Principal component analysis of C-SAR images for flood mapping – Santa Fé province, Argentina

Abstract

Flood events are phenomena associated with heavy rainfall. In Argentina, floods have high economic and social costs, including loss of human life. In this paper, principal component analysis (PCA) is used to map flood-prone areas along the Paraná river in Santa Fe, Argentina. The Sentinel-1B (S1B) images, sensor C-SAR with VH polarisation Interferometric type (IW) Ground Range Detected (GRD) with spatial resolution of 10 m, from 2016, were referenced and the PCA method was used to extract the four first principal components. The flood-affected images make it possible to accurately define the flooded area. In targets with dense vegetation, however, there is no pixel backscatter pattern. PC2 better highlighted the threshold of pixel intensity, with an accuracy of 70%, and 93% of the mapped area was shown to be flood-prone. Procedures to map floods remotely are pivotal because they can quickly obtain precise data on flood areas that may not be accessible for fieldwork or that have not yet been mapped in great detail.

Keywords

• Sentinel-1
• time series analysis
• principal components
• Argentina.