Each folder contains the dataset for the entire CONUS over the corresponding time period.
For folder contains two classes of geotiff images, namely the NW_Day_XXX_9km_Signals.tif and NW_Day_XXX_1km_Signals.tif, corresponding to data at 9 and 1 km spatial resolution respectively (XXX encodes the specific date e.g. 20180601)
Each *9km_Signals.tif image contains 7 bands that correspond to the following variables:
SMAP Tb, Horizontal polarization
SMAP Tb, Vertical polarization
Soil Moisture (estimation)
Clay
Bulk Density
Latitute
Longitude
Each *1km_Signals.tif image contains 12 bands that correspond to the following variables:
Sentinel 1 σ0, Vertical polarization
Sentinel 1 σ0, Horizontal polarization
SMAP TB, vertical polarization;
In-Situ Soil Moisture value
Soil Moisture (estimated) at 1km;
Vegetation Water Content
Surface Temperature
Precipitation
Land cover map
In-Situ location name
Latitute
Longitude
The OMBRIA dataset consists of Sentinel-1 and Sentinel-2 imagery constructed for benchmarking the OmbriaNet deep learning CNN architecture. OmbriaNet was designed for addressing the flood mapping problem.
Each satellite module contains three folders:
The folder named BEFORE contains an image of a region taken before a selected flood event
The folder named AFTER contains an image of the same region taken after the flood event
The folder named MASK contains a binary ground truth image of the region as produced by the EMS Rapid mapping.
The dataset contains time series of surface soil moisture at three locations, namely Idaho, Indiana, and Oklahoma.
For each location, the folder contains multiple gif images, one for each year, each containing 12 frames corresponding to monthly averages. Each example is 28x28 pixels in size