Attribution of Arctic Sea-Ice Melting Patterns to Human Influence

During recent three decades, Arctic sea-ice has been decreasing with its rate accelerating. There have been, however, limited studies which have identified human influence on the Arctic sea-ice using a formal detection approach. In particular, no detection studies have been undertaken with systematic consideration of spatial information on Arctic sea-ice melting. This study conducts an updated detection analysis of Arctic sea-ice decline by comparing observation with Coupled Model Intercomparison Project Phase 5 (CMIP5) multi-model simulations in terms of spatial trend patterns during 1979-2012. The observations are passive microwave sea-ice concentrations (SICs) generated using the NASA Team algorithm, provided by the NSIDC (National Snow and Ice Data Center). The simulated Arctic SICs for the same period are obtained from available multi-model datasets of CMIP5 historical experiments which have been performed under natural plus anthropogenic forcing (ALL), natural only forcing (NAT), and greenhouse gas only forcing (GHG). For spatial pattern analyses, the Arctic is divided into 9 subregions and linear trends in sea ice extents are calculated for each subregion from observations and all model runs. We then apply an optimal fingerprinting technique where spatial trend patterns of observations are regressed onto model-simulated signals (ALL, NAT, and GHG). Pre-industrial control simulations are used here to estimate the range of internal natural variability. Results show that Arctic sea-ice loss has been dominant over Kara-Laptev Seas, E. Siberia-Chukchi Seas, and Barents Sea during warm seasons (JAS, OND) and that model simulations can reproduce the observed sea-ice melting patterns only when including greenhouse gas increases (ALL and GHG). Detection analysis confirms this, representing that ALL and GHG signals are detected across all seasons and also when four seasons are combined. GHG influence is also found to be separable from other forcings, indicating that recent Arctic sea-ice melting is largely attributable to anthropogenic increases in GHGs. Sensitivity of detection results to model performances and influence of the internal variability like Arctic Oscillation will be discussed.