Validation of a decadal OGCM simulation for the tropical Pacific

An ocean general circulation model is forced with NCEP reanalysis over the 1948-1999 period. The simulated dynamic height and sea level are compared respectively to the dynamic height computed from hydrological data and to the sea level measured by tide gauges in the tropical Pacific. The model is shown to capture important features of the temporal structure of variability in the tide gauge data over the last several decades. However, the comparison reveals a largely artificial trend in the simulation, which consists of a decreasing dynamic height and sea level in the southwest and northwest of the tropical Pacific. Model sensitivity experiments show this trend is controlled by the NCEP surface wind stresses and more precisely by a weakening in the trade winds and a trend in the off-equatorial wind curl, with this trend existing mainly before the mid 1970s. For studies of decadal variability, the simple removal of a linear trend is an inadequate way to solve the problem, due to the inhomogeneities in the data used in reanalysis products and the non-linearity of models.