongoing challenges, particularly in accurately delineating narrow water features like canals. Future research directions include refining algorithms to enhance classification accuracy and addressing these challenges in diverse environmental contexts.
Assessment of surface water detection using Sentinel-1 SAR data: Case study Vojvodina province
Nikolić Ratko R. (author) / Bošković Vladan B. (author)
2024
Article (Journal)
Electronic Resource
Unknown
Metadata by DOAJ is licensed under CC BY-SA 1.0
Irrigation in Vojvodina - water quality problems
| British Library Conference Proceedings | 2002
Importance of the Ramsar Site Labudovo Okno for Sustainable Tourism—Evidence from Vojvodina Province
| DOAJ | 2024
| DOAJ | 2020