Evidence record · UDINUS · SDGs
- Unit
- Rektorat UDINUS
- Period
- 2026
- Goals
- Record
- e980e489-5bb9-4b2e-a4ea-a677e4012c1a
- Published
- 2026-10-30
Universitas Dian Nuswantoro conducted research on hybrid SAR-optical remote sensing for flood inundation mapping in a wetland environment. The activity involved UDINUS researchers Puspitasari, Desy Ika; Noersasongko, Edi; Purwanto; Soeleman, Moch. Arief; and Supriyanto. The research addressed recurrent flooding in Banjar Regency, where low-lying wetland topography, high rainfall, and river sedimentation form part of the flooding context. Precise flood mapping remained challenging because cloud cover interferes with optical imaging and diverse wetland ecosystems create spectral ambiguity.
The research proposed a hybrid method combining Sentinel-1 backscatter with Sentinel-2 water indices, FWEI and AWEI, through pixel-level feature stacking and feature contribution analysis. Four machine learning classifiers were evaluated: Random Forest, Logistic Regression, Support Vector Machine, and XGBoost. A spatially independent validation scheme was used to support generalization and address autocorrelation bias. Flood inundation maps were derived through change detection between independently classified pre-flood and post-flood water maps.
The findings indicated that AWEI surpassed individual features in water discrimination. The combination of SAR and AWEI consistently enhanced classification stability and spatial coherence. SVM and Random Forest exhibited similar performance, while SVM attained the optimal classification balance for water detection, with OA of 0.98, Kappa of 0.97, and F1 of 0.98. The difference lacked statistical significance. Feature contribution analysis showed that AWEI provided the strongest standalone classification performance, while SAR contributed complementary structural information that improved classification robustness.
The research also found that multi-sensor integration enhanced the delineation of flood-affected areas in complex wetland environments. Mapping reliability depended on the accuracy of pre-flood and post-flood water classification and the temporal consistency of the input imagery. The activity therefore frames UDINUS research around flood inundation analysis, water classification, and the delineation of affected areas in wetland conditions.