非表示:
キーワード:
-
要旨:
We apply causal effect networks to evaluate the influence of spring North Atlantic extratropical sea surface temperatures (NA-SSTs) on the summer East Atlantic (EA) pattern seasonal predictability during the period of 1908–2008. In the ECMWF Reanalysis of the 20th Century (ERA-20C), we find that the causal link from the meridional NA-SST gradient in spring (expressed by a meridional “SST index”) to the summer EA is robust during the period from 1958 to 2008, with an estimated causal effect expressed by a β coefficient of about 0.2 (a 1 standard deviation change in the spring SST index causes a 0.2 standard deviation change in the EA 4 months later). However, this causal link is not evident when analysing the entire period from 1908 to 2008. When performing the analysis on 45-year-long time series randomly sampled from this late period, we find the strength of the causal link to be affected by interannual variability, suggesting a potential modulation by an external physical mechanism. In addition to the summer EA, we find that the spring SST index has an estimated causal effect of about −0.2 on summer 2 m air temperatures over northwestern Europe. We then use different datasets from the Max Planck Institute Earth System Model in its mixed-resolution set-up (MPI-ESM-MR) to analyse the 1908–2008 period, focusing on a historical simulation and a 30-member initialised seasonal prediction ensemble. We specifically test the model's ability to reproduce the causal links detected in ERA-20C and evaluate their impact on the model's predictive skill for the European summer climate. We find that MPI-ESM-MR generally fails to reproduce the causal link between the spring SST index and the summer EA across the datasets. The 30-member initialised ensemble occasionally reproduces the causal link, though it typically underestimates its strength. We perform a predictive skill assessment conditioned on the spring SST index causal links for July–August sea level pressure, 500 hPa geopotential height, and 2 m air temperatures for predictions initialised in May. Our results suggest that while the overall impact may be limited, leveraging these causal links locally could help to constrain and improve the seasonal prediction skill of European summer climate.
