Meeting Abstracts

<< Previous Abstract | C101 | C102 | C103 | C104 | C201 | C202 | C203 | C204 | C301 | C302 | C303 | C304 | Next Abstract >>

STATISTICS OF MODEL DRIFT ACCURACY

Michael Toner
Naval Oceanographic Office
(Abstract received 04/17/2012 for session C)
ABSTRACT

The Naval Oceanographic Office maintains a real-time operational ocean model capacity that covers the entire ocean in either a global, regional, or coastal form. These data assimilating, general circulation models provide a wide variety of support to the Navy, including drift prediction products.

In an effort to understand the basic question of model drift accuracy, surface and near surface drifters were used to develop regional statistics of model drift accuracy. Model predictions were initialized along drifter paths every three hours. Two time-dependent metrics were computed over a typical model forecast period of 72 hours: distance error and bearing error (measured from the initialization point). Errors from several thousand predictions were then combined to a common axis of hours past initialization.

A surprisingly elegant result came from the statistics. While individual trajectory comparisons produced the expected oscillatory behavior in the error time-series, bulk statistical results were remarkably stable.