High-resolution inundation dataset for coastal India and Bangladesh

This collection of gridded data layers provides the extent of inundation in May 2020 resulting from the cyclone Amphan in 39 coastal districts in India and Bangladesh.

Input data:

These geospatial data layers are derived from Sentinel-1 dual-polarization C-band Synthetic Aperture Radar (SAR) data for pre-Amphan (May 5-18, 2020) and post-Amphan (May 22-30, 2020) periods. We accessed ready-to-use SAR data on Google Earth Engine (GEE). These input data were preprocessed using Ground Range Detected (GRD) border-noise removal, thermal noise removal, radiometric calibration, and terrain correction, to derive backscatter coefficients (σ°) in decibels (dB). We used VH polarisation instead of VV, since the latter is known to be affected by windy conditions as compared to VH.

Methods:

We developed a binary water/non-water classification scheme for the pre- and post-Amphan images using the automated Otsu thresholding approach that finds optimum threshold values based on clusters found in the histograms of pixel values. This analysis resulted in eight images: four each for pre-Amphan and post-Amphan periods (one each for coastal districts of Odisha and West Bengal and two for Bangladesh for each period). The pixels in these images have two values: 0 for non-water and 1 for water.

We then used a decision rule to identify areas that changed from ‘non-water’ to ‘water’ after the cyclone. The decision rule generated the ‘inundation layer’ with the permanent water bodies such as river, lakes, oceans and aquaculture masked out. This analysis resulted in four images, each with pixels with a value of 1 for inundated regions.

Data set format:

The spatial resolution of all the derived datasets is 10m. These georeferenced datasets are distributed in GEOTIFF format, and are compatible with GIS and/or image processing software, such as R and ArcGIS. The GIS-ready raster files can be used directly in mapping and geospatial analysis.

Data set for download:

A. Three data layers for Odisha, India:

1. OD_pre_binary.tif
2. OD_post_binary.tif
3. OD_inundation.tif

These data layers cover 10 districts: Baleshwar, Bhadrak, Cuttack, Jagatsinghpur, Jajpur, Kendrapara, Keonjhar, Khordha, Mayurbhanj and Puri.

B. Three data layers for West Bengal, India:

1. WB_pre_binary.tif
2. WB_post_binary.tif
3. WB_inundation.tif

These data layers cover 9 districts: Barddhaman, East Midnapore, Haora, Hugli, Kolkata, Nadia, North 24 Parganas, South 24 Parganas, and West Midnapore.

C. Six data layers for Bangladesh – three each for lower (L) region and upper (U) region.

1. BNG_L_pre_binary.tif
2. BNG_L_post_binary.tif
3. BNG_L_inundation.tif
4. BNG_U_pre_binary.tif
5. BNG_U_post_binary.tif
6. BNG_U_inundation.tif

The data layers for the lower region cover 11 districts: Bagerhat, Barguna, Barisal, Bhola, Jhalokati, Khulna, Lakshmipur, Noakhali, Patuakhali, Pirojpur, and Satkhira.

The data layers for the upper region cover 9 districts: Chuadanga, Jessore, Jhenaidah, Kushtia, Meherpur, Naogaon, Natore, Pabna, and Rajshahi.