Previous Abstract | Back to Abstracts Page | Next Abstract
The REMO Ocean Data Assimilation System into HYCOM
Davi Mignac, Alex Santana, Filipe Costa, Leonardo Nascimento, Clemente Tanajura
Oceanographic Modeling and Observation Network (REMO), Federal University of Bahia
(Abstract received 04/30/2015 for session X)
ABSTRACT
In this study, the recently implemented REMO Ocean Data Assimilation System (RODAS) into the HYbrid Coordinate Ocean Model (HYCOM) is presented together with a series of Observing System Experiments (OSEs). In these experiments, components of the Global Ocean Observing System (GOOS) were systematically withheld from 3-year assimilation runs over the Atlantic Ocean. RODAS employs a multivariate Ensemble Optimal Interpolation scheme that assimilates Argo Temperature (T) and Salinity (S) profiles, UK MetOffice OSTIA Sea Surface Temperature (SST) analyses and satellite Sea-Level Anomalies (SLA). Using the same initial condition from a full assimilation run, the following OSEs were performed from 1 January 2010 to 31 December 2012 withholding: (i) only Argo; (ii) only OSTIA; (iii) only altimetry data; and (iv) all observation types. These OSEs were also compared with the full assimilation run and the model free run to evaluate the impact of different observations into the model state. The results show that each observation type brings complementary information to the analyses. Assimilation of SST is needed to better constrain the mixed layer temperature, while assimilation of SLA is the main responsible to improve the mesoscale circulation and large-scale patterns such as the Gulf Stream and the Brazil-Malvinas Confluence. In the subsurface, only Argo observations are able to constrain the thermohaline state. When Argo data are excluded, the quality of S is seriously compromised and becomes worse than the free run in the upper ocean. Additionally, the run withholding all observation types shows that the model state in the surface almost reaches the free run state by the end of the third year. However, below 300 m the memory of the Argo data assimilation is longer and the quality of T and S is only degraded by 35% in comparison to the full assimilation run.
Previous Abstract | Back to Abstracts Page | Next Abstract
2015 LOM Workshop, Copenhagen, Denmark June 2nd - 4th, 2015