Observing System Simulation Experiments (OSSEs)

 This is work in collaboration with the joint UM/RSMAS and NOAA/AOML "Ocean Modeling and OSSE Center" (OMOC).

An Observing System Simulation Experiments (OSSE) is a tool to evaluate the impact of specific observation systems on our ability to accurately hindcast and forecast important physical processes. OSSEs aim at testing the performance of an existing or proposed observation system, in order to determine its impact on predictive models and optimize its deployment. Extracted from a reference model simulation that correctly represents the processes that are expected to be observed, a set of synthetic observations is assimilated into a second numerical simulation, which is also realistic but different from the reference one. The ability of the pseudo-observations to drive this second simulation close to the reference one gives indications about the expected performance of the network. Several such experiments may be performed to define the best possible network for observing the processes to study. 

In the GoM, we want to design one or several in situ networks that, when combined to remote sensing observations, are able to better observe the Loop Current (LC) variability. In particular, we are interested in detecting the LC extension towards the Northern Gulf shelf break, and the small cyclonic frontal eddies that surround the LC. These cyclonic eddies play a role in the detachment of large warm-core, anticyclonic ring from the LC, hence forcing the LC to retract to a most southern position.

DESCRIPTION:
Methodology highlights: We have first performed a reference simulation from 2004 to 2008. In order to evaluate the sensitivity of this reference simulation to boundary conditions, we have performed an ensemble of 40 simulations, each forced by perturbed boundary currents. Statistics calculated from this ensemble are a good proxy for the model error due to uncertainties in the boundary forcing. The ensemble can also be used to perform preliminary tests on the ability of the considered networks to detect some of the model error patterns.

Panel (a) is the reference simulation Sea Surface Height (SSH, cm) on May 22, 2007. It gives a good representation of the extended phase of the LC towards the Northern Gulf (associated to high SSH) and the presence of small, cyclonic eddies surrounding the LC (associated to low SSH). Panel (b) is the ensemble standard deviation in SSH (cm) on the same date. It shows how the model uncertainties, initiated at the boundary of the model, have high amplitude in the area associated to these cyclonic eddies.  Reducing these error patterns is expected by the networks to be tested using OSSEs. Panels (c) and (d) are examples of preliminary tests of two in situ networks measuring SSH. They show the signature in model SSH, on June 18, of the dominant error mode detected by each network in black dots. Whereas both networks detect the dominant error patterns located in the LC extension area, only the one extending on the Campeche Bank is able to detect model error extending on the shelf. This assessment technique is used to select, among various scenarios, those that will be further tested using OSSEs.

CONTACT: 
Villy Kourafalou — vkourafalou@rsmas.miami.edu
George Halliwell — ghalliwell@rsmas.miami.edu