Transcript 886 - SuperDARN

Computational Challenges in
Space Research
Joseph B.H. Baker
Bradley Department of Electrical and Computer Engineering
Center for Space Science and Engineering Research (Space@VT)
Virginia Tech
Joseph Baker ([email protected])
Space@VT
CS Seminar April 20th 2012
Presentation Outline

What is Space Science?

The Center For Space Science and Engineering Research (Space@VT)


Computational Challenges in Space Research:

Data Mining Space Physics Datasets (SuperDARN)

Compression Algorithms for Small Satellite (CubeSat) Applications

Numerical Simulations of the Space Plasma Environment

Atmospheric Modeling, Assimilation, and Retrieval
Summary
Joseph Baker ([email protected])
Space@VT
CS Seminar, April 20th 2012
What is Space Science?
Joseph Baker ([email protected])
Space@VT
CS Seminar, April 20th 2012
The Sun-Earth System
Space science is the study of the outflow of the solar wind from the sun’s
Joseph Baker ([email protected])
Space@VT
CS Seminar,
April 20th 2012
atmosphere
and its interaction with
the earth’s magnetic
environment.
The Sun-Earth System
Joseph Baker ([email protected])
Space@VT
CS Seminar, April 20th 2012
The Geospace System

Dynamics in the space environment has impacts on upper atmosphere phenomena:
ionospheric electric fields and currents, plasma structuring, winds, waves, pulsations, etc.
Joseph Baker ([email protected])
Space@VT
CS Seminar, April 20th 2012
“Space Weather”
Perturbations to the geospace environment can have huge impacts on society:
Joseph Baker ([email protected])
Space@VT

Spacecraft damage

Atmospheric drag

Power disruption

Human Health (radiation)

Deviation of airplanes

Degraded navigation

Disrupted communication

Radar clutter

Confused pigeons!
CS Seminar, April 20th 2012
Space Instrumentation
 The space environment is divided into various “spheres” (e.g. helio, magneto,
iono, thermo, strato, etc.) and obtaining sufficient measurements is a challenge.
Joseph Baker ([email protected])
Space@VT
CS Seminar, April 20th 2012
What is Space@VT?
Joseph Baker ([email protected])
Space@VT
CS Seminar, April 20th 2012
Space@VT

The Center for Space Science and Engineering Research (Space@VT) is
relatively new at Virginia Tech, having become recognized as an official
College of Engineering Research Center in September 2007.

MISSION: To provide forefront research, instruction, and educational outreach
in the fields of space science and engineering utilizing a holistic approach of
theoretical modeling, advanced simulation techniques, space system and
instrument design, and experimental data acquisition, analysis and interpretation.

Space@VT comprises 13 core faculty and ~30 postdocs and graduate students.

The building under construction at the edge of the Corporate Research Center
(CRC) will be the new home for Space@VT in summer 2012.

