80 likes | 216 Views
Data Assimilation of MLT (~50-110 km) observations using a 3d chemical-dynamical model in DART. Tomoko Matsuo DAI/GSP *in collaboration with Jeff Anderson(DAI), Dan Marsh(ACD), Anne Smith(ACD). Mesosphere and lower thermosphere. http://www.timed.jhuapl.edu. Scientific Objectives.
E N D
Data Assimilation of MLT (~50-110 km) observations using a 3d chemical-dynamical model in DART Tomoko Matsuo DAI/GSP *in collaboration with Jeff Anderson(DAI), Dan Marsh(ACD), Anne Smith(ACD)
Mesosphere and lower thermosphere http://www.timed.jhuapl.edu
Scientific Objectives • climatological global tidal structure • day-to-day synoptic scale tidal variability • roles of non-migrating tides and planetary waves in creating or modulating the tidal variability.
TIMED-SABER/TIDI TIDI Dayside Measurements Vector Wind O2 (0-0) P15 60 - 100 km O2 (0-0) P9 70 - 115 km OI 557.7 nm 100 - 180 km TIDI Nightside Measurements Vector Wind O2 (0-0) P9 80 - 105 km OI 557.7 nm 90 - 110 km
Data Availability http://www.timed.jhuapl.edu
ROSE 3-D chemical dynamical model [Rose and Brasseur, 1983; 1989] • Model Resolution • 38 levels (pressure coordinate) • 17.5 to 110 km by 2.5 km • 5º latitude x 11.25º longitude • 7.5 min time step • Chemistry • 27 species, 101 gas-phase rxns (JPL-2000) • Semi-lagrangian transport scheme • Airglow package • Offline D-region ion chemistry • Photolysis rates based on T U V • Dynamics • Primitive equations • Hines gravity wave parameterization • NCEP and GSWM forcing at lower boundary Tidal amplitude comparisons Tn (K) Local time
Preliminary Results from synthetic observation experiments. • MODEL • no natural error growth • large uncertainty in forcing Ensemble Spread Reduction
Need of DA system for MLT region (~50-110km) is timely. With the DART facilitation, a prototype ensemble filter assimilation system for synthetic observations with ACD's ROSE model is being constructed. Future Work: Assimilation of the ground-based and satellite observations (15 k scalar observations per TIMED orbit). Estimation of forcing and model parameters Challenges and Open Questions: How to cast a DA problem in strongly forced and dissipative systems when models do not have natural error growth? Model Error v.s. Observation Error: Observed day-to-day variability is significantly higher than the variability reproduced by numerical models. Large uncertainty in forcing Summary