A critical analysis of ocean thermal analysis models in operation at FNOC

Abstract

Horizontal and vertical thermal structures were examined in a region of the Northeastern Pacific Ocean off Northern California. The observations were acquired on three cruises as part of the ONR-sponsored Ocean Prediction Through Observation, Modeling and Analysis (OPTOMA) research program, centered in a region of the California Current System, ca. 37 to 39 degrees N, 124 to 126 degrees W, during June and July, 1983. The horizontal temperature correlation scale was between 30 and 50 km, which was a significant factor when comparisons were made between measured horizontal thermal structures and those retrieved from Fleet Numerical Oceanography Center's (FNOC) analyses, which had grid spacing of 320 km (TEOTS) and 40 km (EOTS). Operational (modeled) analysis fields were not in agreement with the observed fields. The major discrepancies occurred in the magnitude of the mixed layer depth and the shape of the horizontal temperature fields (maps). The operational significance of differences between modeled and observed thermal structures was assessed in terms of their effect on low-frequency; i.e., less than 1 KHz, acoustic propagation utilizing the FACT9H and PE transmission loss models. Mixed layer depth differences produced significant disagreements between direct ranges computer from model and observed temperature profiles. The effect was most pronounced at higher frequencies and when both source and receiver were shallow; i.e. both at 20m. A comparison was made between average depth/temperature profiles from July, 1982, July, 1983, and FNOC climatology to obtain a measure of the effect of interannual variability in the domain. This comparison showed that a significant temperature anomaly existed in the upper 400m in 1983 compared to 1982 due to El Nino and that this anomaly was not represented by the FNOC climatology. The differences between modeled and measured thermal structures are believed to be related to thermal structure model resolution, model sensitivity to input data, short scales of spatial variability and non-representative climatology for the domain.