Biomass Estimation from Forest Vertical Structure: Potentials and - PowerPoint PPT Presentation

Biomass Estimation from Forest Vertical Structure: Potentials and Challenges for Multi-Baseline Pol-InSAR Techniques M. Pardini, F. Kugler, S.-K. Lee, S. Sauer, A. Toraño Caicoya & K. Papathanassiou Microwaves and Radar Institute (DLR-HR) Microwaves and Radar Institute / Pol - InSAR Research Group German Aerospace Center (DLR)

Deforestation, Degradation, Fires* (REDD) Forest Biomasse Change* Biodiversity Earthquakes Volcanic Activities Land Slides Sea Ice Extent* Permafrost* Glacier & Ice Cap Dynamics* Soil Moisture* Flooding *) Essential Climate Ocean Currents* Variables Microwaves and Radar Institute Microwaves and Radar Institute > 30.05.2006

Product Product Science Product Coverage Resolution Accuracy 50 m (global) Forest Height ~ 10 % 20 m (local) 100 m (global) ~ 20 % All forest Above Ground Biomass ≤ 50 m (regional) (or 20 t/ha) Biosphere Areas (Height ≥ 8 m) 50 m (global) Vertical Forest Structure 3 layers 20 m (local) Underlying Topography 50 m < 4 m 100 m (global) 1 mm/year Plate Tectonics all risk areas < 20 m (fault) (after 5 y) all land Volcanoes 20 – 50 m 5 mm/week Geo-/Lithosphere volcanoes Landslides risk areas 5 – 20 m 5 mm/week Subsidence urban areas 5 – 20 m 1 mm/year Glacier Flow main glaciers 100 – 500 m 5 – 50 m/year Soil Moisture selected areas 50 m 5 – 10 % Water Level Change regional 50 m 10 cm Cryo- & Hydrosphere Snow Water Equivalent local (exp.) 100 – 500 m 10 – 20 % Ice Structure Changes local (exp.) 100 m > 1 layer Ocean Currents prio. areas ~ 100 m < 1 m/s Digital Terrain & Surface ~ 20 m (bare) 2 m (bare) All global Microwaves and Radar Institute Model ~ 50 m (forest) 4 m (veg.) Microwaves and Radar Institute > 30.05.2006

Overview Biomass estimation: from “height to biomass allometry ” to “structure to biomass allometry ” Vertical structure estimation from multi-baseline Pol-InSAR data Spectral estimation Polarization coherence tomography Real data experiments with airborne acquisitions (test site over the Traunstein forest) Can “radar” structure express biomass? A preliminary validation Conclusions Microwaves and Radar Institute Microwaves and Radar Institute > 30.05.2006

Height to Biomass Allometry Nationalpark Bayrischer Wald Ebersberger Forst Bürgerwald Traunstein Natural development since 1972 “Close to Nature” Intensely managed Montane spruce forest > 1100m asl. Temperate managed forest Submontane mixed forest Floodplain spruce forest < 600m asl Single species (Spruce) N. Spruce, E. Beech, White Fir Height Range (H100): 5 - 45m Height Range (H100): 5 - 40m Height Range (H100): 10 - 40m Biomass Range: 40 ~ 450 t/ha Biomass Range: 40 ~ 350 t/ha Biomass Range: 40 ~ 450 t/ha Steep Slopes Flat Terrain Moderate Slopes Microwaves and Radar Institute Microwaves and Radar Institute > 30.05.2006

Height to Biomass Allometry 1 . 50 B = la 1 66 H * . Nationalpark Bayrischer Wald Ebersberger Forst Bürgerwald Traunstein Microwaves and Radar Institute Microwaves and Radar Institute > 30.05.2006

Structure to Biomass Allometry H 3  *  * B 3 . 11 a j P = (z ) j i = = i 0 j 1 a 1 a 2 a 3 Nationalpark Bayrischer Wald Ebersberger Forst Bürgerwald Traunstein Microwaves and Radar Institute Microwaves and Radar Institute > 30.05.2006

*   S S Interferometric S S ~  1 2 γ ( S S ) 1 2 1 2 * * Coherence     S S S S 1 1 2 2 SAR Interferometry for Volume Structure h v  ik z f ( z ) e dz z Volume ~  ik z γ o ( f ( z )) e z o Vol h Coherence v  f ( z ) dz f ( z ) o … vertical reflectivity function f ( z ) ~  ~ ~ γ γ γ γ κ Δ θ Temporal SNR Volume z  κ Vertical Wavenumber: θ sin( ) 0 ~ γ … temporal decorrelation Temporal γ … additive noise decorrelation SNR Baseline diversity allows to sample the ~ same vertical structure spectrum at γ … geometric decorrelation Volume different spatial frequencies VU 8 > Autor Name Microwaves and Radar Institute Microwaves and Radar Institute > 30.05.2006

