European summer precipitation changes and the role of the large-scale circulation

Regional climate projections indicate that the future changes in European summer mean precipitation may be substantial, with significant drying in southern Europe and possible weak increases in at higher latitudes. Model uncertainties and natural variability are however large. Here we quantify the role of future large-scale circulation changes on future precipitation change in a 16-member single-model regional climate-model ensemble for the RCP8.5 emission scenario (Global climate model EC-Earth2.3, dynamically downscaled using the regional climate model RACMO2 for the period 1950-2100). Circulation analogues are used to distinguish three contributions. The first is the precipitation change occurring without circulation change. The second contribution measures the effects of changes in the large-scale mean circulation. It has a different spatial pattern and is closely related to high-pressure development west of Ireland. For a large area east of Ireland (including parts of western Europe), it is the major contributor to the overall drying signal, locally explaining more than 90% of the ensemble-mean change. The high-pressure region west of Ireland also appears in CMIP6 ensemble-mean projections, although it is weaker than in the EC-Earth2.3/RACMO2 ensemble because of model spread in the exact location of the high-pressure region. The third contribution records the net effect on precipitation of changes in the circulation variability. This term has the smallest net contribution, but a relatively large uncertainty. The analogues can be used effectively to partition the ensemble-mean change but describe only up to 40% of the ensemble-spread. This demonstrates that natural variability in precipitation drivers other than the large-scale circulation (e.g., SST, soil-moisture preconditioning) will generally strongly influence regional summer precipitation trends derived from single climate realisations and thereby reemphasises the need for using large ensembles or other methods where signal to noise ratios are high (e.g., pseudo-global warming experiments).