Study on the morphology of Arctic melt pond based on ship-based images
With global warming,Arctic temperature is increasing,and melt pond has become an important fea-ture of Arctic sea ice.Due to the positive feedback of ice surface albedo,sea ice is melting rapidly,and thus the size and shape of melt pond are very important for the simulation of Arctic sea ice change and further for the vali-dation of existing remote sensing algorithms.In the present study,6 103 images of sea ice were obtained on-board the icebreaker R/V Xuelong 2 during the 12th Chinese National Arctic Research Expedition,5 489 of them remain after excluding invalid images such as uneven illumination and blurring,and then the spatial distribution and statistical characteristics of melt pond size and morphological parameters are investigated.The watershed al-gorithm is applied to segment the images into 9 uniform copies,and the random forest algorithm classifies the segmented images and creates a training set.Each image is classified into three surface categories including wa-ter,ice,and melt pond by the automated method,and the area fraction of each category as well as the size and shape of individual melt pond is calculated after tilt correction.The results show that the area fraction of melt pond increases and then decreases with increasing latitude,and the area fraction of ponded ice decreases with in-creasing latitude.The shorter melting time and earlier refreezing result in lower area fraction of melt pond at higher latitudes.The area,perimeter,and mean clamp diameter of the melt pond have similar patterns of change with increasing latitude.All of them rise first and then drop,indicating that the parameters of melt pond size have good consistency.The median value of melt pond area is significantly lower than the mean value indicating that there are more relatively small melt ponds.The frequency distribution of different sizes of melt pond areas is consistent with the power-law function,and the indices fitted to different observations are in the range of 1.4～1.8,with some similarity.The fractal dimension of the melt pond is more evenly distributed with increasing lati-tude,indicating that the melt ponds at different locations and evolutionary stages have some similarity.There is a significant difference in convexity values between melt ponds and circles at different latitudes,which indicates that the edges of the melt ponds form many concave shapes,and are more jagged due to the inconsistent melting rate of the edge line or the uneven depression of the sea ice surface.The perimeter-to-area ratio varies more with latitude,which indicates the difference in the number density of small melt ponds.And the average value of roundness of the melt ponds at various latitudes is 2.39±0.23,when the melt pond shape is close to a long rect-angle,indicating the non-round shape and geometric complexity of the melt pond.The roundness and convexity of the melt pond are positively proportional to its area,while the perimeter-area ratio is inversely proportional to the melt pond area,indicating that the melt pond shoreline will become longer and more curved with increase of the melt pond area,and jaggedness of the edge is more obvious.There is no obvious relationship between fractal dimension and melt pond area,which is more uniformly distributed.The average sea surface albedo caused by changes in melt ponds varies between 0.24 and 0.67 with increasing latitude,and is highly correlated with the latitudinal variation of ice area fraction,with a rate of change of 0.10(° N)-1 from 81.7° N to 86.2° N.The melt pond size and morphological parameters studied in this paper can be used for numerical simulation of melt pond evolution and to complement and validate remote sensing algorithms.
NETL
NSTL
万方数据
