FR4.TO5.3.pdf

Introduction Model Results obtained Conclusions

An empirical approach towardscharacterization of Dry Snow layers using

GNSS-R

Fran Fabra1, E. Cardellach1, O. Nogues-Correig1, S. Oliveras1,S. Ribo1, A. Rius1, G. Macelloni2, S. Pettinato2 and S.

D’Addio3

1ICE-CSIC/IEEC, Spain2IFAC/CNR, Italy

3ESA-ESTEC, Netherlands

IEEE International Geoscience and Remote Sensing SymposiumJuly 24-29, 2011, Vancouver, Canada


Outline

1 Introduction

2 Model

3 Results obtained

4 Conclusions


Outline

1 Introduction

2 Model

3 Results obtained

4 Conclusions


The frame of this work

GPS-SIDS project: Sea Ice - Dry Snow

GOAL: to investigate the use of reflected GPS signals to study sea iceand dry snow properties from Space

METHOD: to collect long term data sets from fixed platforms (2 fieldcampaigns) and then extrapolate the results

CHALLENGE: experimental campaigns under polar environmentalconditions

Many institutions involved: ICE-CSIC/IEEC, GFZ and IFAC/CNR(funded by ESA)


Experimental campaign

Location

Dome Concordia (Antarctica)



Location

Dome Concordia (Antarctica): American tower





Main aspects I

Instrument controlled in-situ by IFAC in coordination with IEEC.

Short campaign due to stability (DOMEX experiment [Macelloni et al., 2006])of the dry snow cover: 10th to 21st December 2009

Antennas on top of American tower with a height of 46m.

Not a single surface: reflected signal as a contribution from different layers.



Main aspects II

Maximum elevation achievable through a GPS reflection of 65◦ due tolimitations of the satellites constellation at these Latitudes.

Minimum elevation of 5◦ given by GOLD-RTR’s internal GPS receiver.

Monitorization area of ∼400m x 400m: pristine snow.

255˚

27

0˚

285˚

300˚

315˚

330˚

345˚0˚

15˚

30˚0˚

10˚

20˚

30˚

40˚

50˚

60˚

70˚

4

8

12

16

20

24

28

32

PR

N


The instrument employed: GOLD-RTR

GNSSR dedicated hardware receiver

GPS L1 (1575.45 MHz)

C/A code (Public and with simpleautocorrelation)

10 channels compute cross-correlations(waveforms) of 64 lags every millisecond

50 ns lagspacing ⇒∼ 15 meters

Scan the delay- and/or Doppler-space

3 radio front-ends

One Up-looking antenna for reference signal(internal GPS receiver)

A dual polarization (LHCP and RHCP)Down-looking antenna for reflected signals

Designed, manufactured, and tested at theICE-CSIC/IEEC


Outline

1 Introduction

2 Model

3 Results obtained

4 Conclusions


Motivation

Existing Model [Wiehl et al., 2003]: sub-surface contribution essentially given byvolumetric scattering.

But we consider volume scattering negligible compared to absorption loss.

Motivation: data shows clear interference fringes, better explained bymultiple-layer reflections.

0

5

10

15

20

Am

plitu

de (

a.u.

)

44.00 44.25 44.50 44.75 45.00

Elevation angle (deg)

0

5

10

15

20

Am

plitu

de (

a.u.

)

44.00 44.25 44.50 44.75 45.00


0

5

10

15

20

Am

plitu

de (

a.u.

)

44.00 44.25 44.50 44.75 45.00


0

5

10

15

20

Am

plitu

de (

a.u.

)

44.00 44.25 44.50 44.75 45.00


0

5

10

15

20

Am

plitu

de (

a.u.

)

44.00 44.25 44.50 44.75 45.00


0

5

10

15

20

Am

plitu

de (

a.u.

)

44.00 44.25 44.50 44.75 45.00


0

5

10

15

20

Am

plitu

de (

a.u.

)

44.00 44.25 44.50 44.75 45.00


0

5

10

15

20

Am

plitu

de (

a.u.

