This dataset is provided according to the open access policies adopted by UiT The Arctic University of Norway. It consists of three images (time1, time2, ground truth) which are resampled, co-registered and cropped to cover an area affected by a flood.

The first image is a Landsat 8 acquisition covering Sacramento County, Yuba County and Sutter County, California, on 5 January 2017. It is composed of nine channels covering the spectrum from deep blue to short-wave infrared, plus two long-wave infrared channels. 

The second image was acquired on 18 February 2017 by Sentinel-1A over the same area after the occurrence of a flood. The image is recorded in polarisations VV and VH, and the dataset is augmented with the ratio between the two intensities as a third channel. All these channels are log-transformed.

First, both the images are clipped to remove outliers: for every channel, the minimum (maximum) is set equal to the mean - (+) 3 times the standard deviation. Every value below (above) such threshold is clipped up (down) to the minimum (maximum). Then, all the data in this dataset are normalised between -1 and 1.

The Landsat 8 and the Sentinel-1A acquisitions were processed and distributed by the Land Processes Distributed Active Archive Center (LP DAAC) and the European Space Agency (ESA) respectively, and further details about their data-sharing policies may be found here and here. IF YOU ARE USING THIS DATA PLEASE ACKNOWLEDGE NASA'S LP DAAC AND ESA'S COPERNICUS TO SUPPORT THEIR EFFORTS! 

We obtained the ground truth by the use of two other single-polarisation Sentinel-1 images acquired approximately at the same times as the previous ones. The ratio of these two is computed and manually thresholded to highlight changes.

PLEASE CITE [1] IF YOU ARE USING THE GROUND TRUTH WE ARE PROVIDING.

The .mat file includes three variables: 't1_L8_clipped','logt2_clipped','ROI' and it is available at this link.