Spatial Analysis for Radar Remote Sensing of Tropical Forests

✍ Scribed by Gianfranco D. De Grandi, Elsa Carla De Grandi

• Publisher: CRC Press
• Year: 2021
• Tongue: English
• Leaves: 380
• Series: SAR Remote Sensing
• Edition: 1
• Category: Library

✦ Synopsis

Spatial Analysis for Radar Remote Sensing of Tropical Forests is based on the authors’ extensive involvement in Synthetic Aperture Radar (SAR) mapping projects, targeting the health of an earth ecosystem with great relevance for climate change studies: the tropical forests. The subject is developed from a vantage point provided by analysis in a combined space, scale (frequency), time, wavelength, polarization domain. The combination of space and scale offers the capability to zoom in and out like a virtual microscope to the resolution in tune with the underlying ecological phenomenon. It also enables statistical measures (correlations) related to the forest spatial distribution in case of backscatter, or to the canopy height variations in case of interferometric observations. The time dimension brings into play measures of the ecosystem dynamics, such as the flooding extent in the swamp forests, deforestation or degradation events. Wavelength and polarization agility extend the abovementioned capabilities by radar observations that are in tune with particular characteristics of the forest and terrain layers. The book’s spotlight is on radar spatial random fields, these being populated by either backscatter observations or elevation data from interferometric SAR. The basic tenet here is that the spatial statistic of the fields measured by the wavelet variance (in stationary or non-stationary situations) carries fingerprints of the forest structure.

