Landfast Ice Mapping Using MODIS Clear-Sky Composites: Application for the Banks Island Coastline in Beaufort Sea and Comparison with Canadian Ice Service Data

摘要

Landfast ice (LFI) is a prominent climatological feature in the Canadian Arctic. LFI is generally defined as immobile near-shore ice that remains fast along the coast and forms seaward from the land. It affects the coastline dynamics, is important for the near-shore ecosystems, wildlife, and human socio-economic activities. A method is proposed for mapping the LFI using time series of 10-day clear-sky composites derived at the Canada Center for Remote Sensing (CCRS) from the Moderate Resolution Imaging Spectroradiometer (MODIS) 250-m imagery. The delineation of coastal zone ice utilizes simultaneous analysis of the mean and standard deviation of MODIS monthly reflectance maps. The application of this method is demonstrated for a 20-year period (2000–2019) over the coastal zone of Banks Island in the Beaufort Sea. Detailed analyses have been conducted for three LFI parameters: (1) the total area (spatial extent) occupied by LFI; (2) the distance from the coast to the outer seaward LFI edge, and (3) the water depth at the outer seaward LFI edge. Comparison with the Canadian Ice Service (CIS) data demonstrates good agreement. The average correlation coefficients between CIS and CCRS time series in April-June, when the area reaches a maximum, are equal to 0.87–0.88. The mean differences (CIS-CCRS) are 344km~2 (5,464km~2 vs 5,120km~2) or 6.3% for the spatial extent; 1.3 km (17.6 km vs 16.3 km) or 7.4% for the distance; ～2.7m (～27.4m vs ～24.7m) or 10% for the water depth. Because the CCRS method uses monthly statistics, it tends to exclude potentially more mobile continuous landfast ice zones than the CIS analysis which is based on data collected on a specific date. The longterm trends of the LFI seasonal cycle in our region of interest since 2000 have shown a tendency for an earlier break-up, later onset, and longer ice-free period; however, these trends are not statistically significant.