Annual Research Expenditures of Space@VT are approximately $4 Million.
Joseph Baker ([email protected])
Space@VT
CS Seminar, April 20th 2012
Space@VT Core Faculty
Faculty Member
Department
PhD
Dr. Scott Bailey
Dr. Joseph Baker
Dr. Xai Cai
Dr. Robert Clauer
Dr. Greg Earle
Dr. Chad Fish
Dr. Ray Greenwald
Dr. Troy Henderson
Dr. Bob McGwier
Dr. Mike Ruohoniemi
Dr. Wayne Scales (Director)
Dr. Kevin Shinpaugh
Dr. Dan Weimer
ECE
ECE
ECE
ECE
ECE
ECE
ECE
AOE
ECE
ECE
ECE
AOE
ECE
Colorado
Michigan
Michigan
UCLA
Cornell
Utah State
Dartmouth
Texas A&M
Brown
Western Ontario
Cornell
Virginia Tech
Iowa
Joseph Baker ([email protected])
Space@VT
CS Seminar, April 20th 2012
Space@VT Research
•Space-Based Observations and Instrument Development
•Ground-Based Autonomous Magnetic Observatories
•SuperDARN HF Space Weather Radars
•Space Plasma Environment Simulations
•Space Weather Data Assimilation and Models
•Spacecraft Modeling, Simulation and Design
Joseph Baker ([email protected])
Space@VT
CS Seminar, April 20th 2012
Spacecraft Observations
Aeronomy of Ice in the Mesosphere (AIM) Satellite
• NASA Small Explorer in third year of observations
Small Explorer Satellites
University Class Satellites
Joseph Baker ([email protected])
Sounding Rockets
Space@VT
PolarNOx Sounding Rocket:
• Brief space based observation
•Test bed for HPAC technology
•Experiment built at Va Tech
• Joint with U. Colorado
• Launched in January 2011
CS Seminar, April 20th 2012
Autonomous Observatories
•Funding from NSF MRI and Polar Programs to build a chain of autonomous space
weather observing stations (magnetometer and GPS receiver) in Antarctica.
•The new stations will be magnetically conjugate to similar instrumentation in Greenland,
thus allowing studies of space weather effects at both ends of a single magnetic field line.
Joseph Baker ([email protected])
Space@VT
CS Seminar, April 20th 2012
SuperDARN HF Radars
• The Super Dual Auroral Radar Network (SuperDARN) is an international space weather
radar system that monitors the motion of plasma in the Earth’s ionosphere.
• At the present time there are 25+ radars operational world wide.
• Virginia Tech is the PI institution for five SuperDARN radars in the US and Canada.
Joseph Baker ([email protected])
Space@VT
CS Seminar, April 20th 2012
Space Plasma Simulations
1. Modeling of space plasma
turbulence due to chemical
releases in space for
controlling space weather
processes. (ONR, DARPA)
2. Modeling of artificial
heating of natural dust
layers in the near earth
environment for diagnostic
purposes. (NSF)
3. Modeling of creation,
evolution and associated
plasma turbulence of
artificial dust layers in the
near earth environment.
(ONR , DARPA)
•The effort utilizes advanced computational methods and high performance computing.
Joseph Baker ([email protected])
Space@VT
CS Seminar, April 20th 2012
Space Weather Models
Analogous to “earth weather” models
that assimilate and predict
Space Weather
Earth Weather
Courtesy: Utah State University GAIM
Earth weather models assimilate and
predict
ground
temperatures,
precipitation, winds, pressure
Joseph Baker ([email protected])
Space weather models assimilate and
predict upper atmospheric plasma density,
temperature, wind, electric fields, solar
particle precipitation
Space@VT
CS Seminar, April 20th 2012
COMPUTATIONAL CHALLENGE #1
Data Mining Space Physics Datasets
Dr. Joseph Baker
Dr. J. Michael Ruohoniemi
Dr. Naren Ramakrishnan
Joseph Baker ([email protected])
([email protected])
([email protected])
([email protected])
Space@VT
CS Seminar, April 20th 2012
Motivation

In recent years, the sheer volume and complexity of space physics datasets
has quickly grown to become beyond the reach of the traditional data analysis
tools used by most researchers with space physics training.

At the same time, there has been an increased emphasis on “system science”
themes: interdisciplinary investigations; interconnectedness; cross-scale
coupling; complexity, etc.

Further progress requires an injection of Knowledge and Data Discovery
(KDD) techniques from the computer science community.

Over the past 18 months Space@VT faculty have partnered with Naren
Ramakrishnan in CS on a project to apply advanced KDD techniques to
extract complex features from the SuperDARN HF radar database.
Joseph Baker ([email protected])
Space@VT
CS Seminar, April 20th 2012
The SuperDARN Radars
The Super Dual Auroral Radar Network (SuperDARN) is an
international network of high-frequency (HF) radars for researching the
Earth’s upper atmosphere, ionosphere, and connection into geospace.
Joseph Baker ([email protected])
Space@VT
CS Seminar, April 20th 2012
SuperDARN Targets
B-Field Line
Ionospheric
plasma
irregularities
Ionosphere
• HF signals are refracted in the ionosphere by gradients in electron density
• Backscatter targets for HF radars include:
1) Field-aligned ionospheric plasma irregularities (“half-hop” mode)
2) The ground (“one-hop” mode)
3) Meteor trails at mesosphere altitudes (near-range echoes)
Joseph Baker ([email protected])
Space@VT
CS Seminar, April 20th 2012
1-1/2 hop
Time
1-1/2 hop
1/2 hop
Ground Scatter
1/2 hop
Width
Single beam range-time series (Power, Doppler, Spectral Width)
Polar Potential
Sun
Radar Field-of-View Doppler Map
Doppler
Range
2-minute scan
Power
SuperDARN Data Views
Multi-Radar Convection Map
Joseph Baker ([email protected])
Multi-Radar Key Parameter Index (i.e. Cross-Polar Potential)
Space@VT
CS Seminar, April 20th 2012
SuperDARN Data Products
I = ionospheric scatter, G = ground scatter, M = meteor scatter, VD = Doppler velocity, P = Power, WD = spectral width
SuperDARN data is widely used by geospace scientists to specify ionospheric electric fields
(item-1); however, other more exotic capabilities are under-utilized (items 2-13).
Joseph Baker ([email protected])
Space@VT
CS Seminar, April 20th 2012
Atmospheric Gravity Waves
Range Gate
Gravity Wave

