RADIOMETRY A W Tony England Hamid Nejati and Amanda Mims University of Michigan Ann Arbor Michigan US A IGARSS 2011 Outline Intro to global snowpack sensing Limitations of current snowpack sensing technologies ID: 502846
Download Presentation The PPT/PDF document "1 SNOWPACK AND FRESHWATER ICE SENSING US..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
1
SNOWPACK AND FRESHWATER ICE SENSING USING AUTOCORRELATION RADIOMETRY
A. W. (Tony)
England, Hamid
Nejati
, and Amanda Mims
University of Michigan, Ann Arbor, Michigan, U.S.
A
IGARSS 2011
Slide2
Outline
Intro to global snowpack sensingLimitations of current snowpack sensing technologiesPotential of Wideband Autocorrelation Radiometry (Wideband AR) for snowpack sensing
Demonstrate concept through simulation
Summary of Wideband AR’s advantages
Wideband AR’s challenges
2Slide3
Intro to Global Snowpack Sensing
ApplicationsWeather prediction and climate monitoringWater resource managementFlood hazard predictionDesired coverage
Near-daily of all snow-covered terrains including snowpacks on major ice sheets
Snowpack characteristics of interest
ThicknessSnow Water Equivalent (SWE)
Wetness
F
reeze/thaw state of underlying soil
3Slide4
Current Snowpack Sensing Technologies
Combinations of 19 & 37 GHz brightness temperatures are used empirically to estimate snowpack SWE in simple
terrains
Combination of 10 & 17 GHz
SAR is being developed as an empirical technique to estimate snowpack SWE in all
terrains
The physical basis for both of these microwave techniques is differential frequency dependent scattering guided by theory but ‘tuned’ empirically
Radiometry
–
scatter darkening induced negative spectral gradients
Radar – frequency dependent backscatter strength
4Slide5
Limitations
of Empirical AlgorithmsBecause empirical algorithms are ‘tuned’ for an expected snowpack:
Static algorithms fail:
W
hen anomalous warm periods or diurnal melting causes metamorphic changes in snowpacks
W
here there is sub-pixel snowpack variability, i.e., area averaging has limited utility where processes are nonlinear
Dynamic algorithms:
R
equire a dynamic thermophysical snowpack model that follows the metamorphic evolution of the snowpack, and
Mechanisms to adjust algorithm for changes in snowpack grain size profiles from the thermophysical model
5Slide6
Wideband Autocorrelation Radiometry (Wideband AR
):An Alternative Technique for Snowpack Sensing?6
Downwelling Sky Radiance
AR Sensed Radiance
Snowpack
Upwelling Soil
Radiance
Direct
Ray
Ray
Delayed
By τ
0
Source = Upwelling Soil Radiance
+ Downwelling Sky Radiance
Sensed Signal = Direct Ray
+ Ray Delayed by
τ
oSlide7
Things to Note
Key is observing delayed autocorrelation peak at lag time τo
If thickness,
Δ
, varies over the footprint of the radiometer, the effect will be to broaden the autocorrelation peak at lag time τo
W
etness in the snowpack (< ~7 volume percent) will cause absorption and self emission
A
bsorption will reduce the height of the autocorrelation peak at
τ
oSelf emission will not be observed because it will not correlate with the direct ray
7
Downwelling Sky Radiance
AR Sensed Radiance
Snowpack
Upwelling Soil
Radiance
Direct
Ray
Ray
Delayed
By τ
0Slide8
Necessary Conditions for Sensing a Dry Snowpack with Wideband AR
Frequency, f, must be sufficiently low, and snowpack
thickness,
Δ
, sufficiently thin that neither absorption (or emission) nor scattering will significantly modify rays transiting the snowpack
Requirements generally met
for
f
<
10
GHz and Δ < 2 mInterfaces at top and bottom of snowpack must be nearly parallel and quasi-specular at sensor’s frequency
Requirement generally met for
f
<
10 GHz
Dielectric transitions at top and bottom of snowpack must be distinct
Requirement generally met for
f
<
10
GHz
C
orrelation time of
AR radiometer’s band-limited signal must be less than lag time of delayed autocorrelation peak, i.e., τc < τ
oConsequence of failing this condition is illustrated on next slide
8Slide9
Example Where
τ
c
>
τ
o
Experiment: Freshwater
Ice Over
Water
9
1.4 GHz Tb Profile
20 MHz bandwidth
23
0
beamwidth
100 m
agl