)

44.00 44.25 44.50 44.75 45.00


⇒ Need to develop our own model


Multi-layer Single-reflection model: MLSR

Multiple infinite parallel layers

1 single reflection per layer is considered

Only LHCP reflections reach the receiver

Expressions

ρi = ρ0 +k=i∑k=1

2nkHk

cos(θk )− (

k=i∑k=1

Dk )sin(θ0)

Dk = 2Hk tan(θk )

Uk = Rk,k+1Πi=ki=1Ti−1iTii−1e

−2αdi

Tco =1

2(T‖ + T⊥)

Rcross =1

2(R‖ − R⊥)

α =2π

λ|Ima{

√ε}|


Methodology

Input of the forward model (MLSR)

Depths and permittivity of the dry snow layers are needed

Retrieved from in situ measurements of snow density


Methodology

Complex waveform generated

Incident signal at surface with A=1

Direct signal set to lag 22 (RHCP toLHCP leakage with A=0.1)

Frequency of direct signal as areference

Several contributions


Methodology

Complex waveform generated

Incident signal at surface with A=1

Direct signal set to lag 22 (RHCP toLHCP leakage with A=0.1)

Frequency of direct signal as areference

Several contributions

Example: elevation from 44.884 to 44.955

deg (128 samples)


Methodology

Comparison of the modeled waveforms with real data


Methodology

Comparison of the modeled waveforms with real data


Methodology

Evolution of the different contributions wrt incidence angle

⇒ Interferometric frequency relates to depth of the layer


Methodology

How do we retrieve information?

FFT to each of the lag-time series ⇒ elevation domain (cycles/deg)

Several bands of main contributions below the surface level appear,corresponding to depths with stronger gradients of snow density/permittivity

⇒ Proper inversion might determine dominant layers of L-band reflections

0

100

200

300

Sno

w D

epth

(m

)

−18 −15 −12 −9 −6

Interf. Freq. (cycles/deg−el)

0

100

200

300

Sno

w D

epth

(m

)

−18 −15 −12 −9 −6


0

100

200

300

Sno

w D

epth

(m

)

−18 −15 −12 −9 −6


0

100

200

300

Sno

w D

epth

(m

)

−18 −15 −12 −9 −6


Lag-hologram from previous example


Outline

1 Introduction

2 Model

3 Results obtained

4 Conclusions


Lag-holograms from real data

First impressions

It show assymetrical behavior: consistentwith rotation due to interference fromdifferent contributions.

Apparently, wider frequency fringes thanexpected (high frequency bands at shortdelay lags).

Lag-hologram of 256 samples

PRN 19 at ∼ 40◦ of elevation from Dec-16.

Next processing step: averaging

To fade away those other features of the data which are not consistentlyrepeated (noise, atmospheric and instrumental induced features).

To identify the geometric and experimental parameters which essentially drivethe interferometric signal.


Averaging in elevation

Next slides show a series of

averaged lag-holograms

(using all PRN’s) in

elevation steps of 5◦ from

December 16th.

5◦ ⇒ 65◦

5◦ - 10◦

−20

−15

−10

−5

0

5

10

15

20

Fre

quen

cy (

cycl

e/de

g−el

ev.)

10 20 30 40 50 60

Delay−lag (15 m.−lag)

10◦ - 15◦

−20

−15

−10

−5

0

5

10

15

20

Fre

quen

cy (

cycl

e/de

g−el

ev.)

10 20 30 40 50 60




15◦ - 20◦

−20

−15

−10

−5

0

5

10

15

20

Fre

quen

cy (

cycl

e/de

g−el

ev.)

10 20 30 40 50 60


20◦ - 25◦

−20

−15

−10

−5

0

5

10

15

20

Fre

quen

cy (

cycl

e/de

g−el

ev.)

10 20 30 40 50 60


25◦ - 30◦

−20

−15

−10

−5

0

5

10

15

20

Fre

quen

cy (

cycl

e/de

g−el

ev.)

