1. 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)

2. Mesosphere and lower thermosphere

3. 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.

4. TIMED-SABER/TIDI TIDI Dayside Measurements
Vector Wind O2 (0-0) P km
O2 (0-0) P km
OI nm km
TIDI Nightside Measurements
Vector Wind O2 (0-0) P km
OI nm km

5. Data Availability http://www.timed.jhuapl.edu Horizontal Axis
# of ground-based radars
Total # of observations (50K) per
Given time scale, assimilation window: one orbit a time (90-100min)

6. 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
Chemical and dynamical time scale.
Chemistry has traditionally been a primary diagnosis of the region.
Size of state vector (400K)
Tn (K)
Local time

7. Preliminary Results from synthetic observation experiments.
MODEL
no natural error growth
large uncertainty in forcing
Ensemble Spread Reduction
Ensemble mean and spread
MODEL (no error growth & large uncertainty in forcing)
OBSERVATIONS (the huge variability)
COVARIANCE
DART facilitation
how quickly a given model can be implemented into DART

8. Summary 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
pilot study for WACCAM_DART
CAM’s gravity breaking parameterization,