At ionospheric altitudes gravity waves produce
structuring in ionospheric plasma density that appears
in SuperDARN data as quasi-periodic enhancements in
backscatter from the ground [Bristow et al, 1994].
Gravity waves are an important mode of atmosphereionosphere coupling but reliable identification and
extraction of their signatures from the SuperDARN
data stream is a significant computational challenge.
Joseph Baker ([email protected])
Space@VT
Time

One under-utilized SuperDARN capability is the ability
to monitor atmospheric gravity waves (AGWs).
Universal Time

Ground Scatter
Range
Power (dB)
CS Seminar, April 20th 2012
Event Detection

A traditional approach to automated gravity wave event detection might use a Fast
Fourier Transform (FFT) to band-pass the data for gravity wave frequencies.

However, there are a number of complications with this approach:

SuperDARN data exhibits a high degree of spatio-temporal sparseness, so some
sort of interpolation in space and time is required before an FFT can be applied.

Gravity waves propagate through a radar field-of-view and so analysis at a fixed
range-beam location in space is unsatisfactory.

Gravity waves are subtle features which are often masked within the radar fieldof-view by more dominant features (e.g. auroral zone convection).

For these reasons, previous studies of gravity waves in SuperDARN data have tended
to be “event” driven; that is, focused on manual analysis of the most intense
signatures that are easily detected by eye and tend to occur during active conditions.

At the present time, there has not yet been a full statistical characterization of gravity
wave activity in the SuperDARN dataset covering all levels of geomagnetic activity.
Joseph Baker ([email protected])
Space@VT
CS Seminar, April 20th 2012
Spatiotemporal Process
Discovery Approaches
Interpolation by “Data Driven Surrogates”
Feature Extraction using “Spatial Aggregates”
Activity Hotspot identification by “Subclustering”
Joseph Baker ([email protected])
Recurrent Pattern Identification by “Motif Mining”
Space@VT
CS Seminar, April 20th 2012
AGW Detection Algorithm

BASIC APPROACH: Use temporal “Motif Mining” to identify gravity waves as
recurrent events (i.e. “motifs”) with broadly similar (but not identical) features.

COMPLICATION: Motif mining is generally applied to simple time series data
(either univariate or multivariate) and so the basic framework had to be extended
to work on sparse two-dimensional time-series data.

ALGORITHMIC STEPS:

Thresholding detects enhanced backscatter at each radar range

Cluster analysis identifies temporal hot-spots of backscatter at each range

Edge detection ties together the backscatter clusters over multiple ranges

Motif mining identifies transitions in the clustered data (i.e. wave fronts)
Joseph Baker ([email protected])
Space@VT
CS Seminar, April 20th 2012
Raw Range-Time Data
Range Gate
Power (dB)
Range
Time
Universal Time
Joseph Baker ([email protected])
Space@VT
CS Seminar, April 20th 2012
Gravity Wave Detections
Time
Range Gate
Range
Gravity Wave Fronts
Universal Time
Joseph Baker ([email protected])
Space@VT
CS Seminar, April 20th 2012
COMPUTATIONAL CHALLENGE #2
Compression Algorithms for CubeSats
Dr. Troy Henderson
Dr. Greg Earle
Dr. Chad Fish
Joseph Baker ([email protected])
([email protected])
([email protected])
([email protected])
Space@VT
CS Seminar, April 20th 2012
CubeSats
CubeSats are modular satellites with dimensions of 10 cm on a side
Advantages:
•Can piggy-back rides on rockets carrying larger spacecraft
•Perfect for academic institutions and training students
•Well suited for space science missions
Joseph Baker ([email protected])
Space@VT
CS Seminar, April 20th 2012
The Trouble with CubeSats



