Climatic trends and regional climate models intercomparison over the CORDEX-CAM (Central America, Caribbean, and Mexico) domain

摘要

An intercomparison of three regional climate models (RCMs) (PRECIS-HadRM3P, RCA4, and RegCM4) was performed over the Coordinated Regional Dynamical Experiment (CORDEX)-Central America, Caribbean, and Mexico (CAM) domain to determine their ability to reproduce observed temperature and precipitation trends during 1980-2010. Particular emphasis was given to the North American monsoon (NAM) and the mid-summer drought (MSD) regions. The three RCMs show negative (positive) temperature (precipitation) biases over the mountains, where observations have more problems due to poor data coverage. Observations from the Climate Research Unit (CRU) and ERA-Interim show a generalized warming over the domain. The most significant warming trend (>= 0.34 degrees C/decade) is observed in the NAM, which is moderately captured by the three RCMs, but with less intensity; each decade from 1970 to 2016 has become warmer than the previous ones, especially during the summer (mean and extremes); this warming appears partially related to the positive Atlantic Multidecadal Oscillation (+AMO). CRU, GPCP, and CHIRPS show significant decreases of precipitation (less than -15%/decade) in parts of the southwest United States and northwestern Mexico, including the NAM, and a positive trend (5-10%/decade) in June-September in eastern Mexico, the MSD region, and northern South America, but longer trends (1950-2017) are not statistically significant. RCMs are able to moderately simulate some of the recent trends, especially in winter. In spite of their mean biases, the RCMs are able to adequately simulate inter-annual and seasonal variations. Wet (warm) periods in regions affected by the MSD are significantly correlated with the +AMO and La Nina events (+AMO and El Nino). Summer precipitation trends from GPCP show opposite signs to those of CRU and CHIRPS over the Mexican coasts of the southern Gulf of Mexico, the Yucatan Peninsula, and Cuba, possibly due to data limitations and differences in grid resolutions.