Cyclone Idai Kenneth Flood Mapping

Flooding near Beira March 2019.

What is this?

NEW: Data for Cyclone Kenneth in Cabo Delgado in Northern Mozambique.

Cyclone Idai has resulted in widespread flooding in Central Mozambique. Frequent radar images from Sentinel-1 are now freely available through the Copernicus Programme, and are an excellent resource for mapping floodwater. These data are, however, tricky to preprocess. Here we've produced ready-to-use Sentinel-1 mosaics and some simple flood maps for Sofala and Manica provinces of Mozambique. We hope to provide means to get a general idea of where flood waters are, and how they are changing over time.

Floodwater can be mapped with radar imagery based on a distinctively low backscatter associated with water bodies. Flat water tends to reflect radar energy, resulting in little energy being returned to the sensor from areas covered in water.

DONATE!

The MICAIA foundation, based in Chimoio supports a wide range of social enterprises in the affected area. They are calling for donations to help family and small enterprises and community and volunteer services get re-established later this year. When the aid is gone, so many communities will start again from a depressing base of destroyed infrastructure, lost assets built over years, and no source of support.

Pre-produced maps

We've put together some pre-generated maps of some flooded areas (more detailed maps are now available e.g. here)

DATA ACCESS

Images are shared through The University of Edinburgh datasync service and Google Drive. The datasync service requires a password, which is password.

NEW: Cyclone Kenneth, 1st - 3rd May 2019

Last updated 03/05/2019, 2000 UTC.
Sentinel-1 backscatter mosaic, ~1.0 GB, VH and VV polarisations.
Sentinel-1 backscatter mosaic (simplified), ~230 MB, VH and VV polarisations.
Classified flooding map, ~13 MB, here.

11th - 13th April 2019

Last updated 14/04/2019, 1730 UTC.
Sentinel-1 backscatter mosaic, ~2.0 GB, VH and VV polarisations.
Sentinel-1 backscatter mosaic (simplified), ~290 MB, VH and VV polarisations.
Classified flooding map, ~50 MB, here.

5th - 7th April 2019

Last updated 09/04/2019, 2100 UTC.
Sentinel-1 backscatter mosaic, ~2.4 GB, VH and VV polarisations.
Sentinel-1 backscatter mosaic (simplified), ~320 MB, VH and VV polarisations.
Classified flooding map, ~61 MB, here.

30th March - 1st April 2019

Last updated 02/04/2019, 1200 UTC.
Sentinel-1 backscatter mosaic, ~2.0 GB, VH and VV polarisations.
Sentinel-1 backscatter mosaic (simplified), ~320 MB, VH and VV polarisations.
Classified flooding map, ~60 MB, here.

24th - 26th March 2019

Last updated 27/03/2019, 1600 UTC.
Sentinel-1 backscatter mosaic, ~3.0 GB, VH and VV polarisations.
Sentinel-1 backscatter mosaic (simplified), ~487 MB, VH and VV polarisations.
Classified flooding map, ~61 MB, here.
Preview image, here.

18th - 20th March 2019

Sentinel-1 backscatter mosaic, ~2.8 GB, VH and VV polarisations.
Sentinel-1 backscatter mosaic (simplified), ~480 MB, VH and VV polarisations.
Classified flooding map, ~49 MB, here.
Preview image, here.

12th - 14th March 2019

Sentinel-1 backscatter mosaic, ~2.5 GB, VH and VV polarisations.
Sentinel-1 backscatter mosaic (simplified), ~430 MB, VH and VV polarisations.
Classified flooding map, ~70 MB, here.
Preview image, here.

6th - 8th March 2019

Sentinel-1 backscatter mosaic, ~1.7 GB, VH and VV polarisations.
Sentinel-1 backscatter mosaic (simplified), ~290 MB, VH and VV polarisations.
Classified flooding map, ~42 MB, here.
Preview image, here.

28th Feb - 2nd March 2019

Sentinel-1 backscatter mosaic, ~2.4 GB, VH and VV polarisations.
Sentinel-1 backscatter mosaic (simplified), ~430 MB, VH and VV polarisations.
Classified flooding map, ~69 MB, here.
Preview image, here.

Using the data

Data are provided as compressed GeoTiffs, which can be opened in common GIS software such as QGIS or ArcGIS. We have generated three products: one is a mosaic of backscatter from recent Sentinel-1 images, the second is identical to the first but simplified and compressed into a smaller file for easier download, and the third is a simple classification of the probability of a pixel being flooded.

  1. Sentinel-1 backscatter mosaic: Pixels take values of ~-30 dB to -5 dB, with higher numebrs associated with higher radar backscatter. We have produced mosaics of both VV and VH polarisations. The nodata value for these images is 0. See an example image here.
  2. Sentinel-1 backscatter mosaic (simplified): This is the same as above, but we've scaled the image between 0 and 255. To convert this back to decibels, divide pixel values by 6 then subtract 40. The nodata value for these images is 255.
  3. Classified flooding maps: We've classified pixels by their probability of being flooded. See an example image here. These images are much more lightweight (~50 MB), and it might be possible to download them on a limited internet connection. Each pixel is given one of five values:
    • [0] Unlikely to be flooded, > -15 dB HV backscatter.
    • [1] High probability of flooding, < -20 dB HV backscatter.
    • [2] Medium probability of flooding, -20 to -17.5 HV dB backscatter.
    • [3] Low probability of flooding, -17.5 to -15 HV dB backscatter.
    • [99] No data value

These classes should be considered with care, this is a very simple classification. Note that there will be many false positives, as well as some false negatives. Also take care with locations of no data, this does not indicate an absence of flooding.

All data were processed using our command line tool sen1mosaic, produced as part of the SMFM project.

These data are free to use. Please acknowledge Sam Bowers (The Universiy of Edinburgh) and the data source as Copernicus Sentinel data (2019).

Data from Sentinel-3

The first non-cloudy optical imagery of the region has been captured by Sentinel-3 on 25th March 2019. Download GeoTiffs for red, green, and blue bands here (~104 MB).

Preview image here.

Were these cyclones caused by climate change?

A few people have been asking this. Short answer is that you can't pin any one event on climate change, but that intenses cyclones will become more common in this region as the sea warms due to climate change. They key paper is here (open access). Fig 7 and Fig 9 show that intense cyclones are predicted to become more frequent along all of the Mozambican coast, under even moderate scenarios of climate change (up to 5 per decade). So these events were more likely to have happened because of climate change.

Are you using this data?

Please let us know! We'll continue to generate them as long as there is demand.

Who are we?

We're remote sensing scientists from The University of Edinburgh, who work in Mozambique. Get in touch at:

sam.bowers@ed.ac.uk (data processing), paulanietoq@gmail.com (mapping), casey.ryan@ed.ac.uk (both!)