Long-term behaviors of two versions of FGOALS2 in preindustrial control simulations with implications for 20th century simulations

Climate drift in preindustrial control (PICTL) simulations can lead to spurious climate trends and large uncertainties in historical and future climate simulations in coupled models. This study examined the long-term behaviors and stabilities of the PICTL simulations in the two versions of FGOALS2 (the Flexible Global Ocean-Atmosphere-Land System model Version 2), which have been submitted to the Coupled Model Inter-comparison Project Phase 5 (CMIP5). As verified by examining time series of thermal fields and their linear trends, the PICTL simulations showed stable long-term integration behaviors and no obvious climate drift the magnitudes of linear trends of SST were both less than 0.04oC (100 yr)-1] over multiple centuries. The changed SSTs in a century (that corresponded to the linear trends) were less than the standard deviations of annual mean values, which implied the internal variability was not affected. These trend values were less than 10% of those of global averaged SST from observations and historical runs during the periods of slow and rapid warming. Such stable long-term integration behaviors reduced the uncertainty of the estimation of global warming rates in the historical and future climate projections in the two versions of FGOALS2. Compared with the trends in the Northern Hemisphere, larger trends existed in the SST and sea ice extents at the middle to high latitudes of the Southern Hemisphere (SH). To estimate the historical and future climate trends in the SH or at some specific regions in FGOALS2, corrections needed to be carried out. The similar long-term behaviors in the two versions of FGOALS2 may be attributed to proper physical processes in the ocean model.