10 20 30 40 50 60




30◦ - 35◦

−20

−15

−10

−5

0

5

10

15

20

Fre

quen

cy (

cycl

e/de

g−el

ev.)

10 20 30 40 50 60


35◦ - 40◦

−20

−15

−10

−5

0

5

10

15

20

Fre

quen

cy (

cycl

e/de

g−el

ev.)

10 20 30 40 50 60


40◦ - 45◦

−20

−15

−10

−5

0

5

10

15

20

Fre

quen

cy (

cycl

e/de

g−el

ev.)

10 20 30 40 50 60




45◦ - 50◦

−20

−15

−10

−5

0

5

10

15

20

Fre

quen

cy (

cycl

e/de

g−el

ev.)

10 20 30 40 50 60


50◦ - 55◦

−20

−15

−10

−5

0

5

10

15

20

Fre

quen

cy (

cycl

e/de

g−el

ev.)

10 20 30 40 50 60


55◦ - 60◦

−20

−15

−10

−5

0

5

10

15

20

Fre

quen

cy (

cycl

e/de

g−el

ev.)

10 20 30 40 50 60




55◦ - 60◦

−20

−15

−10

−5

0

5

10

15

20

Fre

quen

cy (

cycl

e/de

g−el

ev.)

10 20 30 40 50 60


60◦ - 65◦

−20

−15

−10

−5

0

5

10

15

20

Fre

quen

cy (

cycl

e/de

g−el

ev.)

10 20 30 40 50 60


Apparently, the elevation

range that shows consistent

features from deep snow

layers goes from 20◦ to

45◦.

The elevation-rate also

plays an important role in

this analysis.


Averaging in elevation-rate

Next slides show a series of

averaged lag-holograms

(using all PRN’s) in

elevation-rate steps of

0.001◦/s from December

16th.

0.001◦/s ⇒ 0.008◦/s

0.0◦/s - 0.001◦/s

−20

−15

−10

−5

0

5

10

15

20

Fre

quen

cy (

cycl

e/de

g−el

ev.)

10 20 30 40 50 60


0.001◦/s - 0.002◦/s

−20

−15

−10

−5

0

5

10

15

20

Fre

quen

cy (

cycl

e/de

g−el

ev.)

10 20 30 40 50 60




0.002◦/s - 0.003◦/s

−20

−15

−10

−5

0

5

10

15

20

Fre

quen

cy (

cycl

e/de

g−el

ev.)

10 20 30 40 50 60


0.003◦/s - 0.004◦/s

−20

−15

−10

−5

0

5

10

15

20

Fre

quen

cy (

cycl

e/de

g−el

ev.)

10 20 30 40 50 60


0.004◦/s - 0.005◦/s

−20

−15

−10

−5

0

5

10

15

20

Fre

quen

cy (

cycl

e/de

g−el

ev.)

10 20 30 40 50 60




0.005◦/s - 0.006◦/s

−20

−15

−10

−5

0

5

10

15

20

Fre

quen

cy (

cycl

e/de

g−el

ev.)

10 20 30 40 50 60


0.006◦/s - 0.007◦/s

−20

−15

−10

−5

0

5

10

15

20

Fre

quen

cy (

cycl

e/de

g−el

ev.)

10 20 30 40 50 60


0.007◦/s - 0.008◦/s

−20

−15

−10

−5

0

5

10

15

20

Fre

quen

cy (

cycl

e/de

g−el

ev.)

10 20 30 40 50 60




0.006◦/s - 0.007◦/s

−20

−15

−10

−5

0

5

10

15

20

Fre

quen

cy (

cycl

e/de

g−el

ev.)

10 20 30 40 50 60


0.007◦/s - 0.008◦/s

−20

−15

−10

−5

0

5

10

15

20

Fre

quen

cy (

cycl

e/de

g−el

ev.)

10 20 30 40 50 60


Consistent features from

deep snow layers are shown

from up to ∼0.005◦/s.

Expected result:

interferometric information

depends on the evolution of

the different paths followed

by each layer-contribution

during the time-series of

data (fixed).