Orbit and attitude control are critical functions for any
space mission and require sophisticated algorithms that
have been developed and refined over many years.
Miniaturization of CubeSats has lowered the threshold
for access to space but it presents a number of
challenges because of limited computational resources
and high operational rate (~100 Hz).
There is a critical need for advanced compression
algorithms to modify existing flight software for:
 Orbit Determination and Control
 Attitude Determination and Control
 Interfacing with hardware
 On-orbit processing of science data
 Determining errors
 Transmitting data to the ground (telemetry)
Joseph Baker ([email protected])
Space@VT
CS Seminar, April 20th 2012
CubeSat Mission: LAICE
LAICE – Lower Atmosphere Ionosphere Coupling Experiment
O2 Photometers
RPA
Pressure
Gauges
LAICE Satellite Design
Experiment Goals:
1. Measure waves in mesospheric airglow
remotely using on-board photometers
2. Measure higher altitude neutral pressure
and plasma density variations in LEO.
3. Correlate in-situ and ground observations to
investigate how waves are related to HF
and UHF scintillation activity.
4. Validate a microtip emitter-based ion gauge.
Mission Overview:
Telemetry Issues:
1. RPA and pressure gauge aperture point
in the ram direction.
2. Photometers (U of I) are nadir viewing.
3. Uplink commands and downlink data
using amateur radio band groundstations at Virginia Tech and U of I.
1. TI OMAP L-137 processor
2. 64 MB ram, ~1 GB program and data
storage, I2C on-board interface
3. AFSK or QPSK modulation
4. 6.5 - 8 Mb/day downlink requirement
5. Two ground stations yield ~20 minutes
contact/day, or 2.9 Mb/day at 2400 bps.
6. On-board data compression req’d.
Joseph Baker ([email protected])
Space@VT
CS Seminar, April 20th 2012
Falling Sphere Sensor
In-situ measurements of high altitude winds (~ 150 km altitudes)
• Similar telemetry issues apply to a “falling
sphere” dropped from a sub-orbital rocket
to measure atmospheric winds.
• The winds are measured by very sensitive
accelerometers inside the sphere.
• Data is telemetered to the ground and later
processed to retrieve wind profiles
• COMPUTATIONAL CHALLENGE:
Processing involves the solution of
multiple first and second order
equations that govern (1) the state of
the atmosphere, (2) interaction of the
sphere with the atmosphere, and (3)
the attitude of the sphere during flight.
Joseph Baker ([email protected])
Space@VT
CS Seminar, April 20th 2012
Orbit Propagation


Satellites in different orbits have different dynamics
We can take advantage of that to build constellations over time by solving the
orbit equation:


r 

r
3
r  ad
where ad is representative of every perturbation in the universe!
CHALLENGE: Generally have less computational power than your cell-phone!
Joseph Baker ([email protected])
Space@VT
CS Seminar, April 20th 2012
Attitude Control

“Attitude Control” refers to the pointing
of a spacecraft or particular sensor

Determination of the attitude amounts to
solving a first-order differential equation
involving dynamic perturbations and
applied torques:






Over-determined system (> 4
measurements for 3 unknowns)
May involve image processing
Uneven spacing in receiving data
Kalman filtering techniques
Error determination
Command controls
Joseph Baker ([email protected])
Space@VT
CS Seminar, April 20th 2012
COMPUTATIONAL CHALLENGE #3
Simulations of Space Plasma Environments
Dr. Wayne Scales
Joseph Baker ([email protected])
([email protected])
Space@VT
CS Seminar, April 20th 2012
Space Plasma Simulations
Capabilities:
Modeling physical processes in the near earth space environment with
physics-based models from a space plasma physics formulation
Goals:
a. Understand basic nonlinear physics such as space plasma turbulence
b. Understand the impact on modern technologies such as communication
and navigation systems
c. Exploitation for remote sensing/diagnostic information of the near earth
space environment
Joseph Baker ([email protected])
Space@VT
CS Seminar, April 20th 2012
Virginia Tech Plasma Physics Model
(e.g. Scales et al. 2010: Bordikar and Scales 2012)
Electrons (Fluid)
Continuity:
Momentum:

ne
dne
   (ne ve )  Q  L 
t
dt
ch arg ing
Ions (Fluid)

ni
dni
   (ni vi )  Q  L 
t
dt
 

EB
ve 
B2


 qi 
vi
T 1
 (vi  )vi 
E   i ni
t
mi
mi ni
ch arg ing
Dust (Particle-in-Cell PIC)

dx j
dt

dv j

 vj
Dust Charging
Q j (t )   

