Assimilation of CODAR HF-RADAR into The Regional Ocean Model System (ROMS)

By Hernan Arango
Presented at ASLO-99, on February 5, 1999, in Santa Fe, NM

During the LEO-15 Coastal Ocean Predictive Skill Experiments of 1998, Coastal Ocean Dynamics Applications RADAR (CODAR by SeaSonde Inc.) surface currents were assimilated into an ocean forecast model (ROMS) for the first time anywhere. The assimilation of CODAR data proved to be huge success in bringing the model output very close to reality. The four figures below show a short outline of the modeling process.

ROMS model output of the upwelling front for July 20 at 9:00 GMT. The colors indicate the ocean surface temperature, while the arrows indicate surface current speed and direction. Here, all the surface currents are moving NE parallel to the coastline. Also, the upwelling front is a single band of cold water along the coast, with little or no pertubations along it. This model output was dissimilar to the actual data pictured on the figure at the bottom of this page.

This model output was created assimilating hourly CODAR vectors. Notice that instead of a broad band of upwelling with currents moving Northeast, there are now 3 upwelling centers (two are partially cut off), with current rotations around the middle upwelling center. The CODAR data used here was detited, low-pass filtered, and then vertically extended using the most energetic (81%) Empirical Orthogonal Functions (EOF) from Acoustic Doppler Current Profiler (ADCP) data. The data were assimilated as horizontal vector components, computed by combining the radial data from each CODAR site. Unfortunately, this procedure tends to eliminate some of the data, unlike assimilation of raw CODAR radials, which does not eliminate any data.

Here we have a model output for the same time, but created by assimilating radial data. Notice that the upwelling center is better defined when the CODAR radials are assimilated and better compares to the AVHRR Sea Surface Temperature image below.

AVHRR Sea Surface Temperature (SST) image for July 20, at 11:57 GMT. Notice how the central upwelling center, and the edges of upwelling centers to the north and south agree with the radial model output above.
There is an item of interest to point out here. The SST data is approximately 2 degrees C cooler than all of the model output. This problem with surface heat flux input from last summer will be improved by coupling to an atmospheric model during this summer's experiment.

This final figure is a cross-shelf temperature section through LEO-15 for July 23, 1998, generated by the Naval Research Lab using the Modular Ocean Data Assimilation System (MODAS). Note how this forecast overestimates general surface temperatures by 1-2 degrees Celsius, and completely misses the upwelling front located at the coast (shown on the AVHRR SST inset). Our research is being done to improve these types of forecasts.