Characterizing the annual-mean climatic effect of anthropogenic CO2 and aerosol emissions in eight coupled atmosphere-ocean GCMs

Abstract

This work examines the spatial patterns of the transient response of mean annual temperature and precipitation to CO2 (or CO2 plus aerosol or aerosol proxy) radiative forcing in eight coupled AOGCMs, generally for the period 1900–2099. Response patterns are characterized using empirical orthogonal functions (EOFs) and the quasi-EOFs of Harvey and Wigley (the first qEOF field, discussed here, is given by the correlation between local year-by-year temperature changes and the global mean temperature change). The first temperature EOF accounts for 80–95% of the space-time variation of the CO2 run in all of the models, and is almost identical to qEOF1 of the temperature response or to the temperature change pattern averaged over the last 30 years of the simulations. EOF1 accounts for 80–95% of the space-time variation in the CO2+aerosol runs in six of the eight models. The CO2 response patterns of different models are highly correlated with one another (R 2 generally >0.5), and are also highly correlated with the CO2+aerosol response patterns (R 2≥ 0.85 in all except one model). The difference between CO2 and CO2+aerosol runs can be represented by EOF1 of the year-by-year differences, by qEOF1 of the year-by-year differences, or by the difference in temperature averaged over the last 30 years of each run. In models where these representations are highly correlated with each other, they are also highly correlated with CO2 EOF1. In other cases, aerosol EOF1 is modestly to highly correlated with control EOF1 (i.e.: the year-by-year differences between CO2 and CO2+aerosol runs are dominated by internal variability), while aerosol qEOF1 and the 30-year difference are highly correlated with each other. For all models, the decadal mean temperature change can be closely replicated by scaling the CO2 EOF1 pattern based on the global mean temperature changes (RMSE for the last decade is <6% of the RMS temperature change for CO2 runs, <8% for CO2+aerosol runs). The first EOF of the precipitation response to increasing CO2 accounts for only 10–30% of the space-time variation, and is generally highly correlated (R 2 up to 0.85) with control EOF1. In all of the models, there is an increase in precipitation in the ITCZ and a decrease in bands at or near 30°S and 30°N. In many models there is an El Nino-like response, including a substantial decrease in precipitation over the Amazon. Global-mean precipitation increases in all models due to CO2 forcing, but aerosols appear to have a disproportionally large effect in suppressing the increase compared to their effect in suppressing the warming. There is evidence in some models that the non-absorbing aerosols considered here reduce summer monsoon rainfall compared to the changes that would be expected based on the globally averaged effect of aerosols on precipitation. When regional precipitation changes over time are predicted by scaling a fixed precipitation-change pattern with the global mean temperature change, the global mean RMSE in the predicted change in decadal-mean precipitation is 25–35% of the global RMS precipitation changes by the end of the simulation.