Evaluation of Mid-Depth Currents of NCEP Reanalysis Data in the Tropical Pacific Using ARGO Float Position Information

1. Introduction The reanalysis dataset (Ji et al., 1995; Kalnay et al., 1996) offered by the National Centers for Environ- mental Prediction (NCEP) has been applied widely in relevant multidisciplinary research. Along with in- creasingly extensive observa…

Evaluation of mid-depth currents of NCEP reanalysis data in the tropical Pacific using ARGO float position information

The global project of the Array for Real-time Geostrophic Oceanography (ARGO) provides a unique opportunity to observe the absolute velocity in mid-depths of the world oceans. A total of 1597 velocity vectors at 1000 (2000) db in the tropical Pacific derived from the ARGO float position information during the period November 2001 to October 2004 are used to evaluate the intermediate currents of the National Centers for Environmental Prediction reanalysis. To derive reliable velocity information from ARGO float trajectory points, a rigorous quality control scheme is applied, and by virtue of a correction method for reducing the drift error on the surface in obtaining the velocity vectors, their relative errors are less than 25%. Based on the comparisons from the quantitative velocity vectors and from the space-time average currents, some substantial discrepancies are revealed. The first is that the velocities of the reanalysis at mid-depths except near the equator are underestimated relative to the observed velocities by the floats.The average speed difference between NCEP and ARGO values ranges from about -2.3 cms-1 to -1.8cms-1. The second is that the velocity difference between the ocean model and the observations at 2000dB seems smaller than that at 1000 dB. The third is that the zonal flow in the reanalysis is too dominant so that some eddies could not be simulated, such as the cyclonic eddy to the east of 160°E between 20°N and 30°N at 2000 dB. In addition, it is noticeable that many floats parking at 1000 dB cannot acquire credible mid-depth velocities due to the time information of their end of ascent (start of descent) on the surface in the trajectory files. Thus, relying on default times of parking, descent and ascent in the metadata files gravely confines their application to measuring mid-depth currents.