Assessment of upper-ocean variability and the Madden-Julian Oscillation in extended-range air-ocean coupled mesoscale simulations

Atmosphere-ocean interaction, particular the ocean response to strong atmospheric forcing, is a fundamental component of the Madden-Julian Oscillation (MJO). In this paper, we examine how model errors in previous Madden-Julian Oscillation (MJO) events can affect the simulation of subsequent MJO events due to increased errors that develop in the upper ocean before the MJO initiation stage. Two fully coupled numerical simulations with 45-km and 27-km horizontal resolutions were integrated for a two-month period from November to December 2011 using the Navy's limited area Coupled Ocean/Atmosphere Mesoscale Prediction System (COAMPS). There are three MJO events that occurred subsequently in early November, mid-November, and mid-December during the simulations. The 45-km simulation shows an excessive warming of the SSTs during the suppressed phase that occurs before the initiation of the second MJO event due to erroneously strong surface net heat fluxes. The simulated second MJO event stalls over the Maritime Continent which prevents the recovery of the deep mixed layer and associated barrier layer. Cross-wavelet analysis of solar radiation and SSTs reveals that the diurnal warming is absent during the second suppressed phase after the second MJO event. The mixed layer heat budget indicates that the cooling is primarily caused by horizontal advection associated with the stalling of the second MJO event and the cool SSTs fail to initiate the third MJO event. When the horizontal resolution is increased to 27-km, three MJOs are simulated and compare well with observations on multi-month timescales. The higher-resolution simulation of the second MJO event and more-realistic upper-ocean response promote the onset of the third MJO event. Simulations performed with analyzed SSTs indicate that the stalling of the second MJO in the 45-km run is a robust feature, regardless of ocean forcing, while the diurnal cycle analysis indicates that both 45-km and 27-km ocean resolutions respond realistically when provided with realistic atmospheric forcing. Thus, the problem in the 45-km simulation appears to originate in the atmosphere. Additional simulations show that while the details of the simulations are sensitive to small changes in the initial integration time, the large differences between the 45-km and 27-km runs during the suppressed phase in early December are robust. Published by Elsevier B.V.