Fig. 1: Observed recent slowdown in September Arctic sea ice (ASI) and its association with multidecadal preceding summer North Atlantic Oscillation (NAO) variability.
From: Recent slowing of Arctic sea ice melt tied to multidecadal NAO variability
a The time series of regional averaged September sea ice concentration (SIC) across the Arctic region (66.5°–90° N, 180° W-180° E), defined as the ASI index, derived from Hadley (dashed red curve), ERA5 (European Centre for Medium-Range Weather Forecasts Reanalysis v5; dashed yellow curve) spanning 1950–2023, and NSIDC (U.S. National Snow and Ice Data Center; dashed purple curve) spanning 1979–2023, as well as the regional averaged September 2 m temperature across the Arctic region (66.5°–90° N, 180° W-180° E), defined as the surface air temperature (SAT) index derived from ERA5 (dashed grey curve) spanning 1950–2023. The dashed and solid blue curves depict the linear trend of the ASI index (unit: %/decade) averaged over the three datasets spanning 1996-2011 and 2012-2023, respectively. The dashed and solid black curves depict the linear trend of the SAT index (unit: K/decade) spanning 1996-2011 and 2012-2023, respectively. An asterisk (*) indicates statistically significance at p < 0.01, highlighting a statistically significant trend. b The time series of standardized September ASI index derived from Hadley (dashed red curve) and ERA5 (dashed yellow curve) and the preceding summer (June-July-August, JJA) NAO index (grey bar; the primary mode of atmospheric variability in the North Atlantic region) from 1955 to 2018, all de-trended and smoothed with an 11-year sliding average to examine their characteristics and relationships on an interdecadal time scale. The 11-year sliding average of the ASI index in year Y (a specific year) is the average ASI index from year Y-5 to Y + 5. The values r and p denote the correlation coefficients and significant levels between the NAO index and ASI index derived from Hadley and ERA5, respectively. The index is “standardized” by subtracting the mean and then dividing the resulting deviations by the standard deviation.
