ABSTRACT

The future projections of climate change envisage a global increase in extreme precipitation events and subsequent flooding. The reliable and rapid flood maps are the critical parameters in preparing the disaster management plans. This study demonstrated an effective flood mapping framework using freely available multi-temporal Earth Observation (EO) images, including C-band Sentinel-1A & 1B Synthetic Aperture Radar (SAR) images and optical WorldView-3 images, for analyzing the 2018 flood event of Kerala, India. Two Change Detection (CD) techniques, i.e. Ratio Index (RI) and Normalized Change Index (NCI) combined with semi-automatic thresholding are implemented on temporal descending pass SAR images for flood identification. For ascending pass SAR images, the statistical-based thresholding method is implemented. The results indicate that combined use of ascending and descending pass SAR images contributed to a better understanding of flood conditions. It is also inferred that the use of a pre-flood image can enhance flood area estimation and helps in minimizing the overestimation errors. The results also found that NCI outperforms RI for Kerala flood event. Flood area extracted from these techniques is plotted against the Indian Meteorological Department (IMD) rainfall datasets, which showed a similar trend. Field photographs and optical images are used for validation purposes.

摘要

对气候变化的未来预测是，全球范围内极端降水事件和随后的洪水将增加。可靠，快速的洪水图是制定灾难管理计划的关键参数。这项研究展示了使用免费的多时相地球观测（EO）图像（包括C波段Sentinel-1A和1B合成孔径雷达（SAR）图像以及光学WorldView-3图像）来分析2018年洪水事件的有效洪水制图框架印度喀拉拉邦的地图。两种变化检测（CD）技术，即比率指数（RI）和归一化变化指数（NCI）与半自动阈值结合，在时间递减的SAR图像上实现，用于洪水识别。对于上升的SAR图像，实现了基于统计的阈值方法。结果表明，上升和下降通过SAR图像的组合使用有助于更好地了解洪水情况。还可以推断出，使用洪水前的图像可以增强洪水面积的估计，并有助于最大程度地减少过高的估计误差。结果还发现，在喀拉拉邦洪水事件中，NCI优于RI。从这些技术中提取的洪水面积与印度气象局（IMD）降雨数据集相对应，其趋势相似。现场照片和光学图像用于验证目的。结果还发现，在喀拉拉邦洪水事件中，NCI优于RI。从这些技术中提取的洪水面积与印度气象局（IMD）降雨数据集相对应，其趋势相似。现场照片和光学图像用于验证目的。结果还发现，在喀拉拉邦洪水事件中，NCI优于RI。从这些技术中提取的洪水面积与印度气象局（IMD）降雨数据集相对应，其趋势相似。现场照片和光学图像用于验证目的。