Data files accompanying Culberg et al. (2024) "Sentinel-1 Detection of Ice Slabs on the Greenland Ice Sheet"

Contact:
Riley Culberg, rtculberg@cornell.edu

Files & Metadata:

FirnAquifers_3413.tif
	- Mask used to exclude regions with firn aquifers from further analysis. This dataset was originally published together with:
	  Brangers, I., Lievens, H., Miège, C., Demuzere, M., Brucker, L., & De Lannoy, G. J. M. (2020). Sentinel-1 Detects Firn Aquifers in 
	  the Greenland Ice Sheet. Geophysical Research Letters, 47(3), Article 3. https://doi.org/10.1029/2019GL085192
	  The original dataset can be accessed at: https://arcticdata.io/catalog/view/doi%3A10.18739%2FA2HD7NS8N.
	  Here, we have reprojected the results from the EASE2.0 grid to EPSG:3413 and resampled from 1km to 500m resolution for consistency with 
	  the rest of the datasets used in our analysis.
	- Spatial Resolution: 500m pixels, 1km intrinsic resolution
	- Data Format: 8 bit unsigned integer
	- DN Value Map:
		- 0: firn aquifer detected
		- 1: no firn aquifer detected

IceMask_3413.tif
	- Ice masked use to remove areas beyond the ice sheet, derived from the BedMachine Greenland dataset and resampled to 500 m resolution.
      Original data available at: https://nsidc.org/data/idbmg4/versions/5
	- Spatial Resolution: 500m pixels
	- Data Format: 8 bit unsigned integer
	- DN Value Map:
		- 0: no ice 
		- 1: ice

melt_mask.tif
	- Mask that discriminates between areas with consistent surface melting (percolation zone or west snow zone) vs. minimal melting (dry snow 
	  zone). Results are based on the difference between summer and winter backscatter, averaged over 1 Nov 2014 - 31 Aug 2020.
	- Spatial resolution: 500 m
	- Data Format: 8 bit unsigned integer
	- DN Value Map:
		- 1: melt detected
		- 0: minimal melt detected

melt_500m.tif	
	- The difference between summer and winter backscatter, averaged over 1 Nov 2014 - 31 Aug 2020, as a proxy for the extent of typical 
	  melting. A threshold of -4.7 dB (50,000 DN) is used to discriminate the dry snow zone. 
	- Spatial Resolution: 500m
	- Data Format: 16 bit unsigned integer
	- DN Value Map:
		difference (dB) = 20*(DN/65535) - 20

HV_2017_500m.tif
	- Average Sentinel-1 HV backscatter from winter 2016-2017 at 35 degree incidence angle. To correct for incidence angle, HV backscatter is 
	  multilooked for 500 m resolution and then regressed against incidence angle in each pixel. The resulting linear relationship is used to 
	  calculated the theoretical backscatter at 35 degrees. All data from 1 Oct 2016 to 30 April 2017 are used.
	- Spatial Resolution: 500m
	- Data Format: 16 bit unsigned integer
	- DN Value Map:
		HV (dB) = 25*(DN/65535) - 25

IceSlabs_2017_Final.tif	
	- Final estimate of the extent of refrozen ice facies on the Greenland Ice Sheet from Sentinel-1 backcatter observations between 1 Oct 2016
	  and 30 April 2017. These data have been validated with OIB ice-penetrating radar observations of ice slab extent.
	- Spatial Resolution: 500m
	- Data Format: 8 bit unsigned integer
	- DN Value Map:
		- 0: no refrozen ice detected
		- 1: maximum estimated extent of refrozen ice
		- 2: most likely estimated extent of refrozen ice
		- 3: minimum estimated extent of refrozen ice

xpol_2017_500m.tif
	- Average Sentinel-1 cross-polarized backscatter ratio (HH-HV) from winter 2016-2017 at 35 degree incidence angle. To correct for incidence 
	  angle, the HH and HV backscatter multilooked to 500 m and then separately regressed against incidence angle in each pixel. The resulting 
	  linear relationships are used to calculated the theoretical backscatter at 35 degrees for each polarization. All data from 1 Oct 2016 to 
	  30 April 2017 are used. Data are then subtracted from one another and multilooked to 500m resolution with boxcar filter.
	- Spatial Resolution: 500m
	- Data Format: 16 bit unsigned integer
	- DN Value Map:
		xpol (dB) = 12*(DN/65535) - 12;

Validation2017_UpperLimit.mat
	- Complete training data set for the upper elevationlimit of the ice slabs from the 2017 OIB data (presence/absence of ice slabs along flight tracks)
	  Ice slab presence or absence was derived from the high end estimate of ice slab extended presented in Jullien et al. (2023). The original 
	  data are available at https://doi.org/10.5281/zenodo.7505426.
	- .mat file with the following columns:
		- x - spatial coordinate of data point in EPSG 3413
		- y - vertical spatial coordinate of data point ESPG 3413
		- slabs - binary (0 or 1) variable where 0 means no ice slab was detected at that location and 1 means an ice slab was detected

Validation2017_Lower.mat
	- Complete training data set for the lower limit of the ice slabs from the 2017 OIB data (presence/absence of ice slabs along flight tracks)
	  Ice slab presence or absence was derived from the high end estimate of ice slab extended presented in Jullien et al. (2023). The original 
	  data are available at https://doi.org/10.5281/zenodo.7505426.
	- .csv file with the following columns:
		- x - spatial coordinate of data point in EPSG 3413
		- y - vertical spatial coordinate of data point ESPG 3413
		- slabs - binary (0 or 1) variable where 0 means no ice slab was detected at that location and 1 means an ice slab was detected

ValidationFolds_Upper.mat
	- OIB training data set divided into 10 regional folds for running the 10-fold cross-validation of the ice slab upper limit.
	  Ice slab presence or absence was derived from the high end estimate of ice slab extended presented in Jullien et al. (2023). The original 
	  data are available at https://doi.org/10.5281/zenodo.7505426.
	- .mat file with a cell array of length 10, one cell entry per fold. Each index in the cell array contains a Z x 3 matrix where the first 
	   column is the x coordinate, the second column is the y coordinate, and the third column is yes/no ice slabs.
		- column 1 - spatial coordinate of data point in EPSG 3413
		- column 2 - vertical spatial coordinate of data point ESPG 3413
		- column 3 - binary (0 or 1) variable where 0 means no ice slab was detected at that location and 1 means an ice slab was detected

ValidationFolds_Lower.mat
	- OIB training data set divided into 10 regional folds for running the 10-fold cross-validation of the ice slab lower limit.
	  Ice slab presence or absence was derived from the high end estimate of ice slab extended presented in Jullien et al. (2023). The original 
	  data are available at https://doi.org/10.5281/zenodo.7505426.
	- .mat file with a cell array of length 10, one cell entry per fold. Each index in the cell array contains a Z x 3 matrix where the first 
	   column is the x coordinate, the second column is the y coordinate, and the third column is yes/no ice slabs.
		- column 1 - spatial coordinate of data point in EPSG 3413
		- column 2 - vertical spatial coordinate of data point ESPG 3413
		- column 3 - binary (0 or 1) variable where 0 means no ice slab was detected at that location and 1 means an ice slab was detected