Preprint of a paper that is submitted to the “Journal of Selected Topics in Earth Observation and Remote sensing” with the following abstract:

It is widely assumed that C-Band Synthetic Aperture Radar (SAR) signal do not reach the forest floor in dense forests, and that hence C-Band SAR cannot be used for sub-canopy flood mapping in tropical forests. Indeed, flooded and non-flooded forests are not distinguishable in single C-Band acquisitions. The question is whether long-term seasonal dynamics in C- Band SAR time series data encode flooding dynamics under- neath the canopy. In this paper we investigate the relationship of Sentinel-1 backscatter with sub-canopy flooding in the Amazon rain forest. We use the Empirical Mode Decomposition to extract annual modes in the backscatter signal and use the correlation to the water level of a nearby river to understand the dynamics of the Sentinel-1 signal. Clusters of these correlations coincide well with known forest flooding areas in the VH signal. The analysis shows that the Sentinel-1 C-Band backscatter is influenced by floodings underneath the canopy in the Amazon rain forest. The presented approach could allow to systematically map flooded areas throughout tropical rain forests.