Sub-surface glacial structure over Nordaustlandet using multi-frequency Pol-InSAR

Transcription

1 ub-surface glacial structure over Nordaustlandet using multi-frequency Pol-InAR Jayanti harma, Irena Hajnsek, Kostas Papathanassiou DLR (German Aerospace Center),.6.8

2 Introduction: Pol-InAR over Land Ice Motivation: A greater understanding of the properties of glaciers/ice sheets including their topography, velocities, melting/accumulation rates, densities and internal structure (related to rate of extinction) AR (synthetic aperture radar) sensors particularly well-suited for cryospheric measurements (polar areas dark for much of the year, high resolution & coverage) polarimetry yields information regarding scattering mechanisms (volume+surface scattering), interferometry regarding their distribution with height Goal: For the first-time, use Pol-InAR observables to estimate land ice extinctions lide

3 Airborne Polarimetric AR Interferometry (Pol-InAR) HH HV VH VV baseline HH HV VH VV AR: synthetic aperture radar enables high resolution in the along-track line-of-flight direction Polarimetry: enables the separation of different scattering mechanisms within a resolution cell Interferometry: sensitive to the height at which scattering occurs Pol-InAR: combined, sensitive to the vertical distribution of scattering mechanisms lide 3

4 Data description ICEAR data campaign conducted in Mar/Apr 7 over Austfonna ice cap (ummit) and Etonbreen drainage basin, valbard, Norway (~ 8 N, 4 E) Joint project between DLR (German Aerospace Center) EA (European pace Agency) AWI (Alfred-Wegener Institute) Etonbreen ummit Corner reflector at ummit DLR s E-AR (left), AWI s airborne platform lide 4

5 E-AR parameters X L P λ [m] f [GHz] PRF [Hz/chan] 4 5 Chirp BW [MHz] lant-rng res. Δrng [m].5 lant-rng spac. δrng [m].5 Az res. LC Δaz [m] Az spac.lc δaz [m] Focus on L-/P-band, Baselines of 5- m (L-band), -4 m (P-band) lide 5

6 Experimental Data, ummit, Pauli images L-band P-band surface (HH+VV) double bounce (HH-VV) volume (HV) lide 6

7 Modelling Ice Extinctions: Volume/urface separation with Polarimetric Decomposition lide 7

8 Isolation of the Ice Volume Goal: In order to estimate extinction of the ice volume, necessary to separate ground and volume contributions Method: Estimate ground-to-volume scattering ratios (m) through decomposition of polarimetric covariance [C] matrix (second-order statistics) = VH VV HH HV [ C] = HH * HH * HH VV HV HH HV * HV * HV VV HH HV VV * VV * VV Model elements of the [C] matrix using the Freeman 3-component decomposition Freeman, 998. A three-component scattering model for polarimetric AR, TGAR, vol.36 no 3. lide 8

9 lide 9 Polarimetric Decomposition Freeman-3 Component Decomposition Originally proposed for modelling forested regions with PolAR Decomposes [C] into surface, dihedral and volume components Assume: random volume of dipoles Adjust volume component for snow-firn transmissivities (T s, T P ) + + = * * P P s P s P s s v d s T T T T T T T T f f f α α α β β β ] [ ] [ ] [ ] [ v d s C C C C + + = Total backscatter = + + urface Dihedral Volume

10 L-band, Freeman-3 decomposition: relative powers P s P d P v azimuth range lide

11 P-band, Freeman-3 decomposition: relative powers P s P d P v azimuth range lide

12 Volume/urface eparation: Experimental data vs. theory Theory: Expected Bragg surface σ as a function of λ with incidence angle θ (s=cm, l= cm) L-band P-band Assuming a homogeneous volume, expect same trend for ground-to-volume scattering ratios m m = P P Ulaby, 98. Microwave Remote ensing, Active and Passive, Vol. II V σ Bragg Homogeneous subset azimuth range lide

13 Ground-to-volume ratios, ubset L-band P-band m HH P HH P VV = mvv = mhv P P V HH V VV = As expected, general trend of decreasing m with θ at both freqs m values higher for P-band than for L-band: perhaps related to significantly less volume backscatter (relative size scatterers with respect to wavelength) lide 3

14 Modelling Ice Extinctions: InAR Coherence Model lide 4

15 γ Coherence modelling vol Fundamental InAR observable coherence: Assuming scattering medium is : = + γ = infinite, homogeneously lossy (constant extinction κ e [/m]) consists of uniformly distributed and uncorrelated scattering centres j cosθrk κ e zvol k zvol 4π ε Δθr = λ sinθ Combine with a ground contribution from snow-firn interface (m = ground-to-volume scattering ratio) and take absolute value: γ r A B A θ B R R z= now Firn = const ε ε = dielectric permittivity γ vol + m γ = + m z Δθ θ r x Δθ r x y harma,7. Multi-frequency Pol-InAR signatures of a subpolar glacier. Pol-InAR, -6 Jan., Frascati. lide 5

16 Expt Data, All B, mask (. < k z <.) κ e VV L-band κ e VV P-band κ e db/m κ e db/m azimuth range Near-range lide 6

17 Expt Data: Extinction Inv. ubset, All B L-band, Volume only (m=) P-band, Volume only (m=) L-band, Freeman-3 P-band Freeman-3 lide 7

18 ummary Extinction model Adapted existing PolAR decomposition/pol-inar coherence model created for vegetation scenarios to a glacier geometry eparated ground and volume contributions using Freeman-3 component decomposition Estimated extinctions using InAR, combining baselines for a more robust solution -D images of extinctions at L- and P-band, useful in identifying areal extents of internal ice structure, not visible with optical systems nor -D GPR profiles Future Work Extend method to polarimetric decompositions modelling an oriented volume instead of a random volume Relate extinctions to geophysical parameters (grain size/density, presence of percolation features such as ice lenses/pipes,etc.) lide 8

19 Questions? lide 9

20 Volumetric decorrelation Assuming scattering medium is: Where: homogeneously lossy (i.e. constant extinction) consists of uniformly distributed and uncorrelated scattering centres σ v = averaged normalized RC per unit vol. [m / m 3 ] κ e = extinction coeff. [/m] γ vol v = σ ( z) e v σ ( z) = σ e jk σ ( z) dz v v zvol zκe cosθ r z dz = cos (θ r ) / d pen z/cos(θ r ) = penetration length in the vol. [m] Evaluate γ vol for - < z < γ vol = + j cosθrk κ e zvol lide

21 Extinction: κ e = cosθ r k zvol ( + m) γ ( + m) γ m Non-linear solution for extinction κ e κ = κ + κ e a s Absorption coeff cattering coeff lide

22 Extinction Inversion: ensitivity analysis P-band, B= m, κ e true =. db/m, δm=. Plot bias in inverted extinctions due to an error in m (ground-to-vol scattering ratio) constant m assumed; decreasing m with θ will exaggerate trend even further non-linear soln for κ e creates large jumps in soln for < m < P-band, m=, κ e true =. db/m, δm=. κ e = cosθ r k zvol ( + m) γ ( + m) γ m lide