Detection of Ice-Jam Floods Using PALSAR Full-Polarimetry Data

Abstract

The Lena River in Russia at its intersection with the Arctic Ocean is subject to spring floods such as those caused by ice jams. In such cases, snowmelt occurs at lower latitudes and river water freezes at higher latitudes. That snowmelt is blocked by the frozen river water, which induces ice-jam floods every year. Full-polarimetric parameters obtained from the Phased Array Type L-band Synthetic Aperture Radar (PALSAR) were used to determine the most suitable ones for detecting ice-jam floods. Representative full-polarimetric parameters were calculated, including sigma(0), entropy/alpha/anisotropy, the four-component decomposition model, and coherence. The PALSAR mosaic image helped to identify two large-scale floods, at Ytyk Kiuel on May 18, 2007, and at Berdigestiakh on July 15, 2007. These flood events were confirmed by articles published in the local newspaper and by optical sensor images. The optimum detectability for the flooding area was derived by the difference of polarimetric coherence, gamma(HH + VV) - (HH - VV), obtained with and without the consideration of flooding. Total accuracy was 92.9\%, and user and producer accuracies for detection of the flooded areas were 53.5\% and 72.8\%, respectively. The four-component decomposition analysis indicated that flood detection using gamma(HH + VV) - (HH - VV) is applicable to low-height vegetation areas and urban regions.