FY 1989

Observational verification of a quasi real time simulation of the tropical Pacific Ocean

Hayes, S.P., M.J. McPhaden, and A. Leetmaa

J. Geophys. Res., 94(C2), 2147–2157, doi: 10.1029/JC094iC02p02147 (1989)

Time series of upper ocean temperature and currents in the equatorial Pacific simulated by a numerical general circulation model run in nearly real time are compared to observations for the period August 1985 through May 1987. The model was forced by monthly mean wind stress and climatological air-sea heat flux. Comparisons with observations near the equator in the western (165°E), central (140°W), and eastern (110°W) Pacific are discussed. Simulated sea surface temperature (SST) was too cool in the eastern Pacific and too warm in the western Pacific. Largest rms deviations were in the east and exceeded 2°C. On and north of the equator in the eastern Pacific, SST and thermocline depth fluctuations on seasonal and monthly time scales were prominent. South of the equator, seasonal variability dominated. The model simulations often reproduced the amplitude and phase of the seasonal changes but not the higher-frequency variability. Model runs which included monthly assimilation of upper ocean temperature observations were included in the study. None of the comparison time series were incorporated in the assimilation. Inclusion of thermal observations generally improved agreement of simulated and observed time series. This improvement was largely due to reduction in the mean offsets of SST and thermocline depths. Data assimilation did little to improve the month-to-month differences in thermocline depth. In addition, south of the equator in the eastern Pacific, relatively large, systematic intra-month SST deviations occurred. These deviations corresponded to an erroneous heat flux of about 80 W m⁻² and indicated problems in the simulated upper ocean circulation. Although no velocity data were included in the assimilation, the improved model thermal structure led to improved velocity simulation at some locations. No comparisons indicated large-amplitude spurious velocity variations which could be associated with data assimilation transients.