Glaciers and Ice Sheet Interferometric Radar

1
Glaciers and Ice Sheet Interferometric Radar

• ESTO Mid Year Review
• January, 2007
• GSFC

2
GISIR/GISMO Team

• The Ohio State University (K. Jezek)
• The Jet Propulsion Laboratory (E. Rodriguez, A.
  Freeman)
• The University of Kansas (S. Gogineni)
• Vexcel Corporation (X. Wu, J. Curlander)
• E.G.G Corporation (J. Sonntag)
• Collaborative with Wallops Flight Facility (W.
  Krabill)
• Science team members
• University of Utah (R. Forster)
• University of New Hampshire (M. Fahnestock)

3
Briefing Overview

• Review Project Goals and Status
• Summary of May 2006 Experiment
• InSAR Results
• Multi-aperture beam processing
• Electromagnetic Modeling Status
• Year 2 Project Goals
• Objectives of April 2007 Airborne Experiment
• TRL Status
• Schedule
• Budget

4
GISMO Project Status

5
Global Ice Sheet Interferometric Radar (GISIR)
PI Prof. Kenneth C. Jezek, The Ohio State
University
Objective
Filtered basal inferogram
InSAR Concept

• Develop and test radars and algorithms for
  imaging the base of the polar ice sheets
• Investigate interferometric and tomographic
  clutter rejection and basal imaging methods
• 3-d topography of the glacial bed
• Images of subglacial conditions
• Develop multiphase center P-band and VHF radars
• Capable of sounding 5 km of ice
• Single and repeat pass interferometric operation
• Assess the requirements for extension to
  continental scale campaigns

Repeat pass tomography
Approach
Key Milestones

• Use available topography data to simulate
  interferograms for testing the InSAR and
  tomographic concepts.
• Modify the SAR simulator to include operating
  characteristics of several aircraft and several
  radar designs
• Develop UHF and VHF radars and antenna systems
• Test methodology by collecting data over the
  Greenland and Antarctic ice sheets
• Algorithm validation and sensitivity assessment.

1/ 06 Phase History Simulations and Algorithm
Testing 5/06 First flight test in Greenland
(Twin Otter 150 MHz) 7/06 InSAR algorithm
refinement 3/07 Radar and Antenna
Development 7/07 Tomography algorithm
refinement 5/07 Greenland Field Campaign (NASA
P-3) 5/08 Second Greenland Campaign (NASA
P-3) 6/08 Algorithm and methodology assessment
7/08 Requirements doc. for continental scale
imaging
Co-Is E. Rodriguez, JPL P. Gogineni, U. Kansas
J. Curlander, Vexcel Corp. John Sonntag, EGG
C. Allen, U. Kansas P. Kanagaratnam, U. Kansas
TRLin 3
http//esto.nasa.gov
6
GISMO Technical Challenges

• Obtain Swath Topography and reflectivity data
• Separate basal return from surface clutter

Surface Clutter
Weak Echoes Strong Attenuation
7
GISIR IIP Concept Evaluation Objectives

• Ice sounding performance at P-band and VHF
• SAR imaging of basal ice from aircraft
• Clutter rejection (Interferogram filtering
  tomography multi-aperture beam steering)
• Evaluation of ionospheric effects

8
Project Accomplishments

• Theoretical concept well defined
• Phase history simulations confirm theoretical
  predictions
• Radar design trade completed
• Scaling study completed
• 150 MHz radar system deployed for May 06 test
  flight in Greenland
• For the first time, SAR data acquired from
  aircraft and successfully processed to SAR images
  and interferograms of glacier bed

9
May 2006 Experiment Summary
10
May 2006 Experiment Summary and Objectives

• Flight of opportunity to acquire early GISMO data
• Data acquired using KU 150 MHz radar and WFF
  navigation equipment
• Single pass and repeat pass data acquired
• Objectives in priority order
• investigate whether data of suitable signal
  strength and with suitable knowledge of aircraft
  navigation parameters could be acquired for
  successful InSAR processing for measurements of
  basal topography and reflectivity
• evaluate phase filtering clutter rejection
  concept
• evaluate tomographic imaging concept for clutter
  rejection
• Evaluate clutter rejection using multiple antenna
  elements

