This might be true in some places but not in the tropics, at least prior to 2001. shows the opposite pattern to the one invoked by Cowtan et al.Marine buoys in the tropics show that MAT trends were negative even as the SST trended up, and a global data set using MAT would show less warming than one relying on SST, not more.The pattern in the models had enough dispersion in CMIP3 to encompass the observed probabilities, but in CMIP5 the model pattern had a smaller spread and no overlap with observations.
found that the discrepancy between models and observations declines somewhat. shows that the use of SAT/SST (“blended”) model output data doesn’t actually close the gap by much: the majority of the reconciliation happens by using “updated forcings”, i.e. With the 2016 El Nino at the end of the record a crossing between the observations and the modified CMIP5 mean occurs.
In my version (using the unmodified CMIP5 data) the change to a 1970-2000 baseline would yield a graph like this: The 2016 Had CRUT4 value still doesn’t match the CMIP5 mean, but they’re close. method compresses the model data above and below so in Zeke’s graph the CMIP5 mean crosses through the Had CRUT4 (and other observed series’) El Nino peak.
This would appear to confirm the claim in Millar et al.
that climate models display an exaggerated recent warming rate not observed in the data.
The ranges of observed trends reflect observational uncertainty, whereas the ranges of model trends reflect forcing uncertainty, as well as differences in individual model responses to external forcings and uncertainty arising from internal climate variability.