✦ Table of Contents

Half Title
Series Page
Title Page
Copyright Page
Table of Contents
Preface
Acknowledgements
The Authors
List of Abbreviations
List of Figures and Tables
Part I: Sarcheology: The Era of the Big Radar Mosaics
Chapter 1: The Dawn of the SAR Mosaics Era: The ESA–JRC Central Africa Mosaic Project
1.1 Radar Mosaics: What and Why
1.2 The CAMP Data Processing Machine
1.3 Radiometry
1.3.1 Radiometric Changes in Time
1.3.2 Within Tile Radiometric Changes in range
1.3.3 Quantization Noise
Note
References
Chapter 2: The L-Band Breed: The GRFM Africa Radar Mosaic
2.1 The GRFM project
2.2 The GRFM Africa Processing Chain
2.2.1 Input Datasets
2.2.2 Data Flow
2.3 Geolocation
2.3.1 The Block Adjustment Method
2.3.2 Geolocation Validation
2.4 Wavelet Multiresolution Decomposition
2.4.1 Multiresolution Products
2.5 The GRFM Africa Mosaic Second Edition
References
Chapter 3: The GRFM–CAMP Thematic Products
3.1 From Backscatter to a Thematic Map
3.2 Vegetation Classes
3.3 Map Compilation Methods
3.4 Complementarity of Radar Sensors
3.5 Validation
3.6 Tour of Relevant Features
References
Chapter 4: Evolution of the Species: The ALOS PALSAR Africa Mosaic
4.1 Introduction
4.2 The Mosaic Processing Chain
4.3 Correction of Range Dependent Radiometric Bias in Path Images
4.4 Correction for Additive Thermal Noise in HV Strip Images
4.5 Radiometric Inter-strip Mosaic Balancing
4.6 Geocoding
4.7 Radiometric Normalization for Topographic Effects
4.7.1 Correction of Effective Scattering Area
4.7.2 Correction for the Dependence of the Backscattering Coefficient on Incidence Angle
4.7.3 Assessment of the Radiometric Correction for Topography
4.8 Overview of the Thematic Information Content
4.8.1 Comparison with the GRFM Africa Dataset
4.8.2 Grass and Woody Savannas
4.8.3 Flooded Forest
4.8.4 Plantations
4.8.5 Secondary Forest
References
Part II: Measures of SAR Random Fields in the Scale–Space–Time Domain
Chapter 5: The Stuff Backscatter Random Fields Are Made Of
5.1 Introduction
5.2 Transport Theory
5.2.1 An Illustrative Case: Propagation Through A Plane Parallel Medium
5.3 The UTA Wave Scattering Model for Layered Vegetation
5.4 Backscatter Simulation for a Dense Tropical Primary Rain Forest
References
Chapter 6: Statistical Measures of SAR Random Spatial Fields: Fingerprints of the Forest Structure
6.1 Introduction
6.2 Random Fields from Backscatter Observations
6.3 Random Fields from InSAR Coherence Observations
6.4 Wavelet Based Textural Measures of Random Fields
6.5 Connection between Wavelet Space–Scale Analysis and Fourier Spectral Analysis
6.5.1 White Noise
6.5.2 1/f Process
6.5.3 Correlated Surface (Gamma Distributed RCS) with Exponential ACF (Lorentzian Spectrum)
6.5.4 Correlated Surface with Exponential Cosine ACF
6.5.5 Effects from Coherent Imaging and Illumination Beam Size
6.5.6 Cross-correlation between Two Stationary Processes with a Gaussian CCF
6.6 Accuracy of Wavelet Variance Estimators
6.6.1 Prelude: Probability Density Function of the Wavelet Coefficients of a Speckle Pattern
6.6.2 Expected Value and Variance of the Wavelet Variance Estimator
6.6.2.1 Uncorrelated Speckle Pattern
6.6.2.2 Correlated Speckle
6.7 Tools for Textural Analysis of SAR Random Fields
6.7.1 A Multi-Voice Discrete Wavelet Transform
6.7.2 Wavelet Signatures
6.7.3 Wavelet Spectra
6.8 WASS Analysis of SAR Backscatter Fields
6.8.1 Lowland Rainforest and Swamp Forest Signatures in ERS-1 Data
6.8.2 TanDEM-X Signatures in the same Thematic Context
6.8.3 Intact and Degraded Forest Detection by Functional Analysis of WASS Signatures
6.9 WASS Analysis of InSAR and LiDAR Digital Surface Models
6.10 2D Wavelet Variance Spectra of Backscatter Fields: Toward a Textural Classifier
6.10.1 A Test Case: Texture-Based Forest Mapping in the Congo Floodplain by ERS-1 Data
6.10.2 Floodplain Mapping Revisited by Sentinel-1 data
6.10.3 An (Experimental) Wavelet Spectrum Functional Classifier
6.11 Extension to Polarimetry
6.11.1 The WASP of Correlated Backscatter Textures: A Numerical Model
6.11.2 WASP Analysis of a PALSAR Full-Pol Data Set
Note
References
Chapter 7: Hitting Corners: The Lipschitz Regularity, a Measure of Discontinuities in Radar Images Connected with Forest Spatial Distribution
7.1 Introduction
7.2 The Lipschitz Condition
7.3 Singular Functions and Lip Parameters Estimated by Wavelet Maxima Trajectories in the Scale Domain
7.3.1 Step Function
7.3.2 Cusp
7.3.3 Impulse
7.3.4 Smoothed Singularity
7.3.5 Non-Isolated Singularities
7.3.6 Effect of Speckle
7.4 A Monte Carlo Simulator of Polarimetric SAR Backscatter Discontinuities
7.5 Experiments Using Simulated Signals
7.5.1 Toy Signals with Simple Discontinuities
7.5.2 Margin between a Clear-Cut and a Dense Forest
7.5.3 Edge on Tilted Terrain
7.6 Lipschitz Regularity in Real SAR Data
7.6.1 TanDEM-X Backscatter Data
7.6.2 TanDEM-X Coherence Data
7.7 Image-Wide Representations of Lipschitz Parameters
References
Chapter 8: The Beauty Farm: A Wavelet Method for Edge Preserving Piece-wise Smooth Approximations of Radar Images
8.1 The Image Model and a Conceptual View of the Method
8.2 The Computational Engine
8.3 Problems Related to Multiplicative Speckle Noise
8.4 Issues Related to Textural Edges
8.5 Maxima Linking
8.6 From Theory to Practice: A Tropical Forest Cover Mapping Exercise Using Smooth Approximations of GRFM SAR Data
8.6.1 Processing Methods
8.6.1.1 Region Growing
8.6.1.2 NMP Classifier
8.6.2 Test Sites and Thematic Class Definition
8.6.3 Selected Results
References
Chapter 9: The Cleaning Service: A Multi-temporal InSAR Coherence Magnitude Filter
9.1 Rationale
9.2 The Filter Machinery
9.3 Generation of a Testing Dataset
9.4 Test Cases Using TanDEM-X Data
9.5 Temporal Features
References
Chapter 10: Proxies of Forest Volume Loss and Gain by Differencing InSAR DSMs: Fingerprints of Forest Disturbance
10.1 Motivation
10.2 Study Site
10.3 TanDEM-X Data
10.4 Methods
10.4.1 DSM Difference Data Set Generation and Calibration
10.4.2 Object-Based Change Detection
10.4.3 Change Objects Refinement
10.4.4 Variance of the Within-Object Mean Height Difference Estimator
10.4.5 Effect Size
10.4.6 Probability of object detection by statistical decision theory
10.4.6.1 Neyman–Pearson approach
10.4.6.2 Bayesian Approach
10.4.7 Object Shape
10.4.8 Characterization of Objects by Contextual Information
10.4.8.1 Distance from Roads
10.4.8.2 Attributes by Land Management
10.5 Factors Influencing the DSM Change Magnitude
10.5.1 Forest Vertical Structure and Spatial Distribution (Forest Density)
10.5.2 Environmental Conditions (Seasonality and Rainfall)
10.5.3 Dependence on Instrument Parameters
10.5.3.1 Volume Only
10.5.3.2 Volume over Ground
10.6 Analysis
10.6.1 ∆ DSM Magnitude and Area Descriptive Statistic
10.6.2 Standard Error of the Object Mean
10.6.3 Effect Size
10.6.4 Object Detection by Statistical Decision Theory
10.6.5 Spatial Location of Objects
10.6.6 Objects’ Proximity to Roads
10.6.7 Change in Objects by Land Management
10.6.8 Shape Analysis
10.6.8.1 Fractal Exponent
10.6.8.2 Rectangularity
10.6.8.3 Regular Boundary Shapes in Land Management Units
10.7 Comparison between Objects Detected by InSAR ΔDSM and by Optical Imagery
10.8 Concluding Remarks
References
Appendix: A Wavelet Tour
A.1 Signal Representation in a Basis
A.2 The Fourier Kingdom
A.3 Extension to Linear Transforms with More Interesting Atoms – Where Wavelets Finally Appear
A.4 The Wavelet Transform in a Discrete Time Setting
A.5 Computing the Wavelet Frame Transform: The “à trous” Algorithm
A.6 The Multiresolution Wavelet Representation
References
Index
A
B
C
D
E
F
G
H
I
J
K
L
M
N
O
P
Q
R
S
T
U
V
W