【摘要】：Here we assessed the performances of IAP/LASG climate system model FGOALS-g2 and FGOAS-s2 in the simulation of the tropical Pacific Walker circulation(WC). Both models reasonably reproduce the climatological spatial distribution features of the tropical Pacific WC. We also investigated the changes of WC simulated by two versions of FGOALS model and discussed the mechanism responsible for WC changes. Observed Indo-Pacific sea level pressure(SLP) reveals a reduction of WC during 1900–2004 and 1950–2004, and an enhancement of WC during 1982–2004. During the three different time spans, the WC in FGOALS-g2 shows a weakening trend. In FGOALS-s2, tropical Pacific atmospheric circulation shows no significant change over the past century, but the WC strengthens during 1950–2004 and 1982–2004. The simulated bias of the WC change may be related to the phase of the multi-decadal mode in coupled models, which is not in sync with that in the observations. The change of WC is explained by the hydrological cycle constraints that precipitation must be balanced with the moisture transporting from the atmospheric boundary layer to the free troposphere. In FGOALS-g2, the increasing amplitude of the relative variability of precipitation(?P/P) is smaller(larger) than the relative variability of moisture(?q/q) over the tropical western(eastern) Pacific over the three time spans, and thus leads to a weakened WC. In FGOALS-s2, the convective mass exchange fluxes increase(decrease) over the tropical western(eastern) Pacific over the past 53 a(1950–2004) and the last 23 a(1982– 2004), and thus leads to a strengthened WC. The distributions of sea surface temperature(SST) trends dominate the change of WC. Over the past 55 a and 23 a, tropical Pacific SST shows an El Ni?o-like(a La Ni?a-like) trend pattern in FGOALS-g2(FGOALS-s2), which drives the weakening(strengthening) of WC. Therefore, a successful simulation of the tropical Pacific SST change pattern is necessary for a reasonable simulation of WC change in climate system models. This idea is further supported by the diagnosis of historical sea surface temperature driven AGCM-simulations.