Revisiting the observation-model discrepancy in Southern Hemisphere winter storm track trends

Reanalysis datasets show wintertime storminess in the Southern Hemisphere (SH) has significantly increased since 1979. Previous work reported an observation-model discrepancy whereby CMIP6 models were unable to reproduce the trend, calling into question the ability of climate models to accurately project the anthropogenic climate change in the SH extratropics. Here we revisit the observation-model trend discrepancy in SH winter storminess focusing on the impact of observational uncertainty, model ensemble size, and a like-for-like comparison. We find when storminess trends in all available reanalysis and model datasets are quantified on the same time and spatial grids, the trend distribution from prescribed sea surface temperature (SST) models (AMIP) does not exhibit a discrepancy from the reanalysis trend distribution. However, trends from coupled models (CMIP) still exhibit a discrepancy. The difference between AMIP and CMIP suggests SST trend biases affect the discrepancy. We test the importance of SST trend biases using tropical Pacific and Southern Ocean pacemaker simulations, which involve constraining coupled model SST trends to match observations. Tropical Pacific pacemaker simulations resolve the storminess trend discrepancy in the South Pacific and Southern Ocean pacemaker simulations resolve the storminess trend discrepancy in regions outside the South Pacific. Together, they remove the discrepancy in all SH sectors in the coupled simulations, confirming the importance of SST trend biases for the coupled model-observation discrepancy. Our results show that observation-model trend comparisons should involve all available datasets and like-for-like calculations. Furthermore, regional SST trend biases can lead to non-local storminess trend discrepancies between observations and models.