Sea-Ice Floe Segmentation Products Derived from MEDEA Imagery for 1999 through 2014 in the Canada Basin

This dataset contains sea-ice floe segmentation products derived from 78 Medea images from the United States Geological Survey (USGS) Global Fiducials Library (GFL) acquired in the Arctic's Canada Basin from 1999 to 2014. Sea-ice floe segmentation is the identification of individual sea ice floes in imagery for the purpose of retrieval of unique properties of those individual floes, such as their size or shape. The product was developed for and used in the analysis of the accompanying paper, Denton and Timmermans (2022). The product was created using an algorithm written by Alexis Denton (Yale University) in tandem for the manuscript analysis. The script is also available at github.com/dentonaa/sea-ice-floe-segmentation. This image product was developed to contribute a novel, multidecadal dataset of identified sea ice characteristics, namely floe size, to the Arctic science community. If you use this product in your work or research or for any other reason, credit the authors here and cite the code DOI issued by Zenodo. The DOI badge to the right points to the latest released version of the repository. The development of this product and the research presented in the manuscript was funded by the Office of Naval Research as a part of their Multi-University Research Initiative (MURI) Mathematics and Data Science for Physical Modeling and Prediction of Sea Ice. To learn more about the work of the Sea Ice MURI, please visit https://seaicemuri.org/. For further details about the dataset, please see Denton and Timmermans (2022). A full description of the product files is as follows: This dataset is comprised of the sea-ice floe segmentation products derived from 78 MEDEA images acquired at various locations in the Canada Basin from 1999 through 2014, and one ReadMe file. For each sea-ice floe segmented image, there are 3 files. Each of the 3 files starts with the naming convention, "sifloeseg_MEDEAImageName_segmentationParameters_", followed by a filename ending designated by the type of segmentation data stored, as described below. sifloeseg stands for sea-ice floe segmentation. MEDEAImageName is the original filename of the MEDEA image, usually a series of 3 or 4 strings delimited by "_" in which the first string is the site/project of acquisition, the second is the date (YYYYMMDD), and the third and fourth describe whether the image is a mosaic or a tile in a mosaic. e.g., "beaufo_19990728_2" indicates a MEDEA image acquired in the Beaufort Sea on 28 July 1999 which is the 2nd tile of a larger mosaic. Please visit the USGS GFL for more details on the naming of MEDEA image files. segmentationParameters describes the parameters used to segment the image and follows the format "bwt0##e##hs". bwt stands for "black-white threshold" and is followed by the value of the classification threshold (0.##) between 0 and 1 used to classify the image. e stands for "erosions" and is followed by the number of erosions used to erode the image during the first heirarchical step. hs indicates that the erosion scheme used a half-step, or halving of the number of erosions between hierarchical iterations. e.g., "sifloeseg_beaufo_19990728_2_bwt015e1hs_" indicates a segmentation of the above MEDEA image example which used a classification threshold of 0.15 and 1 erosion for the first hierarchical segmentation step. The 3 files for each segmented image are: A MATLAB file containing the areas (in square meters) of the segmented floes in an image. The file has the naming convention, "sifloeseg_MEDEAImageName_segmentationParameters_areas_sqm.mat", in which sqm stands for "square meters". This file contains the following variables: L_areas_nobc_sqm, which is a vector containing floe areas retrieved from the original segmentation (no border-clearing; contains floe areas wholly and partially in the image), where L stands for "Label". L_areas_bc_sqm, which contains the floe areas retrieved from the border-cleared segmentation (contains only floe areas wholly in the image). A MATLAB file containing the image segmentation, which is a labeled numeric array in which each positive integer numeric label corresponds to a unique sea-ice floe. The file has the naming convention, "sifloeseg_MEDEAImageName_segmentationParameters_labeled.mat". This file contains the following variables: rows: the number of image rows (y-direction array size, double) cols: the number of image columns (x-direction array size, double) bwt: the classification (black-white) threshold used to classify the image during segmentation (fractional value between 0 and 1, double) numErosions: a vector containing the number of erosions used to erode the image during segmentation at each hierarchical step, in which the first value indicates the number of erosions used during the first step, and the last indicates the number of erosions used during the last step, which is always 0 (1 x n double, in which n is the number of hierarchical steps used) numfmax: the total number of wholly and partially segmented floes in the original segmentation (before border clearing, double) L: the original image segmentation containing wholly and partially segmented floes (prior to border clearing), which is a numerical label array, in which positive integer values represent unique floes and values of 0 represent ocean pixels (rows x cols, double). Lbc: the border-cleared image segmentation containing only wholly segmented floes (after border clearing), which is a numerical label array in the same vein as L (rows x cols, double) A PNG representation of the segmented image in RGB (red-green-blue) color, in which non-unique colors represent different floes. This file is for visualization and display purposes. The same image and its border-cleared counterpart can be produced with the following lines of code in MATLAB: load(sifloseg_MEDEAImageName_segmentationParameters_labeled.mat); Lrgb = label2rgb(L,'jet','k','shuffle'); imshow(Lrgb) Lbcrgb = label2rgb(Lbc,'jet','k','shuffle'); imshow(Lbcrgb) For more information on any of the above terms, please see Denton and Timmermans (2022).

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.