Zonal structure and variability of the Western Pacific dynamic warm pool edge in CMIP5

The equatorial edge of the Western Pacific Warm Pool is operationally identified by one isotherm ranging between 28° and 29 °C, chosen to align with the interannual variability of strong zonal salinity gradients and the convergence of zonal ocean currents. The simulation of this edge is examined in 19 models from the World Climate Research Program Coupled Model Intercomparison Project Phase 5 (CMIP5), over the historical period from 1950 to 2000. The dynamic warm pool edge (DWPE), where the zonal currents converge, is difficult to determine from limited observations and biased models. A new analysis technique is introduced where a proxy for DWPE is determined by the isotherm that most closely correlates with the movements of the strong salinity gradient. It can therefore be a different isotherm in each model. The DWPE is simulated much closer to observations than if a direct temperature-only comparison is made. Aspects of the DWPE remain difficult for coupled models to simulate including the mean longitude, the interannual excursions, and the zonal convergence of ocean currents. Some models have only very weak salinity gradients trapped to the western side of the basin making it difficult to even identify a DWPE. The model’s DWPE are generally 1–2 °C cooler than observed. In line with theory, the magnitude of the zonal migrations of the DWPE are strongly related to the amplitudes of the Nino3.4 SST index. Nevertheless, a better simulation of the mean location of the DWPE does not necessarily improve the amplitude of a model’s ENSO. It is also found that in a few models (CSIROMk3.6, inmcm and inmcm4-esm) the warm pool displacements result from a net heating or cooling rather than a zonal advection of warm water. The simulation of the DWPE has implications for ENSO dynamics when considering ENSO paradigms such as the delayed action oscillator mechanism, the Advective-Reflective oscillator, and the zonal-advective feedback. These are also discussed in the context of the CMIP5 simulations