Characteristics Of Inherent Coupling Structure Of Model Climates

This work proposes a framework to examine interactions of climate modes that are identified as leading EOF modes; their coupling structure is unveiled through correlation analysis and helps in constructing a regression model, whose performance is compared across GCMs, thereby providing a quantitative overview of model performances in simulating mode interaction. As a demonstration surface temperature is analyzed for five CMIP5 preindustrial control (PiControl) simulations. Along with the seasonal land and ocean modes, four interannual modes are identified: the tropical mode (TM) associated with the Hadley circulation, the tropical Pacific mode (TPM) characterizing a zonal temperature contrast between the eastern tropical Pacific and the Atlantic/Indian Oceans, and two annular modes, the Arctic mode (AM) and Antarctic mode (AAM). All GCMs converge on the following points: 1) TM strongly couples with seasonal signals of the previous year; 2) TPM leads TM by 1 year, and thus a weaker zonal temperature contrast in the tropics contributes to warming in the entire tropical band 1 year later; and 3) AM weakly couples to TM at a 1-yr lead, suggesting that a colder North Pole may contribute to colder tropics. In addition, all GCMs do not support a linear coupling between AAM and TM. The above-learned coupling structure is incorporated to construct an optimum regression model that demonstrates considerable predictive power. The proposed approach may both serve as a useful tool for dynamical analysis and lend insight into GCM differences. Its merit is demonstrated by the finding that different representations of the mean seasonal cycle in GCMs may account for the GCM dependence of relative contributions of seasonal and interannual modes to TM variability.