Enhancing Flood Detection in Surabaya: A Comparative Study of VV and VH Polarizations with Sentinel-1 Data

Abstract

Flood mapping is critical to strengthen urban resilience, particularly in Surabaya,
where flooding is a major and recurring threat. Sentinel-1 satellite data offers
significant advantages for flood model calibration due to its high-resolution imagery
and frequent revisits. This study utilizes Google Earth Engine to process and analyse
Sentinel-1 data for mapping flood extents using two different polarizations: VV and
VH. The research compares the capabilities of these polarizations in detecting flood
areas. The results show that VV polarization consistently identifies a larger flood area
compared to VH polarization under similar processing conditions. However, the
Kappa coefficient was used to assess classification accuracy, with VV achieving a
Kappa of 0.8 and VH reaching a higher Kappa of 0.92, reflecting better classification
performance. These findings suggest that while VV provides a broader flood
detection, VH offers more reliable flood mapping, highlighting the trade-offs between
sensitivity and accuracy in flood monitoring using Sentinel-1 satellite.