UCLA-LANL Reanalysis Project

www.atmos.ucla.edu/reanalisys

Yuri Shprits 1

Collaborators: Binbin Ni 1, Dmitri Kondrashov 1, Yue Chen 2, Josef Koller 2, Reiner Friedel 2, Geoff Reeves 2, Michael Ghil 1, Richard Thorne 1, Tsugunobu Nagai 3

1Department of Atmospheric and Oceanic Sciences, UCLA, Los Angeles, CA

2 Los Alamos National Lab, Los Alamos, NM

Talk Outline

• Acceleration and loss processes in the Earth’s radiation belts

• Combining radial diffusion model with observations by means of Kalman filtering (performing reanalysis)

• Comparison between ensemble and exact Kalman filters

• Comparison between reanalysis obtained with Akebono and CRRES observations

• Sensitivity of the reanalysis to the assumed magnetic field model

• Summary and Conclusions

Dominant acceleration and loss mechanisms of relativistic electrons in the outer radiation belt

Losses

1) Plasmaspheric Hiss ( whistler mode waves) loss time on the scale of 5-10 days

2) Chorus waves outside plasmapause provide fast losses on the scale of a day

3) EMIC waves mostly in plumes on the dusk side very fast localized

4) Combined effect of losses to magnetopause and outward radial diffusion

Sources

1) Inward radial diffusion

2) Local acceleration due to chorus waves

Kp

ind

ex

Lif

etim

e, d

ays

Ph

ase

Sp

ace

Den

sity

Phase Space Density

Time, days Time, days

L-value Time, days

L-v

alu

e

Monotonic profiles of PSD obtained with a radial diffusion model.

Comparison of the radial diffusion model and observations, starting on 08/18/1990.

L

Radial Diffusion Model

3

4

5

6

7

-1

-0.5

0

0.5

L

Hourly averaged CRRES observations

=700 MeV G-1 K=0.11 G0.5 RE

3

4

5

6

7

-1

-0.5

0

0.5

0 10 20 30 40 500

2

4

6

8

Time, days

Kp

Make a prediction of the state of the system and error

covariance matrix, using model dynamics

Kalman Filter

fkk

fk ww 11

Compute Kalman gain and innovation vector

Update state vector using innovation vector

Compute updated error covariance matrix

Assume initial state and

data and model errors

fkk

fk ww 11 i

kfk

ak www

ik

fk

ak www

Comparison of the model with data assimilation with Daily-averaged CRRES observations.

L

Model with data assimilation

3

4

5

6

7

-1

-0.5

0

0.5

0 10 20 30 40 500

2

4

6

8

Time, days

Kp

L

Daily averaged CRRES observations

=700 MeV G-1 K=0.11 G0.5 RE

3

4

5

6

7

-1

-0.5

0

0.5

Comparison between UCLA Kalman filter approach and LANL ensemble Kalman filter

Comparison between reanalysis obtained with Akebono and CRRES observations

Ni et al.,| 2009a

Comparison Between the Radial Diffusion Model and Reanalyses

Ni et al.,| 2009a

Global Coherency

Ni et al.,| 2009a

Comparison of reanalyses obtained with various magnetic field models.

Ni et al.,| 2009b

Inacuracies associated with a choice of magnetic field model for various satellites

Ni et al.,| 2009b

Simulations with VERB diffusion code. Phase Space Density at =850 MeV/G ; K=0.025 G0.5 RE

Summary

• Data assimilation allows to blend observations from various satellites with a model, minimize errors of individual measurements and produce high resolution in time and space reconstruction of the phase space density.

• Comparison of reanalyzes from the polar orbiting Akebono and nearly equatorial CRRES spacecraft shows that data assimilation can be used to accurately reconstruct radiation belt phase space density.

• Results of the reanalysis are insensitive to a choice of magnetic field model.

• Reanalysis shows persistent peaks in phase space density which are consistent with the local acceleration processes.

• Global coherency of the radiation belt PSD indicates that pitch-angle distributions reach the lowest normal mode and decay as whole on the time-scales of a day.

Data assimilation with synthetic data produced with a radial diffusion model with Kp

L

Radial Diffusion Model with = 1/Kp

4567

-1-0.500.5

LSparce data, Radial Diffusion Model with = 1/Kp

=700 MeV G-1 K=0.11 G0.5 RE

4567

-1-0.500.5

L

Radial Diffusion Model with = 5/Kp

4567

-1-0.500.5

L

Radial Diffusion Model with = 5/Kp, data = 1/Kp

4567

-1-0.500.5

0 10 20 30 40 500

5

Time, days

Kp

204060

Data assimilation with synthetic data produced with a radial diffusion model with Kp

L

Radial Diffusion Model with = 5/Kp

4567

-1-0.500.5

LSparce data, Radial Diffusion Model with = 5/Kp

=700 MeV G-1 K=0.11 G0.5 RE

4567

-1-0.500.5

L

Radial Diffusion Model with = 1/Kp

4567

-1-0.500.5

L

Radial Diffusion Model with = 1/Kp, data = 5/Kp

4567

-1-0.500.5

0 10 20 30 40 500

5

Time, days

Kp

204060

Innovation vector

4 5 6 7-20

0

20

40

60

80

100

120

140

160

L

Inn

ova

tio

n v

ecto

r, %

of

aver

age

PS

D

Model with = 1/Kp, data = 5/Kp

4 5 6 7-100

-80

-60

-40

-20

0

20

L

Model with = 5/Kp, data = 1/Kp