11
May 06 Experiment

• Twin otter flight from Thule Camp Century
  interior
• 150 MHz Radar
• 5 transmit and 5 receive elements (1 m spacing)
• 2 m baseline outbound (achieved 8-11 m)
• Return flight offset 25 m to the south for larger
  baseline
• Maximum comfortable altitude (achieved 3000 m)
• Range window setting procedure

12
May 2006 Radar Deployment and Results
13
Background

• Radars
• Installation in Calgary
• Measure antenna pattern
• Antennas
• Twin Otter

Antenna arrays
Equipment rack
14
Background-- Specifications

• VHF Radar
• To measure thickness, map layers and image
  ice-bed interface
• Center freq 150 MHz
• Bandwidth 20 MHz
• Loop sensitivity gt 210 dB
• Depth resolution 3-4.5 m in ice
• Image
• Resolution
• Cross-track 40-100 m
• Along track 100 m
• Swath width
• Min 1000 m
• Maximum 1500 m

15
Nadir Results
16
A-scopes
17
Noise problem

• The laser was operated during the flight.
• The laser increased the noise floor by about 10
  dB every now and then (it was triggering at 5
  kHz).
• (noise analysis presented later)

18
Issues

• Calibration
• To determine imbalance in channels
• Amplitude
• Phase
• Ocean data
• We are analyzing these data
• Noise problem
• Radiated noise

19
Wet surfaces
20
Higher frequency
21

• May 2006 Navigation

22
May 06 flight route
23
Navigation Techniques

• Two navigation tools available
• Soxmap
• Used for May 2006 GISMO
• Standard Twin Otter tool
• Visual aid for flight crew
• Best for following curved path
• Course Deviation Indicator (CDI)
• Will be used for 2007 GISMO
• Can couple to aircraft steering
• Good repeatability for long straight lines

24
23 May 2006 Mission Plan

• Flight plan was out-and-back
• Thule to Camp Century, then southeast along 18
  May 1999 ATM/KU flight track
• Inbound leg offset 25 m to south of outbound
• Constant 10,000' pressure altitude

25
23 May 2006 Steering ErrorOutbound from Thule
26
23 May 2006 Steering ErrorInbound to Thule
27
24 March 2006P-3 Steering Error with CDI
28
Navigation Lessons Learned from May 2006

• 2006 GISMO flight used Soxmap / Twin Otter
  combination
• Configuration more suitable for outlet glacier
  work
• Steering within /-50m, could be better
• 2007 GISMO flights will use CDI with P-3
• Better repeatability for straight flight lines
• Soxmap backup
• More info atm.wff.nasa.gov click Aircraft
  Navigation

29
May 06 Data Processing Status
30
May 06 ExperimentFlight Route(MODIS Mosaic)
31
Range and Azimuth Compressed Slant Range Images
(log scale)5200 m along track
First Airborne SAR Images of the Base of the
Greenland Ice Sheet
Base
Left Wing
Right Wing
Internal Layers
Surface
32
Ground Range Image(5800 m along track)
1.5 km (ground range)
33
First Airborne SAR Interferogram of the Base of
the Ice Sheet
Base
surface
noise
layers
34
Filtered (left) and Unwrapped (Right)
Interferograms Intensity modulated with
coherenceFirst steps towards computing swath
topography
filtered with an adaptive spectral filter then
unwrapped with a correlation threshold of 0.45.  
Color assignment of 360 deg 1 color cycle
35
Laser Noise
36
Data Processing Steps
37
SAR Image
38
Interferograms from two adjacent channels
39
Interferogram using left wing and right wing
channel combinations
40
Ping Pong ModePing pong mode can be used to
increase the baseline but requires that multiple
transmitters and receiver systems be highly
correlated to maintain usable signal to noise
41
Range Offset SensitivityImage separation is to
small to use traditional image cross correlation
techniques so registration optimized by manually
sliding the images in range
42
Data Processing Plans

