220 likes | 378 Views
Overview of my past research activities in atmospheric and oceanographic modelling. Srdjan Dobricic Air and Climate Unit, JRC. Outline. Simulation of the spreading of the desert dust Coupled atmosphere-ocean model for the Adriatic North Atlantic sea level response to buoyancy forcing
E N D
Overview of my past research activities in atmospheric and oceanographic modelling • SrdjanDobricic • Air and Climate Unit, JRC
Outline • Simulation of the spreading of the desert dust • Coupled atmosphere-ocean model for the Adriatic • North Atlantic sea level response to buoyancy forcing • An oceanographic data assimilation scheme • Bayesian Hierarchical Modelling in data assimilation • Global ocean monitoring and forecasting system with a very high horizontal and vertical resolution. • Discussion
Short term prediction of the spreading of desert dust • A module for the spreading of the desert dust was added into the ETA limited area model • The major sensitivity was found in the specification of the input of the dust into the atmosphere • For the first time the input was defined at a threshold velocity that lifts the dust from the ground and creates surface concentration • The desert soil type was specified only for valleys
Short term prediction of the spreading of desert dust • Vertically integrated dust concentration (g/m2) • Dashed lines show satellite observations of 0.2 g/m2
Coupled atmosphere-ocean model for the Adriatic ETA atmospheric model (25 km) ISPRAMIX ocean model (5-10 km resolution)
Coupled atmosphere-ocean model for the Adriatic • The evaluation made by using in-situ and satellite observations: • The coupling had a strong impact on the ocean model, especially on near-surface parameters • The air-sea fluxes (temperature and evaporation/precipitation) were significantly improved • There was no significant impact on the atmospheric model (as measured in comparison with coastal observations)
North Atlantic sea level response to buoyancy forcing • On average the ocean steady state is in the geostrophic balance. • The currents are strong at the surface and weak in deep layers (but not everywhere). • What determines the mean sea level distribution and general circulation given the density distribution (which changes slowly)?
Sea level (m) Full buoyancy forcing Compensated buoyancy forcing Uncompensated buoyancy forcing North Atlantic sea level response to buoyancy forcing Transport (Sv)
The European GMESMarine Core Service – Mediterranean Sea Satellite and in situ data Geoportal For Data Access The Marine Core Service delivers regular and systematic reference information on the state of the oceans and regional seas of known quality and accuracy
The European GMESMarine Core Service – Mediterranean Sea • Hydrodynamics (INGV) • Horizontal res: 1/16°x1/16° ~ 6.5Km • Vert res: 72 unevenly spaced levels • Lateral open boundary condition in the Atlantic: nesting in Mercator global model • Pelagic Biochemistry (OGS) • Horizontal res: 1/16°x1/16° ~ 6.5Km • Vertical res: 72 unevenly spaced levels • Offline Coupling with MFS physics • 10 days forecast every three days
OceanVar finds the minimum of a cost function linearized around the background state: An oceanographic data assimilation scheme Preconditioning using a control vector v defined by:
An oceanographic data assimilation scheme is modelled as a sequence of linear operators: - Vertical EOFs to physical space. • Diagnose u and v. - Horizontal covariances. - Divergence damping filter. - Barotropic model.
A barotropic model estimates covariances between T/S perturbations and sea level. It finds the stationary solution of free-surface equations forced by constant perturbations of salinity and temperature: An oceanographic data assimilation scheme
SeaLevelAnomalymisfits: 10 yearsof increaseot the quality MOM1.1 +SOFA OPA8.2 +SOFA (sys2b) OPA8.2 +3DVAR (sys3a2) NEMO+3DVAR (sys4a) rmse rmse
Assimilation of trajectories An oceanographic data assimilation scheme Day 1 Day 10
Summer Winter Bayesian Hierarchical Modelling in data assimilation Temperature (full line) and salinity (dashed line) error covariances with temperature error at 10 m (green – control, black – BHM).
Bayesian Hierarchical Modelling in data assimilation Relative change of RMS of misfits: cyan – control, blue – no d, red – with d
Ocean and ice models: • Global tripolar grid (5600*3200*100 points) • Specification of initial and forcing fields • OceanVar: • Fast interpolation on the non-regular grid and cyclic boundary conditions • Third order accurate recursive filter for simulating horizontal error covarinaces • The exact coastal boundary condition instead the use of ghost points • Computational time on 750 cores: • Model: 60 minutes (possible to reduce to 45 minutes) • OceanVar: 7-10 minutes (reduced from 60 minutes) Global ocean monitoring and forecasting system with 1/160 horizontal resolution
Intensity of surface currents (m/s) Global ocean monitoring and forecasting system with 1/160horizontal resolution
Intensity of surface currents (m/s) Global ocean monitoring and forecasting system with 1/160horizontal resolution