Regional and large-scale influences on seasonal to interdecadal variability in Caribbean surface air temperature in CMIP5 simulations

We evaluate the ability of global climate models participating in phase 5 of the Coupled Model Intercomparison Project (CMIP5) to reproduce observed seasonality and interannual variability of temperature over the Caribbean, and compare these with simulations from atmosphere-only (AMIP5) and previous-generation CMIP3 models. Compared to station and gridded observations, nearly every CMIP5, CMIP3 and AMIP5 simulation tends to reproduce the primary inter-regional features of the Caribbean annual temperature cycle. In most coupled model simulations, however, boreal summer temperature lags observations by about 1 month, with a similar lag in the simulated annual cycle of sea surface temperature (SST), and a systematic cold bias in both climatological annual mean air temperature and SST. There is some improvement from CMIP3 to CMIP5 but the bias is still marked compared to AMIP5 and observations, implying that biases in the annual temperature cycle may originate in the ocean component of the coupled models. This also suggests a tendency for models to over-emphasize the influence of SSTs on near-surface temperature, a bias that may be exacerbated by model tendency to over-estimate ocean mixed layer depth as well. In contrast, we find that both coupled and atmosphere-only models tend to reasonably simulate the response of observed temperature to global temperature, to regional and large-scale variability across the Caribbean region and the Gulf of Mexico, and even to more remote Atlantic and Pacific influences. These findings contribute to building confidence in the ability of coupled models to simulate the effect of global-scale change on the Caribbean.