• Time reference functions
• In range compression the ideal chirp is used for
  each receive channel. We plan to measure the
  received chirp from each receive channel and use
  them to do range compression. These actual range
  reference functions may give us some improvement
  in focus and SNR in the range compressed images.
• Motion compensation
• We are quite sure the motion data are quite
  accurate for the 150 MHz carrier frequency data.
  But in April 2007 we are going to collect data
  using 450 MHz carrier frequency. Motion data may
  become less accurate relatively. We plan to use
  the current 150 MHz data to investigate motion
  compensation methods and try to find appropriate
  approaches to improve the azimuth compression for
  the special cases of the ice sounding SAR images.
• Imaging model with ice mass refraction
• Any image formation algorithms assume that the
  electromagnetic wave which carries the radar
  waveforms is traveling in the same homogeneous
  media like the air. It is not the case for ice
  sounding radars. For the data collection in May
  2006 the ice thickness is about 2000m and the
  slant range between the radar sensors and the ice
  surface is only about 1000m long. There are two
  main differences between the ice sounding radar
  and the normal surface mapping radar. The first
  one is the refraction which happens at the
  air-ice boundary and changes the travel
  directions of the electromagnetic wave. The other
  is that the travel velocity within the ice is
  about 1.8 times slower than in the free air. We
  plan to model the ice mass with two layers and
  try to improve the azimuth compression results by
  taking into account the refraction and the
  different travel velocity.
• Tomography processing
• Try to verify tomography technique for generating
  3D volumetric images of the regions of interest
  in Greenland and/or in Antarctica using the data
  acquired in May 2006 and the data yet to be
  acquired in April 2007. The methods to be tested
  include direct convolution back-projection from
  the phase history data and the method of creating
  3D images from already-formed 2D complex images.

43
May 06 Data ProcessingLessons Learned

• 1) Single pass, across track SAR imaging from
  aircraft is possible even in areas where the base
  of the ice sheet appears to be relatively smooth.
  
• 2) Across track interferometry is possible in
  the area where backscatter is relatively weak.
  This is consistent with theory. The fringe rates
  we observe are reasonable for the short (7 m)
  baseline we achieved on the Twin Otter aircraft.
• 3) Given the measured fringe rate patterns, we
  expect to retrieve across track measurements of
  basal topography.
• 4) Data processed so far steer the beam 20
  degrees off nadir. Depending on the product of
  the beam pattern with the backscatter falloff,
  this may or may not be optimum. We will analyze
  the data with different degrees of beam steering.
• 5) We did not observe fringes from the ice sheet
  surface in the most recently processed data. Yet
  we can clearly see internal layers, which should
  have a much lower backscatter value than the
  surface return. We will investigate how beam
  steering angle influences the measured
  backscatter from the ice sheet surface.
• 6) We observe detailed internal layers in the
  range and azimuth compressed data. We also
  observed the frequently described internal layer
  free zone near the base of the ice sheet.
• 7) 150 MHz backscatter strength is sufficient to
  yield a measurable signal. We will test and
  compare 150 MHz and 450 MHz systems.
• 8) The May 23 data collected observations along
  the same in and out bound track. We will
  investigate how longer baselines derived from
  repeat pass data effect data quality.
• 9) We observed a systematic noise pattern in the
  amplitude and interferometric data. The noise
  artifacts in the InSAR data will be an additional
  complication for interferogram filtering. The
  noise source is not always on and we will attempt
  to identify the origin of the noise source.
• 10) We must measure the time reference functions
  prior to the experiment.

44
Multi-aperture Beam Processing

• Approach is to manually positions nulls in the
  antenna pattern to do surface clutter rejection
• Test over Jacobshavn Glacier
• Theoretical calculations of antenna designs

45
Jacobshavn Channel at 48.45W
Focused data
A-scopes of data with arrows pointed to approx
location
46
Jacobshavn along the channel
Calving front
Calving front
Focused data
Zoomed in to highlight bed echoes
A scope with arrow pointed to approximate location
47
Lessons Learned

• Using multi-aperture arrays and spatial
  filtering, measured for the first time the ice
  thickness across Jacobshavn Glacier and to the
  calving front
• Analyses indicate that increasing the number of
  antenna elements from 4 to 10-15 at 450 MHz
  improves spatial filtering sufficiently to
  develop an automatic clutter rejection algorithm

48
Electromagnetic Scattering Model Performance
Assessment of the Global Ice Sheet Mapping
Orbiter Concept
Goal Preliminary formulation and evaluation of
scattering model.

