Trends of surface solar radiation in unforced CMIP5 simulations

Abstract

We consider decadal scale trends of annual mean all-sky surface solar radiation (SSR) that occur solely because of internal variability of the climate system. We give statistical estimates of their magnitude and probability of occurrence. The estimates are based on 43 preindustrial control (piControl) experiments of the Coupled Model Intercomparison Project phase 5 (CMIP5). Trends are found to depend strongly on geographical region and on whether they are quantified in absolute units or relative to the long-term mean SSR. We find it to be sufficient to provide one map for absolute and one for relative trends, as approximate analytical relations are shown to hold between trends of different length and likelihood and the standard deviation of the underlying SSR time series. We estimate that a positive trend over 30 years and with 25% chance of occurrence (75th percentile of all possible trends) has a magnitude between 0.15 and 1.7 W/m2/decade or 0.11 and 1.4% of long-term mean SSR per decade, depending on geographical location. Comparison with present-day observations and intermodel spread suggests an average uncertainty of these estimates of about 30%. Intermodel spread suggests that regional uncertainties can be up to about 3 times larger or smaller. We give examples of how these results may be used to obtain statistical estimates of how (un)likely it is that observed SSR trends or part thereof are due to internal variability alone.