Identification of flood events using Sentinel 2 and Landsat 8 data on the Google Earth Engine in Hyderabad City

Abstract:

Inundation of dry areas is a result of complex interactions between atmosphere and land, and assessment of its extent is essential for risk studies. The remote sensing data is key in mapping the extent of inundation for data-scarce regions. The web-based computing platforms like the Google Earth Engine (GEE) is a great asset in understanding the city system’s response in the context of hydrology. The chief objective of this study is to estimate the area or the part of the area, which is inundated during flooding with the aid of GEE. Landsat-8 images at 30m resolution and Sentinel-2 at 10m resolution are used for this study. Here, we present an algorithm that utilizes all existing Sentinel-2 images in combination with Landsat-8 that are accessible on the GEE to identify flood events using three water extraction indices. The three water indices are mNDWI, AWEInsh, AWEIsh used to extract percentage of water area (PWA). The period of the record for a period of 2013 April to 2019 December is considered. Statistical techniques using rainfall and cloud cover as explanatory variables are being pursued in the analysis. The Google Earth Engine platform allows in saving time by avoiding downloading and preprocessing the data, and the flexibility of the proposed framework allows the development of flood databases in the data-scarce regions which can be used in qualitative understanding of flood mechanisms.