1. Electromagnetic models for glaciers. Help answer
   the question Where is the water?
2. Physical Optics approximation
3. Simulation and results
4. Summary

49
Electromagnetic models
EM Models
surface

• Neglecting volume scattering, problem reduces to
  scattering from multi-layer rough surfaces
  (incl. ice, water and rock layers)

Near Nadir Radar
base

• Interested in modeling both deterministic and
  stochastic surfaces

• 1-D surface profiles are considered here to
  simplify the analysis methods also applicable to
  2-D but requires more computation

500 m 5 km

• Although GISMO operates at somewhat low
  frequencies, surface height variations of
  interest are on scales much larger than the EM
  wavelength.
• Near normal incidence geometry motivates
  examination of Physical Optics (PO) approximation
  plus extension to Geometrical Optics (GO) limit.

unknown depth
50
Physical Optics Approximation
PO approximation 1-D surface
One interface problem
Fields on surface estimated using a local tangent
plane approximation
Multi-layer problems Neglecting multiple
interaction, we can cascade scattering effects
from each layer.
Using a plane wave spectrum approach,
deterministic PO theory involves a set of
transition matrices coupling incident and
scattered plane waves
51
Simulation
Simulation
Surface Profile
Freq-response Time-response

• Permittivity of ice/pure water/rock labeled above
  (Debye formula/Malmberge and Maryott Model)
• Large scale surface domain split into 100m
  sections for analysis
• For each 100m-surface, scattered fields versus
  frequency computed from 140-160 MHz.
• Fourier transform provides scattered field
  amplitude envelope versus time (0-sec delay
  0-meter height)

52
Results
Results Time-response along the profile
Exact Solution - MoM
Approximate Solution - PO

• Bottom plots outline the rock (red) and water
  (blue) surfaces.
• PO solution is consistent with MoM note
  potential for observing weak scattering from
  basal rock even below pure water region
• Time resolution is limited by the bandwidth
  (20MHz) of the system.

53
Results (cont.)
Results Thin layer of water
Surface profile thin water
Time response MoM
S 0 ppt

• As water layer becomes thinner, multiple
  interactions between interfaces can be observed
  (not captured in current PO model)
• However, for water with a slight salinity (2
  ppt), returns below water surface as well as
  multiple interactions vanish due to larger
  attenuation
• Detailed salinity properties of sub-glacier water
  an important issue

S 2 ppt
54
Results (cont.)
Results Local roughness
Surface Profile with local roughness
Time response
4 meters

• local-scale power-law (k-3 spectrum) roughness
  (rms height 1m) has been added to the base-rock
  profile.
• Time response due to 6 distinct surface
  realizations are shown on the right.
• Presence of local roughness affects both signal
  strength and range estimation.

55
Summary

• PO solution for (1-D) deterministic multi-layer
  surface implemented
• PO approximation matches MoM well for reasonable
  assumed sub-glacier surface properties
• Simulations show possibility of imaging sub-water
  layer rock if water layeris pure and relatively
  thin possibility of multiple reflections also
  shown.
• Sub-water effects eliminated if water is even
  slightly saline
• As expected, dielectric contrast (i.e. presence
  of water) and local large/small scale roughness
  determine scattering amplitudes observed
• Next steps
• - Ensemble averaging for stochastic surfaces
• - Formulation for 2-D surfaces

56
Year 2 Project Goals

• Radar Development Build sub-system and assemble
  the complete system. Perform laboratory tests
  using delay lines to document loop
  sensitivity,radar waveforms and impulse response.
• System Integration (KU, WFF, Aircraft Operator)
  a)Install the radar and navigational equipment on
  P-3 or similar aircraft and conduct flight tests
  over the ocean.
• Algorithm Development. Develop a strip IFSAR
  processor and compare against the results of the
  exact time-domain processor. Iterate the clutter
  removal algorithm based on experimental results
  (JPL). Develop software and apply software to
  process multiple 2-D complex SAR images
  coherently (Vexcel).
• Data acquistion and Analysis Field experiments
  over the ice sheet Finalize interferometric SAR
  processor and pre-processor and process data from
  first campaign (JPL). Extract basal topography
  from result.. Iterate interferometric filter
  design based on assessment of the results.
• Science and Management Participate in field
  measurements Conduct design and performance
  review assess quality of results in context of
  science requirements.