First conclusions from averaging

A minimum elevation and elevation-rate are needed for obtaining persistentfeatures in the lag-holograms.

The apparently bad results at higher elevations may be due to the impact of lowelevation-rate samples in the elevation averaging.

Test: Data set divided in cells of elevation and elevation-rate ⇒ Averaging overmost populated cell.

0.000

0.002

0.004

0.006

0.008

Ele

vatio

n ra

te (

deg−

elev

/s)

10 20 30 40 50 60 70


5

10

15

20

25

30

PR

N

Cell division

0.000

0.002

0.004

0.006

0.008

Ele

vatio

n ra

te (

deg−

elev

/s)

10 20 30 40 50 60 70


0 5 10 15 20 25 30 35 40 45 50 55 60 65 70

Counts


Averaging in cell: Repeatability

Dec 16th

−20

−15

−10

−5

0

5

10

15

20

Freq

uenc

y (cy

cle/de

g−ele

v.)

10 20 30 40 50 60


Dec 18th

−20

−15

−10

−5

0

5

10

15

20

Freq

uenc

y (cy

cle/de

g−ele

v.)

10 20 30 40 50 60


Dec 17th

−20

−15

−10

−5

0

5

10

15

20

Freq

uenc

y (cy

cle/de

g−ele

v.)

10 20 30 40 50 60


Dec 19th

−20

−15

−10

−5

0

5

10

15

20

Freq

uenc

y (cy

cle/de

g−ele

v.)

10 20 30 40 50 60



Depth of the contributing layers

Most of the reflected signals comes fromfirst 40 meters: especially 10 first meters–5 to 7 cycle/deg–, and from 30 to 40meters depth –around 9 cycle/deg–.

Remarkable contributions between 75 and100 meters –10 to 12 cycle/deg–, andaround 110 and 140 meters –14 and 16cycle/deg–.

Layers corresponding to depths withstronger gradients of snowdensity/permittivity from the in-situmeasurements.

0

100

200

300

Sno

w D

epth

(m

)

−18 −15 −12 −9 −6


0

100

200

300

Sno

w D

epth

(m

)

−18 −15 −12 −9 −6


0

100

200

300

Sno

w D

epth

(m

)

−18 −15 −12 −9 −6


0

100

200

300

Sno

w D

epth

(m

)

−18 −15 −12 −9 −6


MODEL

DATA (averaged cell)


Outline

1 Introduction

2 Model

3 Results obtained

4 Conclusions


Conclusions

Persistent signatures have been found in the collected data-set of GPSreflections over dry snow.

The frequency-domain analysis shows interferometric behavior(asymmetrical spectrum) with contributions below the snow surface level.

No success when attempting a total inversion of the dry snow layersprofile from the data.

In spite of the poor results, the link between frequency bands in thelag-holograms and the depth of a reflecting interface is still a source ofinformation.

Alternative estimates

In [Hawley et al., 2006], the delays of radar measurements are not used to

estimate the snow density, but combined with a given density profile to

estimate the snow accumulation rates. A similar application can be

envisaged for this data set.


Thank you for your attention


References

Macelloni et al., 2006: G. Macelloni et al., ”DOMEX 2004: An Experimental Campaign at Dome-CAntarctica for the Calibration of Spaceborne Low-Frequency Microwave Radiometers”, IEEE Trans. Geosci.and Remote Sens., vol. 44, 2006.

Wiehl et al., 2003: M. Wiehl, R. Legresy, and R. Dietrich, ”Potential of reflected GNSS signals for icesheet remote sensing”, Progress in Electromagnetics Research, vol. 40, pp. 177–205, 2003.

Hawley et al., 2010: R.L. Hawley, E. M. Morris, R. Cullen, U. Nixdorf, A. P. Shepherd and D. J.Wingham, ”ASIRAS airborne radar resolves internal annual layers in the dry-snow zone of Greenland”,Geophysical Research Letters, vol. 33, 2006.

Technology

FR4.TO5.3.pdf