Antarctic SAM-temperature relationships in the historical CMIP5 model runs

Variability in the Southern Annular Mode (SAM) is associated with a distinct spatial pattern of near-surface temperature anomalies across the Antarctic continent: a more positive SAM is generally linked with warming over the Antarctic Peninsula and cooler temperatures over East Antarctica. However, recent studies have shown that on decadal timescales these SAM-temperature relationships (STRs) can break down and even reverse in sign. Due to the relatively short (~50 years) timescale of available observations it is difficult to determine whether such reversals have occurred in response to anthropogenic forcing or are simply part of natural internal climate variability. Future Antarctic temperature change over the 21st Century is likely to be highly dependent on how the SAM responds to a combination of ozone recovery and greenhouse gas increases, which modelling studies suggest will have opposing effects on the sign of the SAM. Decadal variability in Antarctic STRs will be a major contributing factor to the uncertainty attached to such predictions. Therefore, as a first step in refining such predictions, we examine how well the Coupled Model Intercomparison Phase 5 (CMIP5) models, which provide climate projections for the latest IPCC assessment report, capture STR decadal variability across Antarctica. We analyze the seasonal STR at six Antarctic stations in the historical CMIP5 model runs, which typically encompass the period from 1850-2005. The stations were chosen because they have different seasonal patterns of STR; e.g. consistent across all seasons, one season where the STR is much stronger, or a change in the STR sign across the year. We utilize data from 50 different CMIP5 models comprising 183 separate model runs. First, we examine how well the models determine the long-term seasonal STR correlation and regression coefficients for the 1961-2000 period at each station, as compared to equivalent observations. Second, we investigate whether the models capture the observed decadal STR variability by analyzing SAM-temperature correlations derived from 11-year periods. We compare the model-derived statistics (standard deviation, range and probability distribution function) of these sets of correlations with observations for the period of overlap. We also examine these parameters for the full length of the historical model runs and determine the stability of the STR for each season/station combination. These statistics provide an indication of whether recent observed reversals are likely to be anthropogenically-forced rather than internal climate variability. They also allow us to establish which CMIP5 models may be expected to provide the best estimates of future STR variability and the uncertainty this will contribute to predictions of Antarctic temperature.