57
GISMO Flights 07-08

• Plans and Objectives

58
Technical Objectives for April 07 Experiment

• 1) Acquire data over the May 2006 flight line to
  compare high and low altitude observations and to
  compare interferometry acquired with different
  baselines. Are results consistent with theory?
• 2) Acquire data at 150 MHz and 450 MHz along
  every flight line and compare backscatter and
  interferometric frequency response? Are the
  results consistent with theory?
• 3) Acquire data over areas where we expect to
  find subglacial water. Is water detectable
  either from backscatter maps or from topography?
• 4) Acquire data over regions of increasing
  surface roughness. This may require observations
  over heavily crevassed shear margins such as
  those found around Jacobshavn Glacier. Can we
  successfully implement interferogram phase
  filtering?
• 5) Acquire data for tomographic analysis
• 6) Investigate repeat pass interferometry over
  repeat periods of days.
• 7) Verify volume clutter is weak (all snow zones)
• 8) Collect data over thick and thin ice to test
  for absorption effects

59
April 07 Experiment

• P-3 flights from Thule and Kangerdlussuaq
• 150 MHz and 450 MHz Radars
• Maximum altitude allowable

60
Update to May 06 Experiment Plan
Parameter Value
Frequency 150 Mhz, 450 MHz
Band width 20 MHz, 50 MHz
Range window Start 4 us to 44 us with pulse 1 (lo-gain) Then 15 us to 55 us pulse 2 (hi-gain)
Pulse width 3 us
PRF 10 KHz (5 Khz for each pulse)
Baseline offset Return flight 25 m south of outbound flight
Calibration Rough ocean observations at these specs
Aircraft elevation above ellipsoid (geoid) 26000 ft (install additional external attunuators into the receiver
Antennas configured for two frequencies
At least one flight with multiple repeats for tomography
High elevation flights on any flights of opportunity 26,000 ft
Early evaluation of Greenland data VECO assisted DVD or electronic file transfer to KU after first GISMO flight Process to depth sounder mode Process to SAR image
61
Flight Planning Tools

• http//planet.sr.unh.edu/MOG/gismo_mog

62
(No Transcript)
63
(No Transcript)
64
Aircraft Configuration
65
P-3 Modifications

• Multiple conductors to antenna array (one
  conductor used in past experiments)
• Antenna elements modified for dual frequency
  operation
• GPS and Inertial navigation information on
  aircraft position and attitude

66
Airborne Experiment Design
67
Single Pass Interferometry
Maximize altitude Maximize antenna array
separation
6 km swath
68
Multi-Pass SAR Imaging
Synthetic Aperture
Synthetic Elevation Aperture
Ground Reference Point
69
Constraints on Flight Operations

• Fly at maximum allowable altitude
• Limit flight duration to allow for daily data Q/A
  and experiment modifications (about 6 hours
  assuming 150 Gb/hour and 3, 300 Gb disks)
• Allow enough field time to repeat flight lines
• Fly over high and low clutter areas
• Fly over areas where some information on basal
  properties is known
• VHF and UHF radars cannot operate simultaneously
  repeat P-band and VHF along same track to
  within 30 m
• Schedule 2 to 4 repeat flights at 30 m
  horizontal offsets for tomography

70
Aircraft NavigationExpected Performance

• 20 m ground track repeatability
• 0.02 degree post flight knowledge on aircraft
  roll and pitch
• 1 degree post flight knowledge on yaw

71
Proposed Flight Lines
72
Flight Description

• Each flight flown twice 150 and 450 MHz
• Flight 1 is highest priority at 450 Mhz
• Each flight is between 2000 and 2500 km
  (roundtrip)
• Flights 1 and 2 include segments over the ocean
• Flight 3 should include a segment down the
  Sondrestrom Fjord

73
Proposed Flight Lines

1. Ice Streams
2. Outlet Glaciers
3. Jacobshavn

74
Flight Details
Flight 1 Segment over open ocean Repeat segment
flown at 150 MHz in May 2006 Flights at 150 and
450 MHz Overflight of NGRIP and North East Ice
Stream Segment from GITS to Thule reflown several
times for Tomography (20-30 m Horizontal
offsets) Flight 2 Flight of NEMES drilling site
location Flights across crevassed areas of outlet
glaciers and across grounding lines Reflight of
GITs Thule segment for Tomography Segment over
open ocean Flight at 150 and 450 MHz Flight
3 Open Ocean Segment down Sondrestrom
Fjord Several passes over Jacobshavn glacier with
at least two segments separated By 20 30
m Flight over GRIP GISP drill sites Outbound at
26000 ft, Return flight at 500 feet Flights at
150 MHz and 450 MHz
75
Radars Systems
76
Background-- Specifications

