(1700) CROWLEY Dragster - Space Weather Prediction Center

Dr. Geoff Crowley

M. Pilinski, Eric Sutton, M. Codrescu

T. Fuller-Rowell, Mariangel Fedrizzi, S. Solomon, L. Qian, J. Thayer

Atmospheric & Space Technology Research Associates LLC www.astraspace.net

REDUCING CONJUNCTION ANALYSIS ERRORS WITH AN ASSIMILATIVE TOOL FOR SATELLITE DRAG SPECIFICATION

orImproved Orbit Determination and Forecasts

with an Assimilative Tool for Atmospheric Density and Satellite Drag

Satellite Aerodynamics

Modeling

Ground-based

Instrument

Development

Data

Assimilation

Data

Services

Ionospheric Electron Density

Physics-BasedModeling

(TIMEGCM)

High-latitudeElectrodynamics

Space Based Data

Ground Based Data

HF TID Mapper

Space

Systems

GPS-based Space Weather Monitor

CubeSat Instruments

Scanning

UV Photometer

E-field Double Probe

GPS-based Space

Weather Monitor

RF Waves & Sounder

Wind Profiler

CubeSat Missions

NASA: SORTIE &

MiRaTa

AF: DIME, SIPS & TSS

NSF: DICE & LAICE

Plug-N-Play Avionics

Hosted Payloads

ASTRA: Space Weather Focus

ThermosphericNeutral Density

Lidar Systems

E-fields andMagnetometers

Forensic Space Weather Analysis

Real-Time Specification

of Ionosphere/

Thermosphere

Low Power Ionospheric Sounder

Magnetometer &

Langmuir Probe

Celebrating our

12th Anniversary

Satellite Drag & Ballistic

Coefficients

3

Where is the Thermosphere?

What is satellite drag

4

Winds

Motivation for Neutral Density and Satellite Drag Specification

5

Satellite drag errors degrade ability to:• Maintain accurate catalog of all space objects• Predict and avoid space collisions• Predict satellite reentry time & location

from Picone et al. 2005

4/2/2017 6

Dragster Goals

Improve the state of the art in orbit nowcast & prediction, and conjunction analysis for LEO satellites by reducing the errors associated with atmospheric drag modeling

• Dragster consists of ensemble of world-class well-validated full-physicsatmospheric models combined with ensemble of well-validated empiricalmodels

• Dragster is assimilative: ingests drag information from resident spaceobjects

Improve versus:o JB08 - empirical modelo HASDM – assimilative model

• In other words…Outperform HASDM for time-integrated density for a selection ofvalidation objects in a variety of orbits

Super-Ensemble Approach

Image credit: TerraMetrics, Google

7

4

Dragster top level design

9

TIEGCM, TIMEGCM, CTIPe, Empirical Models

Ensemble Kalman Filter

Design Features

• Multiple model (super-ensemble) approachallows for graceful degradation in case of input-stream or model interruption

• Inclusion of TIME-GCM allows for specification ofdensities in the re-entry regime, down to 30km

• Inclusion of Helium in several models allows fordrag computation up to 1500km

• Dynamically tuned (Kzz) models result inoptimum background atmospheric state

10

Start with a Background Model

11

Local time, latitude distribution of neutral density at 400 km

Assimilate/Test with Orbital Data

Local time and latitude distribution of assimilation and validation satellites 12

Preliminary Validation

• Assimilating orbital data from approximately 75 objectswith perigees between 200 and 700km altitudes (this isconfigurable)

• Processing TLE information for this experiment

• Data is assimilated in a 36 hour window and thewindow is advanced at 12 hour intervals (this isconfigurable)

• THESE RESULTS ARE PRELIMINARY: We are in theprocess of expanding the validation to other years,satellites, and data-types.

13

Test-data coverage:focus on well-characterized objects

14

…and other, already flying, resident space objects

DANDE#39267

SL-3 rocket body outline#04814

SPINSAT#40314

PAM-D rocket body outline#28476

SORCE#27651

OV3-1#02150

Overview of Work:Sample Test Results

6

NRLMSISE-00

Dragster model (bright green above) outperforms all other models, including HASDM (gold), by matching most closely the observed

densities (black diamonds) for this validation object.

GRACE Accelerometer

10

Dragster more closely matches GRACE data, predicting subtle variations in Density (blue arrows) vs. other models - including the operational HASDM.

Small Scale Density Features: Comparison with Satellite Accelerometers

Overview of Work: Sample Test Results

7

These results were generated by assimilating one year of real data from 75 LEO satellites.

Dragster attains superior performance by solving both state corrections (like densities) and model parameter corrections (forcing).

The latter enables the data assimilation to have a much more global impact while remaining physically realistic. This also enables better use of first principles models.

This approach has not been possible with current operational methods.

Overview of Work: Sample Test Results, Orbit Data Validation

8

Satellite Name (Altitude)

Model Standard Deviation

SORCE

MSIS 28%

JB08 20%

HASDM 21%

Dragster 10%

SpinSat

MSIS 21%

JB08 11%

HASDM 5%

Dragster 6%

DANDE

MSIS 15%

JB08 10%

HASDM 10%

Dragster 6%

GRACE-A

MSIS 22%

JB08 11%

HASDM 5%

Dragster 5%

Dragster outperforms or closely matches three

leading atmospheric models including the operational

HASDM. (lower SD is better)

Conclusions: Dragster Benefits

• In preliminary tests, Dragster outperforms several atmosphericdensity models. This is in-spite of using only publicly availableorbits!

• Tests including General Circulation Models and SpecialPerturbation Orbit solutions are ongoing.

• When comparing to the current assimilative density standardfor orbit prediction (HASDM), Dragster performance is better orequivalent in the spectral domain of the input data

• Dragster spatial resolution improves over that of other drag-assimilative models and can be set by the user.

• Dragster approach is compatible with current operationalindices and datasets but can be easily extended to potentialfuture operational datasets

18

4/2/2017 20

Options for model inputs:A flexible approach

Data Type Notes

Orbit Average Dragi.e. Calspheres, DANDE,POPACS

Infer observed energy dissipation rate (EDR) from general perturbations (TLEs) or special perturbations (high task tracking data). Select 30-90 RSOs with stable ballistic coefficients.

HASDM Densities EDR is inferred from HASDM density outputs

Orbit Average Densities Already processed high-task tracking data

Orbit Resolved Drag: GPS Observed EDR from special perturbations and GPS measurements

Orbit Resolved Drag: accelerometers (GOCE)

Observed acceleration at 10-45 sec cadence (in-track and cross-track), binned to 15 min

O/N2 (GOLD) Dayside disk composition

Mass Spectrometer In-situ day and night composition

Assimilated Data Types

194/2/2017

Future WorkFocus in current tests

Overview of Work: Sample Test Results, Orbit Data Validation

Fig. 1: Standard deviation errors for all validation objects

relative to JB08 standard deviations. Values above the dotted line indicate performance worse than JB08 while values below the dotted line indicate performance better

than JB08. 9

Perigee Altitude [km]

Mod

el S

tand

ard

Devi

atio

n di

vide

d by

Sta

ndar

d De

viat

ion

of JB

08

200 300 600 400 500 700 0.0

0.5

1.0

1.5

2.0

Input Data Includes Aerodynamic Panel Models

M. Pilinski

SORCE

C/NOFS

Allows Dragster to assimilate datafrom objects whose A/m ratios arenot constant. This means Dragstercan ingest more data.

GRACE

22

Dragster Capabilities

2/22/2017 G. Crowley, M. Pilinski, G. Thompson 26