One known issue with climate models is that they tend to suggest a larger response to a volcanic eruption than is observed. A recent paper by Lehner et al. shows that all large eruptions since 1951 coincided with El Niño events. El Niño events are a form of unforced variability that causes a period of warming. Hence, if a volcano occurs at the same time as an El Niño event, the influence of the volcano will appear smaller than if it occured when it was ENSO neutral or at the same time as an Le Niña event.
Consequently, what Lehner et al. (2016) consider is how climate models and observations compare if you correct the observations for ENSO events, or if you consider only models with co-incident El Niño events. This is shown in the figure below. The solid black line is the observed temperature, while the dashed black line is the observations corrected for ENSO events. The blue line is all the climate models considered, while the red line is only those that have co-incident El Niños. It’s clear that correcting the observations for ENSO events produces a better match (dashed black line and blue), as does considering only those climate models with co-incident El Niño events (solid black line and red line).
Of course, this is only one paper and – as it acknowledges – it doesn’t rule out that there may still indeed be model errors in the response to volcanic eruptions (forcings, aerosols, atmospheric circulation). It does, however, make a very good point in the conclusions
If a phenomenon under study is rare and the observational record short ….., potential superposition of forced and unforced variability needs to be taken carefully into account. These results are also important for paloclimate research, as potential superposition of eruptions and El Niño events during the last millennium adds another source of uncertainty to paleoclimate model-data comparisons.