Optimization of threshold ranges for rapid flood inundation mapping by evaluating backscatter profiles of high incidence angle SAR images

Abstract

Rapid satellite-based flood inundation mapping and delivery of flood inundation maps during a flood event can provide crucial information for planners and decision makers to prioritize relief and rescue operations. The present study is undertaken to optimize the threshold ranges for the classification of flood water in Synthetic Aperture Radar (SAR) images (of 20° to 49° incidence angles) for quick flood inundation mapping and response during flood disasters. This is done through assessing the signature of flood water in Horizontal transmit and Horizontal received (HH), Horizontal transmit and Vertical received (HV), Vertical transmit and Horizontal received (VH), and Vertical transmit and Vertical received (VV) polarization radar data. The mean backscattering signature profiles of various water bodies were analyzed to discriminate flood water from other water bodies. The study shows that there is better demarcation of land-water surface in HH polarization. VV polarization has the potential to identify partially submerged features, which can be useful in flood damage assessments. The backscatter of flood water in HV and VH is the same and both HV and VH polarizations are adequate for the mapping of flood water. At near range to far range, −8 to −12 dB, −15 to −24 dB, and −6 to −15 dB can be used as optimum ranges for the classification of flood water in HH, HV, and VV polarizations. These optimum threshold ranges can be applied to the automation of flood mapping using SAR images in near-real time, where much time was often spent on finding the thresholds in order to produce flood inundation maps in a short time from the onset of flood disasters and deliver such maps to the concerned agencies.

Operational Surface Water Detection and Monitoring Using Radarsat 2

“…Modes that have large extents (beyond 70 km) are available in one or two polarizations (HH and HV or VV and VH), due to the storage and downlink capacity of the satellite, while smaller swaths can provide all four polarizations. Previous studies [5,20] had shown that HH is the preferred polarization for mapping flooded vegetation, because it maximizes canopy penetration and enhances the contrast between forest and flooded vegetation. Also, HH has lower backscatter over rough open water (roughness induced by wind) than VV, and therefore is better to map open water under windy conditions.…”

“…Due to its all-weather capabilities, and its image acquisition capacity during day or night or in cloudy conditions, SAR imagery offers better alternatives for water mapping than optical imagery [5]. As a result of the radar's unique response to water, water mapping using intensity thresholding methods on SAR image had been extensively used [4][5][6][7]20,21]. In this method, water is separated from land in intensity images, but the accuracy of the results relies on the ability to differentiate land vs. water pixels in the intensity domain, which becomes especially difficult when the water backscatter is affected by wind-induced roughness.…”

Detection of inundation areas due to the 2015 Kanto and Tohoku torrential rain in Japan based on multi-temporal ALOS-2 imagery

“…Thus, SAR images are effective for extracting inundation areas. Several methods, both pixel-and object-based, have been proposed to extract inundation zones from SAR images (Martinis, 2009 and2015;Hoque et al, 2011;Manjusree et al, 2012;Pulvirenti et al, 2014;Kundu et al, 2015;Nakmuenwai et al, 2017). Thresholding is a 5 common and effective pixel-based approach.…”

“…Thresholding is a 5 common and effective pixel-based approach. It is difficult to judge the most suitable value objectively in manual thresholding; therefore, automated thresholding methods have been introduced to overcome this issue (Fan and Lei, 2009;Martinis, 2009 and2015;Manjusree et al, 2012). In this study, the threshold value for water was investigated automatically using reference areas.…”

“…Different from other land covers, the backscattering intensity of urban areas still showed high values after inundation owing to multiple reflections from buildings and water surfaces (Mason, 2010 and2012). Thus, it is difficult to extract the inundated urban areas using the proposed simple water threshold values.…”