Caribbean Precipitation in Observations and IPCC AR4 Models

摘要

A census of 24 coupled (CMIP) and 13 uncoupled (AMIP) models from the Intergovernmental Panel on Climate Change (IPCC) fourth assessment report (AR4) were compared with observations and reanalysis to show varied ability of the models to simulate Caribbean precipitation and mechanisms related to precipitation in the region. Not only were errors seen in the annual mean, with CMIP models underestimating both rainfall and sea surface temperature (SST) and AMIP models overestimating rainfall, the annual cycle was also incorrect. Large overestimates of precipitation at all SSTs (and particularly above 28 degrees C) and at vertical circulations less than -10hPa/day (the deep convective regime) were inherent in the atmospheric models with models using spectral type convective parameterizations performing best. In coupledmodels, however, errors in the frequency of occurrence of SSTs (the distribution is cold biased) and deep convective vertical circulations (reduced frequency) lead to an underestimation of Caribbean mean precipitation. On daily timescales, the models were shown to produce too frequent light rainfall amounts (especially less than 1mm/day) and dry extremes and too few heavy rainfall amounts and wet extremes. The simulation of the mid-summer drought (MSD) proved a challenge for the models, despite their ability to produce a Caribbean low-level jet (CLLJ) in the correct location. Errors in the CLLJ, such as too strong magnitude and weak semi-annual cycle, were worse in the CMIP models and were attributed to problems with the location and seasonal evolution of the North Atlantic subtropical high (NASH) in both CMIP and AMIP models. Despite these discrepancies between models and observations, the ability of the models to simulate the correlation between the CLLJ and precipitation varied based on season and region, with the connection with United States precipitation particularly problematic in the AMIP simulations. An observational study of intraseasonal precipitation in the Caribbean showed an explicit connection between the Madden-Julian oscillation (MJO) and Caribbean precipitation for the first time. Precipitation anomalies up to 50 percent above (below) the annual mean are observed in phases 1 and 2 (5 and 6) of the MJO and are related to changes in the CLLJ, that is also modulated by the MJO. Considerable progress has been made on identifying both problems and successes in the simulation of Caribbean climate in general circulation models, but many areas still require investigation.