Winds calm
C
alibration flight during late fall, near Boulder, CO, England and Johnson, 1977Slide10
Note: Interference Patterns Are Not Reliably
Diagnostic of Snowpack ThicknessPhase of ‘Delayed’ ray is modulo 2π
for equivalent outcomes yielding
uncertainties in thickness corresponding to 2πn phase differences of ‘Delayed’ ray (where n
is an integer)
Variations in thickness over the footprint of the radiometer will average the interference effects
As snowpacks thicken, variations in thickness necessary to average the interference effects become smaller fractions of overall thickness
For sufficiently thick snowpacks, fringe-washing leads to an incoherent average
10Slide11
Consider a Hypothetical <10 GHz Wideband AR Sensor
11
Antenna
LNA
Analog
BDF
A/D
Digital Processor
Analog Band Definition Filter (BDF) has ~1.5 GHz passband
A/D
Downconversion
,
A/D converter has bandwidth >10 GHz and sampling rate of
>3
Gsamples
/s, i.e., >
Nyquist
rate for a 1.5 GHz passband
Low Noise Amplifier (LNA) system having sufficient gain for A/D conversion Slide12
Assuring that
τc < τo 12
Digital
LPF
Averaging
<
Φ
(
τ
)>
Autocorrelation
Φ
(
τ
)
Digital Processor
Digital Lowpass Filter (LPF)
Unbiased autocorrelation for sample lengths of twice expected
τ
o
Average autocorrelations to drive down noise floorSlide13
Constraints Upon Digital Lowpass Filter
Fourier transform of the autocorrelation of a zero-mean, white noise signal is the power spectrum of the signal, i.e:
τ
c
is inversely related to the bandwidth of the power spectrumFor a
Gaussian-shaped passband
τ
c
= (
Bandwidth)
-1In this case, minimum sensed snowpack thickness for Bandwidth = 1 GHz is ~70 cm
Better filter design and/or wider bandwidth will reduce the minimum sensed snowpack thickness
13Slide14
Hypothetical Wideband AR Sensor viewing a 1 m Snowpack
14
Bandwidth = 1 GHz
Delay = 10 ns
Attenuation = 37 dB
Integration = 1 s
8
th
order
Chebyshev
LPF with -45 dB stopbandSlide15
Conclusion: Potential
of Wideband AROffers a deterministic measure of microwave travel time in snowpack and, when combined with average snowpack density from a thermo-physical model and index of refraction from a dielectric mixing model,
Y
ields estimates of snowpack thickness and SWE
Width of delayed autocorrelation peak will yield an estimate of sub-pixel variance in snowpack thickness
Brightness of direct autocorrelation peak should yield freeze/thaw state of underlying soil
Attenuation
of
delayed autocorrelation peak might yield estimate
of snowpack
wetnessAdditional potential applications:
Sensing freshwater ice
thickness
Sensing planetary ice
thickness
15Slide16
Secondary Advantages of Wideband AR
Low power and low data rates characteristic of radiometersSimplified
thermal
design relative to traditional radiometers
Relaxed requirement for absolute
calibration
Because frequencies below 10
GHz
are within the band-widths of available
A/D
converters, the architecture of the analog front end can be greatly simplified at the cost of digital complexity16Slide17
Significant Challenges
Digital LPF will determine minimum sensed snowpack thicknessRequired Bandwidth is likely >
1
GHz, but how much greater?
Critical filter characteristics:
Minimum spectral width of transition to the stopband?
Needed depth
of
stopband probably
> 45
dBRadio Frequency Interference (RFI) with wideband system:
All RFI will
impact
Φ
(0
) but none are likely to cause false positives
Pulse RFI will likely require avoidance or removal
Communications RFI with long correlation times will raise noise floor
Within footprint, multi-source communications RFI might average out
17Slide18
Thank you!
Questions and/or Suggestions?18Slide19
Future Work
Experiment with design of Digital Lowpass Filter to achieve:A minimum necessary bandwidth
Minimum spectral width of autocorrelation skirt
Perhaps agile notch filtering of RFI
Develop full simulation of Wideband AR sensor to explore full parameter space of sensor design
Build proof-of-concept radiometer for boom on Microwave Geophysics Group’s field laboratory
Test proof-of-concept sensor on various snowpacks
19