models are only as good as our understanding. From the abstract:

> Here we account for the influence of three main natural variability factors: El Niño, volcanism, and solar variation.

All of these events are decades-long (or longer) cycles that don't have a substantial amount of data points... Sure, solar cycles seem to be 11 years, but we don't have a lot of scientifically usable (for forecasting) data points on that -- maybe 8 cycles? less? And the cycles are not consistent. It's not like Year 4 of one cycle is like year 4 of another cycle, we just determined there's a period of about 11 that looks significant.

Same with El Niño -- it's not like its 'true' or 'false', there's degrees of it.. and when it starts, and if other conditions are right to make additional hurricanes that year, and how much cloud cover that generates, etc. etc. a lot of which we don't have data on past 1960 when we launched our first weather satellite ...

As for volcanos... there's lots of them, and we are not great at predicting the high-impact events... we certainly don't have sufficient data to accurately predict what happens if we had a huge eruption on an El Niño strong year during the height of a solar cycle.