Using altimeter measurements for quantitative assessment of high resolution ocean models
LuAnne Thompson(1) and Kathryn Kelly(1)
University of Washington,
Seattle, WA 98195,
Satellite altimeter missions since the 1990s have provided an important benchmark for the evaluation of the fidelity of ocean models. Eddy kinetic energy and sea surface height variability have been used extensively as a comparison point for models of northern hemisphere western boundary currents and their extensions. As longer time series of altimetry observations have become available, they can be used, along with other observations, to evaluate the model’s ability to reproduce longer term variability. Quantitative comparisons then become possible. During the 1990s, ocean models have improved immensely, benefiting from improved forcing fields, as well as from an amazing increase in computational power. These developm ents have resulted in increased model complexity as well as an increase in resolution and length of model runs. The realism of western boundary current extensions in these models in particular has improved. The increase in resolution to better than 10 km has allowed the western boundary current to separate from the coast at about the correct latitude, making possible more detailed comparisons with altimeter observations and observationally derived budgets.
Comparisons that are possible with the long time series of altimetric observations include an evaluation of the mean path strength of the current and now, because of the long time series of altimeter data, path variability. In addition, the upper ocean heat budget can be examined using as observational estimates diagnostic model results that rely on the altimeter observations (Vivier et al, 2002; Dong and Kelly, 2004). Hindcast model runs over some of the same time period can be evaluated against these metrics. Two very high resolution model runs are evaluated critically against the altimeter observations, one in the North Atlantic, and one in the North Pacific. These two model runs are from diference ocean models and use two different model representations of the vertical coordinate, but each, at least qualitatively, has a western boundary current that has a good separation latitude, penetration into the interior and strength.
However, with a more quantitative analysis, model biases can be used to infer deficiencies in model physics. In the North Pacific, the 11 year run at 1/12 degree resolution using the HYCOM (Hybrid Coordinate Ocean Model) show good qualitative performance; however, in a quantitative assessment, the western boundary current path has systematic errors that lead to errors in both the SST and the regional heat budget. In addition, the interannual variability of the strength of the Kuroshio and its path is weaker than observed, with the western boundary current extension being less stable than is observed. The errors have potentially many sources, including errors in the forcing fields; however advection errors in the model ocean explain at least some of the model biases. A similar analysis is done for a North Atlantic simulation that is run using POP (Parallel Ocean Program).
With the advent of increased computational power, coupled climate models will begin to be run at eddy resolving resolution in the ocean. It is important that the model performance is understood in an uncoupled framework and without data assimilation so that model biases are known and can potentially be corrected. The forced ocean model simulation evaluations described here are a first step in determining the validity of high resolution coupled simulations of future climate.