what approach did you select for the detection of flooded areas? Threshold-based identification or supervised classification?
Unlike open areas, flooded vegetation is often characterized by an increase of backscatter (instead of a decrease), because if higher occurrence of double-bounce effects
Please check this review article, it contains nice explanations on the nature of flooded vegetation: SAR-based detection of flooded vegetation – a review of characteristics and approaches .
Also, the discussions in these topics might be helpful: