Paper Reviewed
Hidalgo, H.G. and Alfaro, E.J. 2015. Skill of CMIP5 climate models in reproducing 20th century basic climate features in Central America. International Journal of Climatology 35: 3397-3421.

In discussing their work, Hidalgo and Alfaro (2015) describe how they evaluated "a total of 107 climate runs from 48 coupled Model Inter-comparison Project 5 (CMIP5) general circulation models (GCMs)" in terms of "their ability to skillfully reproduce basic characteristics of late 20th century climate over Central America," where they ranked the models in terms of "metrics that take into consideration the mean and standard deviation of precipitation (pr) and surface temperature (tas), as well as the El Niño-Southern Oscillation (ENSO)-pr teleconnection," where "verification was performed by comparing model runs to observations and a reanalysis dataset."

So what did the two Costa Rican researchers discover? First of all, they report that (1,2) "the models had greatest difficulty in accurately reproducing the monthly standard deviation pr and tas patterns over the region of interest," that (3,4) "with some exceptions, the standard deviations for pr and tas were reproduced with low skill, even in the best models," that (5) "reproducing precipitation mean monthly patterns was also challenging for the models," and that other features that "are not simulated well" are (6-9) "ENSO teleconnection patterns and ENSO characteristic SST [sea surface temperature] patterns, ITCZ [Inter-tropical Convergence Zone] patterns and the seasonal cycles of the CLLJ [Caribbean Low-Level Jet] index."

And, therefore, we once again find ourselves ever learning ... but never quite getting to where we need to go in the realm of climate modeling.