S 1 S K S 2 S 3 Vertical structure estimation as a spectral estimation Equivalent cross-track array problem f (z) Multi-baseline S S data vector Elevation steering Data-dependent Normal-to-slant range (tuning) Interference rejection g (z 1 ) g ( z 2 ) g ( z M ) height Sweeped variable structure bandpass (variable beamshape) f (z 1 ) f (z 2 ) f (z M ) Classic beamforming (Fourier-based) Adaptive beamforming (Capon-based): high resolution, low sidelobes, but radiometrically non linear Covariance matrix brings all the information Need for an adequate number of baselines Baseline distribution and total baseline length play a primary role Microwaves and Radar Institute Microwaves and Radar Institute > 30.05.2006

Polarization S 1 S K S 2 S 3 Coherence Tomography h v  ik z f ( z ) e dz z Volume ~  ik z γ o ( f ( z )) e z o Vol h Coherence v  f ( z ) dz f ( z ) o f ( z ) … vertical reflectivity function κ Δ θ z  κ Vertical Wavenumber: f ( z ) θ sin( ) 0 k h k h h h 1 h z v z v v v i i z '      ik z ik z v 2 2 f ( z ) e dz e ( 1 f ( z ' )) e dz ' f ( z ) e dz z z 2 ~   ik z 0 1 γ o ( f ( z )) e z o Vol h h 1 v v h      v f ( z ) dz f ( z ) dz ( 1 f ( z ' )) dz ' 2  o 0 1 1   2 n 1    f ( z ' ) a P ( z ' ) a f ( z ' ) P ( z ' ) dz ' Fourier Legendre Series: where n n n n 2 Microwaves and Radar Institute n  1 Microwaves and Radar Institute > 30.05.2006

The Traunstein dataset 0 m 50 m Height Point-Spread Functions (PSF) 570 m 775 m 0 m 50 m Forest type Temperate Topography Moderate slopes Height 25 ~ 35m Species N. Spruce, E. Beech, White Fir Biomass 40 ~ 450 t/ha Frequency L-band (1.3 GHz) Baselines 0,5,10,15, 20 nom. Full-pol Acq. date June 12, 2008 (1 hour) Vertical resolution 2  / max( k z ) HV Amplitude Image LIDAR DTM LIDAR forest height Slant range Slant range Slant range Slant range Microwaves and Radar Institute Microwaves and Radar Institute > 30.05.2006

Tomographic slices & profiles (1/2) Range bin 364 Range bin 500 Range bin 584 HH HV Microwaves and Radar Institute Microwaves and Radar Institute > 30.05.2006

Tomographic slices & profiles (2/2) PCT profiles can be calculated also with a very low number of baselines Master From full-baseline profiles to dual baseline profiles B 2 B 1 Microwaves and Radar Institute Microwaves and Radar Institute > 30.05.2006

Traunstein test site: a preliminary validation Measured PCT Measured PCT 50 40 30 20 10 0 m Microwaves and Radar Institute Microwaves and Radar Institute > 30.05.2006

Some conclusions Multi-baseline Pol-InSAR techniques for biomass estimation: Here we are … Accurate (<10%) estimation of forest top height at high spatial resolutions (20-50m grid); vertical “radar” forest structure achievable by means of a “realistic” number of acquisitions. Potentials … Structure-based (AG) biomass estimators promise accuracy and stability across very different forest conditions; PCT shows potentials for biomass estimation even with dual baseline data. Challenges … Mapping of “radar” structure to biomass structure needs to be further investigated, especially with reference to acquisition-dependent parameters (frequency, polarization, temporal decorrelation … ). Microwaves and Radar Institute Microwaves and Radar Institute > 30.05.2006

Thank you! … Questions? Biomass Estimation from Forest Vertical Structure: Potentials and Challenges for Multi-Baseline Pol-InSAR Techniques M. Pardini, F. Kugler, S.-K. Lee, S. Sauer, A. Toraño Caicoya & K. Papathanassiou Microwaves and Radar Institute (DLR-HR) Microwaves and Radar Institute German Aerospace Center (DLR) Microwaves and Radar Institute > 30.05.2006