Detection of melt onset over the northern Canadian Arctic Archipelago sea ice from RADARSAT, 1997-2014

We present an algorithm to detect melt onset over Arctic sea ice using high resolution synthetic aperture radar (SAR) images from RADARSAT. The algorithm is based on the temporal evolution of the SAR backscatter coefficient (σ°), using an ice type specific threshold approach that also corrects for incidence angle variation. Using 4457 RADARSAT images, the algorithm was applied over sea ice in the northern Canadian Arctic Archipelago that low spatial resolution microwave sensors have difficulty resolving, thus generating a new time series of melt onset from 1997 to 2014. Over the 1997-2014 period, the mean annual average melt onset date in the northern CAA was on year day 164 (mid-June) with a standard deviation of 4 days. No significant trend for melt onset date was found over the 18-year period. An earlier (later) melt onset was associated with increased (decreased) solar energy absorption (r = 0.64, significant at 95% confidence) and subsequently associated in lighter (heavier) September sea ice coverage in the northern CAA (r = 0.62, significant at 95% confidence). RADARSAT estimates of melt onset were found to be in good agreement but more robust compared to passive microwave and scatterometer estimates over the northern CAA.