Detection of pan-Arctic terrestrial snowmelt from QuikSCAT, 2000-2005

New pan-Arctic (> 60 degrees N) snowmelt onset and end date records (2000-2005) were produced from enhanced resolution (4.45 km) QuikSCAT (QSCAT) Ku-band backscatter measurements. A modified threshold method was used to detect multiple melt events and the main melt event was identified by the longest melt duration or the largest melt intensity. Algorithm validation was performed with a Circum-Arctic network of station measurements extracted from the Global Surface Summary of Day product maintained by the National Climatic Data Center (NCDC). Estimates of melt onset date corresponded to a clear shift in the statistical distribution of mean daily air temperatures from largely below freezing to above freezing. Melt was not detected in 7% of the cases mainly over dense boreal forest cover, or very shallow tundra snow cover. QSCAT-derived melt end dates were significantly correlated with station-derived snow-off estimates (r(2)= 0.79: RMSE = 11.6 days). The difference between the two estimates was within one week for 70% of the cases so there is good potential to use QSCAT-derived melt end dates to approximate snow-off dates across the pan-Arctic region. There was a tendency for QSCAT-derived snow-off dates to be biased early over dense forest regions due to a diminished backscatter melt signature. Crown Copyright (c) 2008 Published by Elsevier Inc. All rights reserved.