Estimation of degree of sea ice ridging based on dual-polarized C-band SAR data

Abstract

Abstract. For ship navigation in the Baltic Sea ice, parameters such as ice edge, ice concentration, ice thickness and degree of ridging are usually reported daily in manually prepared ice charts. These charts provide icebreakers with essential information for route optimization and fuel calculations. However, manual ice charting requires long analysis times, and detailed analysis of large areas (e.g. Arctic Ocean) is not feasible. Here, we propose a method for automatic estimation of the degree of ice ridging in the Baltic Sea region, based on RADARSAT-2 C-band dual-polarized (HH/HV channels) SAR texture features and sea ice concentration information extracted from Finnish ice charts. The SAR images were first segmented and then several texture features were extracted for each segment. Using the random forest method, we classified them into four classes of ridging intensity and compared them to the reference data extracted from the digitized ice charts. The overall agreement between the ice-chartbased degree of ice ridging and the automated results varied monthly, being 83, 63 and 81 % in January, February and March 2013, respectively. The correspondence between the degree of ice ridging reported in the ice charts and the actual ridge density was validated with data collected during a field campaign in March 2011. In principle the method can be applied to the seasonal sea ice regime in the Arctic Ocean.

Estimating the Speed of Ice-Going Ships by Integrating SAR Imagery and Ship Data from an Automatic Identification System

“…We also tested the usefulness of some textural features. In our earlier studies, we observed that spatial autocorrelation ( [32,33]) and entropy ( [8,34,35]) are effective in SAR classification. Here, we computed them in a window size at a scale of 5.5 km in both polarizations.…”

“…For this reason, the detection and quantification of ridging from SAR imagery has been investigated from different aspects since the late 1980s. For Baltic Sea studies, see, for example, references [3][4][5][6][7][8].…”

“…If the resolution of SAR imagery is too coarse (like 500 m) to allow ridge sails to be visible, the ridging parameters can barely be obtained from the aggregated contribution of ridges to σ o in the pixel area. Recently, learning methods, such as the random forest classification of feature statistics [8], have shown promise in the determination of DIR from SAR images.…”

Interannual sea ice thickness variability in the Bay of Bothnia

“…The 2011 campaign was used by Gegiuc et al (2018) as validation data for the ice chart deformation numeral (degree of ice ridging). The agreement between the two datasets was generally good.…”

“…the point of view of navigational difficulty. No rules for estimating total thickness from ice chart thicknesses and the numerals have been established yet although a clear correlation exists (Gegiuc et al, 2018).…”