Problems with solar, volcanic, and ENSO attribution using multiple linear regression methods on temperatures from 1979–2012

Abstract. The effectiveness of multiple linear regression approaches in removing solar, volcanic, and El Nino Southern Oscillation (ENSO) influences from the recent (1979–2012) surface temperature record is examined, using simple energy balance and global climate models (GCMs). These multiple regression methods are found to incorrectly diagnose the underlying signal – particularly in the presence of a deceleration – by generally overestimating the solar cooling contribution to an early 21st century pause while underestimating the warming contribution from the Mt. Pinatubo recovery. In fact, one-box models and GCMs suggest that the Pinatubo recovery has contributed more to post-2000 warming trends than the solar minimum has contributed to cooling over the same period. After adjusting the observed surface temperature record based on the natural-only multi-model mean from several CMIP5 GCMs and an empirical ENSO adjustment, a significant deceleration in the surface temperature increase is found, ranging in magnitude from −0.06 to −0.12 K dec⁻² depending on model sensitivity and the temperature index used. This likely points to internal decadal variability beyond these solar, volcanic, and ENSO influences.