How well do climate models reproduce North Atlantic subtropical mode water?

Formation and the subsequent evolution of the Subtropical Mode Water (STMW) involve various dynamic and thermodynamic processes. Proper representation of mode water variability and contributions from various processes in climate models is important in order to predict future climate change under changing forcings. The North Atlantic STMW, often referred as Eighteen Degree Water (EDW), in three coupled models (GFDL coupled data assimilation (CDA), GFDL CM2.1, and NCAR CCSM3.0) are analyzed to evaluate how well EDW processes are simulated in those models, and to examine whether data assimilation improves or degrades the model's response to forcing. In comparison with estimates from observations, the data assimilating model gives a better representation of the formation rate, the spatial distribution of EDW, and its thickness, with the largest EDW variability along the Gulf Stream path. However, GFDL CDA does not capture well the observed relationship between EDW volume and the seasonal destruction of EDW. Observations show a robust anti-correlation between the upper ocean heat content and air-sea heat flux, with upper ocean heat content leading air-sea heat flux by a few months. This anti-correlation is well captured by GFDL CM2.1 and CCSM3.0, but not by GFDL CDA. Only GFDL CM2.1 captures the observed anti-correlation between the upper ocean heat and EDW volume. This suggests that data assimilation degrades the model's thermodynamic response to forcing. The weak EDW formation rate in GFDL CM2.1 is related to the weak air-sea heat loss. Unlike the observed dominant southward movement of the EDW, the EDW in GFDL CM2.1 and CCSM3.0 moves eastward after formation in the excessively wide jet stream in the models.