Internal variability and unforced trends of all-sky and clear-sky SSR: quantitative estimates using CMIP6

Quantifying trends in surface solar radiation (SSR) of unforced simulations is of substantial importance when one tries to quantify the anthropogenic effect in forced trends, as the net effect may be dampened or amplified by the internal variability of the system. In our analysis, we consider trends on different temporal scales (10, 30, 50 and 100 years) from 58 global climate models, participating in the Coupled Model Intercomparison Project - Phase 6 (CMIP6). We calculate the trends at the grid-box level for all-sky and clear-sky SSR using annual mean data of the multi-century pre-industrial control (piControl) experiments. The trends from both variables are found to depend strongly on the geographical region, as the most pronounced trends of the all-sky variable are observed in the Tropical Pacific, while the largest clear-sky trends are found in the large desert regions. Inspecting for each grid cell the statistical distribution of occurring N-year trends  shows that they are normally distributed in the majority of grid cells for both all-sky and clear-sky SSR. The 75-th percentile taken from these distributions (i.e. a positive trend with a 25 % chance of occurrence) varies with geographical region, taking values in the ranges 0.79 - 12.03 Wm^-2/decade for 10-year trends, 0.15 - 2.05 Wm^-2/decade for 30-year trends, 0.07 - 0.92 Wm^-2/decade for 50-year trends and 0.02 - 0.29 Wm^-2/decade for 100-year trends for all-sky SSR. The unforced trends become less significant on longer timescales – the trend medians, corresponding to the above ranges, are 3.18 Wm^-2/decade, 0.62 Wm^-2/decade, 0.29 Wm^-2/decade, 0.10 Wm^-2/decade respectively. The trends for clear-sky SSR are found to differ from the all-sky SSR by a factor of 0.16 on average, independent of the trend length. The model spread becomes greater at longer trend timescales, the differences being more substantial between large model families rather than between individual models. To elucidate the dominant causes of variability in different regions, we examine the correlations of the SSR variables with ambient aerosol optical thickness at 550 nm, atmosphere mass content of water vapour, cloud area fraction and albedo.