Archive for the ‘modelling’ Category

Ocean current visualization

16 April 2012

A friend recently comments about how impressed he was by the degree of accord in ocean current features between the NASA ocean model with data assimilation — and the GFDL ocean model without data assimilation —  It’s a good news/bad news situation.  The bad news being that the ocean data assimilation isn’t doing as much as you might hope.  But the good news is, this means that the models must be doing something right to be representing the ocean so similarly in spite of their different histories.

Data assimilation is a great thing, but for ocean currents, what you see in a global ocean data assimilation is very largely what the model chose to do, regardless of the data.  ARGO, for instance, provides information at about 300 spots per day (3000 floats, reporting once every 10 days).  Given 300 million square km in the ocean, that’s 1 million square km per float.  If the ocean’s correlation scale were like the atmosphere’s, upwards of 1000 km Rossby radius, that would be ok sampling, and more or less comparable to the radiosonde network (1000 per day, give or take).  Unfortunately, the mid-latitude ocean’s Rossby radius is more like 100 km, making the ocean undersampled by a factor of 100 or so.

Satellite altimetry, if it was used for assimilation, improves the sampling problem.  But not by as much as satellite sounders do for the atmosphere.  The problem here is that the altimetry tracks are very narrow, so, again, little of the ocean is seen per day.

Where there are no data, the result of assimilation is the model’s output.  So ca. 99% of what you’re seeing in the video is still model.

The good news is that the two models, with different lineages, are producing such similar results.  Suggests that maybe we do know something about ocean behavior.

If you want to see current ocean model forecasts, to 6 days, see


%d bloggers like this: