|Abstract||Icings, sheet-like masses of layered ice that form over the winter by freezing of successive flows of water extruded to the surface, are poorly understood arctic and sub-arctic hydrological phenomena.
For the first time, icings have been successfully mapped from a 30-year archive of satellite images, as a part of a case study undertaken to assess the utility of Landsat optical satellite imagery for mapping icings and to develop a methodology that
may be applied to other regions. Remotely sensed data enable mapping of large areas, and Landsat data have the added benefits of being free and easily available, and having a relatively long historical archive. Both hydrologically and practically
significant, icings indicate the location of springs, are important water storage components, can increase flooding potential by acting as major flow restrictions during spring freshet, and are a potential hazard to northern ice roads and all-weather
roads, decreasing the number of days of operational use. Being controlled by local and regional factors, icings reoccur at the same location, but not each winter, nor to the same extent. Consequently, icing process studies and risk assessments
require icing dynamics and distribution data from long-term datasets, but none have been compiled. Here we present the first semi-automated icing mapping approach, utilizing archival images acquired in late-spring when this region is largely
snow-free, but icy bodies remain. All ice bodies are mapped, but a water mask is used to clip out frozen water bodies (lakes and rivers), and the remaining ice bodies are considered to be land-fast icings, having formed from winter overland flow of
water. Maps from successive years are overlaid in a Geographic Information System to determine the total occurrence of icings and reoccurrence intervals. Nearly 5,500 icings were mapped in the Great Slave region around Yellowknife (21,886 km2) from
24 Landsat archival images (1985 to 2014). Icing size, which ranged by four orders of magnitude, was inversely related to return frequency, with 90 % of the total icing area returning 10 years or less. Substantial spatial variation occurred between
ecoregions within the study area where geological setting affects icing dynamics. Interannual variation of total icing area was considerable, ranging from 1.3 to 29.4 km2, and was in general agreement amongst the ecoregions, but was not coincident,
suggesting likely ecoregional climate variation. In addition, the groundwater source distances likely vary among the ecoregions according to geology and permafrost conditions. Consequently, icings in some regions may also be influenced by
meteorological conditions outside of the study region. This case study successfully demonstrates a relatively uncomplicated methodological approach to icing mapping, which may be operationalized with modifications to map icings throughout the north.
This approach takes advantage of Landsat archival data that allows for most icings to be mapped, and, being free, it is extremely cost effective. The data generated from this approach establish the baseline for winter hydrological variability in the
region against which future icing conditions under a changing climate regime may be compared.|