A real-time synoptic ocean thermal analysis/forecast system

Abstract

An ocean thermal analysis/forecast system is functioning in real time at the U.S. Navy's Fleet Numerical Oceanography Center (FNOC), Monterey, California. This paper provides a detailed description of the system and presents results illustrating its performance during a four-mouth test and evaluation period. The forecast component, designated as the Thermodynamic Ocean Prediction System (TOPS), is a synoptic mixed-layer model that employs the MELLOR and Yamada (1974) Level-2 turbulence parameterization scheme. It includes advection by instantaneous wind-drift and climatologically averaged geostrophic currents, and is forced by surface fluxes supplied by FNOC atmospheric models. The objective analysis component, designated as the TOPS-Expanded Ocean Thermal Structure (TOPS-EOTS) analysis, is a modified version of the conventional EOTS analysis ( Holl, Cuming and Mendenhall, 1979), which was the Navy's official ocean thermal analysis product during the period in question. It uses information blending techniques to map XBT and surface ship observations daily to a three-dimensional grid. It is coupled to TOPS in cyclical fashion, providing initial conditions, on any given day, for a 24 hr TOPS forecast that is subsequently fed back into TOPS-EOTS as a first-guess field for the following day's analysis. This supplies additional information to the analysis by linking it to the atmospheric forcing via the physics of TOPS, and allows representation of upper-ocean variability on time scales too short to be resolved adequately by the ocean thermal observations. Unlike those of the conventional EOTS analysis, day-to-day changes of the sea surface temperature (SST) field produced by the TOPS-EOTS analysis exhibit a low noise level and increase following the spring transition of the mixed layer. In addition, changes of the TOPS-EOTS thermal field tend to be consistent with the predictions of TOPS and, hence, the atmospheric forcing, while those of conventional EOTS do not. Time series of net surface heat flux and mixed-layer depth (MLD), spatially averaged over regions of area 0(10 7 km 2), show that the spring transition of the mixed layer predicted by the TOPS/TOPS-EOTS system occurs in qualitative agreement with the atmospheric forcing. Although the spatial averaging tends to smooth temporal variability, the spatially averaged MLD still shallows fairly abruptly, indicating that the transition occurs almost at once over very large regions. Concomitant with the shallowing of the layer, the spatially averaged SST begins to increase rapidly. The response of the model mixed layer to diurnal solar heating during the spring is also illustrated by time series of spatially averaged MLD and SST. In a relative sense, the mixed layer tends to be shallow and warm following the daytime heating and deep and cool following the nighttime cooling, as expected. Moreover, the capability of the system to represent variability on time scales too short to be resolved adequately by the ocean thermal observations is demonstrated. Composites of forecast verification statistics for the month of May indicate that TOPS exhibits skill consistently in forecasting the patterns of MLD and SST change, even for a forecast period (referred to as “TAU”) of 72 hr. Root-mean-square (RMS) forecast errors for MLD, again composited for May, show that TOPS betters persistence in all cases, except TAU = 72 hr for the Pacific test region. Similar RMS statistics for SST, however, indicate that TOPS betters persistence only at TAU = 24 hr for this parameter. This is a result of a warm bias in the TOPS SST forecast, which is probably due primarily to a bias in the net surface heat flux predicted by the FNOC atmospheric model, although a tendency of the turbulence parameterization scheme to underpredict MLD may also be a contributing factor. The first-guess fields of both the conventional and TOPS-EOTS analyses are compared daily to raw data consisting of the new observations that have not yet been assimilated. Although the conventional EOTS analysis is tuned to draw closely to the rather noisy ocean thermal data base, and consequently exhibits an excessive noise level itself, it generally does not agree with the new data substantially better than TOPS-EOTS does. An exception to this occurs for SST observations for the month of May, during which time the warm TOPS SST forecast bias discussed above is reflected in the comparison with the data.