• VHF Radar
• To measure thickness, map layers and image
  ice-bed interface
• Center freq 150 or 450 MHz
• Bandwidth 20 or 50 MHz
• Loop sensitivity gt 210 dB
• Depth resolution 3-4.5 m in ice
• Image
• Resolution
• Cross-track 40-100 m
• Along track 100 m

77
Existing antennas
78
Antennas

• Dipoles length
• 150 MHz
• Dipoles 14 829 mm or 32.6 in
• Dipoles 23 0.817 mm or 32.2 in
• 450 MHz
• Dipoles 14 265 mm or 10.4 in
• Dipoles 23 262 mm or 10.3 in

79
Simulations at 150 MHz

• Finite ground plane

80
Simulations
81
Experiments
82
Experiments
83
Modifications

• Dipole length
• Adjustable
• Each arm consists of two pieces
• Operate at 150 MHz with full length
• Remove
• 450 MHz
• Take out extra length

84
Plans

• Duplicate radars
• Integrate and test systems in Feb-March 2007
• Collect data over The Greenland ice sheet in
  April-May 2007

85
GISMO Flights 07-08

• Plans and Objectives

86
Arctic 07 Potential Co-Principal Investigators

• Dr. Prasad Gogineni - gogineni_at_ittc.ku.edu
• Mr. William Krabill William.B.Krabill_at_nasa.gov
• Dr. Keith Raney - Keith.Raney_at_jhuapl.edu
• GSFC Greenbelt

87
Documentation Requirements

• Flight Requests OSU and WFF completed
• Danish/Canadian Over-flight Clearances
• General Experiment Information and Request
  Worksheet (PIs) completed
• Project Plan (840)
• Mission Operations Safety Directive (840)
• ECF MOSD (548)
• Approval to Proceed (800/840)

88
Reviews / Meetings

• Airworthiness Review Board (840/548)
• Mission Readiness Review (840)
• Final Installation Inspection Review (840/548)
  Subsequent Safety of Flight Release
• Weekly/Monthly Technical Flight telecons

89
IIP TRL Objectives
Item Entry TRL Justification Exit TRL Success Criterion
IFSAR processing under ice 3 IFSAR processing has only been demonstrated for land surfaces. Imaging under ice requires new techniques to account for ray bending and ice surface. 5 Successfully image basal layer from data collected in deployments (low altitude flights)
IFSAR clutter rejection 3 Extends angle of arrival techniques to develop a new technique for clutter rejection. 5 Successfully reject clutter from high altitude flights results agree with sounder low altitude flights
Ionospheric effects 3 Calibration techniques exist for data far from nadir. They will be extended to near-nadir polar data. 5 Simulation and theoretical results to validate calibration technique
Our goal is to advance the technique to a TRL
level 5 or 6, so that our instrument could be
ready to go to a phase A/B after completion of
the IIP. We estimate that the schedule and
resources required for this are compatible with a
NASA ESSP class mission.
90
2006 TRL Assessment
Item Current TRL Progress Exit TRL Success Criterion
IFSAR processing under ice 5 Demonstrated ability to acquire SAR SLC image data from basal ice (estimate TRL 5 by end of year 2) 5 Successfully image basal layer from data collected in deployments (low altitude flights)
IFSAR clutter rejection 3 Simulations demonstrate that IFSAR filtering technique is feasible (estimate TRL 5 by year 2/3) 5 Successfully reject clutter from high altitude flights results agree with sounder low altitude flights
Ionospheric effects 3 Calibration techniques exist for data far from nadir. They will be extended to near-nadir polar data. 5 Simulation and theoretical results to validate calibration technique
91
GISIR Milestones 1/07
92
(No Transcript)
93
Budget Summary
94
Cummulative Spending
95
IPY Flight Request Pending
Campaign Aircraft Base Location Total Experiment Flight Hours Individual Flight Duration Elevation (ft) Equipment
May 2007 NASA P-3 Kangerlussuaq/Thule Greenland 50 5-7 hours 26,000 150 MHz 450 MHz U. Kansas Radar
May 2008 NASA P-3 Kangerlussuaq/Thule Greenland 50 5-7 hours 26,000 150 MHz 450 MHz U. Kansas Radar