The Extraction of InSAR Information from Imagery of a Wind-Blown Tree Canopy with a Ground-Based SAR Keith Morrison & Muhammad Yasin Department of Aerospace, Power and Sensors, University of Cranfield, Shrivenham, UK & DLR, Institut für Hochfrequenztechnik und Radarsysteme Weßling, Germany
The GB-SAR System Portable SAR / InSAR Imaging System All-weather L through X-band (1-12GHz) Fully polarimetric VV,HH,VH,HV
Rationale Particular open questions relate to the conditions under which PolInSAR produces accurate measurements of biomass, with respect to: canopy structure (species, density, height distribution) technical sensor specifications imaging conditions (spatial and temporal)
Presentation Can the GB-SAR system be used to obtain meaningful PolInSAR measurements of forest canopies? Considerations GB-SAR imaging timescale on order of tens of minutes Can expect wind-induced target motion Can the results be related to air- and space-borne ?
SAR Imaging of Tree
Sweet Chestnut (castanea sativa mills)
Tree spatially isolated in grassy parkland
Tree Dimensions Trunk Height = 25m Trunk diameter at DBH = 1.7m Trunk Circumference at DBH = 5.6m Maximum tree width (2m from ground) = 15m Tree width at ¾ of tree height = 12 m Maximum tree depth = 18m Tree depth at ¾ of tree height = 11m
9.6m 14m 25m 5m
Winter View, from back
Radar Parameters SF-CW Radar Type 13th July 2005Date of observation 4.000GHzStart frequency (GHz) 6.000GHzEnd frequency 1601Number of frequencies per sweep 1.25MHzFrequency step interval 3000HzVNA IF bandwidth +8dBmEffective transmit power at antenna VVPolarisation 20mmAperture elemental sampling, dx 3680mmAperture size, D 185Number of aperture samples 1 or 8Data averaging factor 9.6mAntenna height above ground 0.9s Tsweep, frequency sweep time 1.1sTmove, antenna movement time
ScanStart TimeEnd TimeDurationAveragingLapsed 116:02:1816:08:586.7 min1- 216:32:3716:39:186.7 min10 min 316:41:1816:47:586.7 min18.7 min 416:48:4616:55:276.7 min116.2 min 516:57:1017:03:516.7 min124.6 min 617:04:4417:11:246.7 min132.1 min 717:12:5017:19:306.7 min140.2 min 817:20:4517:27:256.7 min148.1 min 917:28:3417:35:156.7 min156.0 min 1017:38:0018:08: min877.1 min 1118:13:0618:43: min min
Figure 4.1: The log-amplitude images of Scans (top to bo Scans 1-9. Av. Factor 1
Figure 4.2: The log-amplitude images of Scans 10 (top) and 11. The plots are 3-29m in rang Scans 10 & 11. Av Factor 8
Antenna & Space-loss
Corrected Images
Bulk Averaging - Tree
Canopy Attenuation
InSAR Decorrelation γ = γNoise. γSpatial. γSystem. γTemporal 1/(1+SNR -1 )
Figure 8.7: Coherence maps of (top left): Scans 2 & 3; (top right) Scans 8 & 9; (bottom left) Scans Coherence Analysis
Coherence vs Amplitude
Coherent Summation
Distribution of Coherence
Regression Analysis
y = m.x + c
Regression Fit – Gradient (m)
Regression Fit - Constant (c)
Standard Deviation From Fit
Model Simulations
Effects of Wind-Motion
Motion Simulation
Sim_1a vs Sim_1b Sim_1a vs Sim_2a
InSAR Phase
InSAR Phase vs Coherence The curves show the frequency of occurrence with phase for varying coherence ranges. The outermost curve is over the entire coherence range The next innermost curve shows the distribution 0.1-1, then 0.2-1, and so on. The innermost curve shows the phase distribution
Non-Zero Baseline
Conclusions Meaningful SAR Imaging of trees is feasible Wind motion produces spreading of IPR into broadband unstructured azimuthal arcs Good coherences obtained by observation in low wind conditions Recovery of ‘static’ backscatter pattern by temporal averaging Averaging also improves the coherence However, latter might bias InSAR phase / height retrieval to stronger coherent features in canopy Investigation into whether the GB-SAR system can be used for InSAR & PolInSAR