Long-Term Trends of Two Versions of FGOALS2

Abstract

Climate drift can lead to spurious long-term trends and large uncertainties in historical simulations and future projections in coupled models. The present study examines the long-term behavior of the preindustrial control (PICTL) simulations in FGOALS-g2 and FGOALS-s2 (the Flexible Global Ocean–Atmosphere–Land System Model Grid-point version 2 and Spectral version 2, respectively), which were submitted to the Coupled Model Intercomparison Project phase 5 (CMIP5). Driven by preindustrial level of solar radiation and greenhouse gases, FGOALS-g2 and FGOALS-s2 had no obvious climate drift by examining time series over 900 and 600 model years, respectively. Global average sea surface temperature (SST) exhibited the linear trends of −0.02 °C per century and −0.04 °C per century in FGOALS-g2 and FGOALS-s2, respectively, owing to a loss of net heat fluxes. These trends are less than 10 % of those of global average SST based on observations and twentieth century (or historical) runs during periods of slow and rapid warming. Moreover, these trends are smaller their own standard deviations in both versions of FGOALS2; this reduces the uncertainty of the estimated global warming rates in the historical and future climate projections and may not affect internal variability. Long-term trends will be large at some specific regions, especially for the middle to high latitudes of the Southern Hemisphere. The similar long-term behavior in both versions of FGOALS2 can be attributed primarily to the adoption of realistic physical processes in the ocean model.