Do we need another way to characterize sea ice floe size distributions?

Sea ice floe size distributions characterize how many ice floes of particular sizes are found in particular areas. They are commonly used to describe properties of the Arctic sea ice pack, such as variability in seasonality between floe sizes. Image segmentation algorithms, ranging from supervised learning methods to more analog erosion-expansion methods, are commonly used to distinguish these floe size distributions. Here we propose a semi-unsupervised k-means clustering method to quickly characterize bulk properties of floe size distributions. We find reasonable agreement between the floe size distribution slopes found between our k-means methodology and the erosion-expansion methodology of Denton and Timmermans [2022a]. We conclude by arguing that a particularly precise method of segmenting images is unnecessary for determining a reasonable characterization of floe size distribution relevant to the cryosphere community.