( E  v j  B)
dt
mj
dQ j
dt
 I ej  I ij
Poisson’s Equation
 02  e(ne  ni  Zd nd ) ;
E  39
Ionospheric Heating
5-10 km
Ionospheric Interaction Region (~200 km)
High Frequency Active Auroral Research
Program (HAARP) facility in Gakone, Alaska
HAARP
Transmitter
(high power HF
transmitter)
echo
Diagnostic radar
(Diagnostic information due to reradiated electromagnetic energy)
Joseph Baker ([email protected])
Space@VT
3.6 Mw transmitter power !!!
CS Seminar, April 20th 2012
Model-Data Comparisons
Experimental Data From HAARP
Nonlinear Plasma Simulation Model
•Good agreement is obtained between the experimental data, theory and nonlinear plasma
simulation model developed at Virginia Tech (Scales et al. 2011, Samimi et al. 2012).
•The sideband structures provide diagnostic information about ionospheric density, ion
species, temperature, and plasma turbulence characteristics and provide predictive
capabilities for communication/navigation systems.
•The simulations currently take about 1 month to run on a single processor (3GHz PC).
•More realistic simulations will take 1 – 2 orders of magnitude more computational time
and must be parallelized. Collaborations are needed.
Joseph Baker ([email protected])
Space@VT
CS Seminar, April 20th 2012
Atmospheric Dust
Courtesy of Dr. P. Bernhardt
NRL CARE Sounding Rocket Campaign
•Artificial dust layers are created in the near earth space environment by sounding rocket burns.
•Artificial layers provide insight into:
• Natural dust layers linked to global climate change
•Plasma turbulence effects on communication/navigation systems
•De-orbiting spacecraft with the dust debris clouds.
Joseph Baker ([email protected])
Space@VT
CS Seminar, April 20th 2012
Simulated Plasma Turbulence
ymagnetic field
X (expansion direction)
• Broad spectrum of turbulence is produced from VLF to ULF during the dust cloud release.
• Computational models must be developed to study turbulence in each frequency regime
since temporal/temporal scales are so different.
• Virginia Tech has developed computational models to investigate several of these but the
current picture is far from complete
• Current FORTRAN codes must be parallelized with better computer resources to provide
more realistic models of the actual experiments.
Joseph Baker ([email protected])
Space@VT
CS Seminar, April 20th 2012
COMPUTATIONAL CHALLENGE #4
Atmospheric Modeling, Assimilation and Retrieval
Dr. Chad Fish
Dr. Dan Weimer
Joseph Baker ([email protected])
([email protected])
([email protected])
Space@VT
CS Seminar, April 20th 2012
Space Weather Models
Courtesy: ASTRA AIME

The Global Assimilation of Ionospheric Measurements (GAIM) uses Kalman
filtering to combine real-time measurements (>300) with physics-based models to
produce realistic specifications of ionospheric conditions.

Large data sets and iterated simulations of differential and coupled equations
requires significant computational power (10s of CPUs in parallel) to produce high
fidelity forecast solutions.
Joseph Baker ([email protected])
Space@VT
CS Seminar, April 20th 2012
Atmospheric Retrievals
Develop atmospheric profiles of physical parameters and constituent concentrations
(e.g., altitude versus concentration of the green house gases of carbon dioxide and
methane). Very important for climate modeling and earth weather.
Courtesy: P. Y. Foucher et al., 2011
Courtesy: GATS SOFIE
Retrieval problem (or inverse problem) of
multiple iterations using best guess profiles and
then weighting functions (kernels, eigenvalues)
until a solution is within a certain error value
compared to measured radiances
Joseph Baker ([email protected])
Space@VT
CS Seminar, April 20th 2012
Summary


Hopefully, you should now have a better understanding of:

Space Science and Space@VT Research Activities

Computational Challenges associated with:

Data Mining Space Physics Datasets

Compression Algorithms for CubeSat Applications

Numerical Simulations of the Space Plasma Environment

Atmospheric Modeling, Assimilation, and Retrieval
Collaborations are welcome!
Joseph Baker ([email protected])
Space@VT
CS Seminar, April 20th 2012
Joseph Baker ([email protected])
Space@VT
CS Seminar, April 20th 2012