Statistical Comparison Of SAR Backscatter From Icebergs .
“Statistical Comparison of SAR Backscatter from IcebergsEmbedded in Sea Ice and in Open Water using RADARSAT-2Images of in Newfoundland waters and the Davis Strait”ByUmma Hafsa Himi, B.Sc.A thesis submitted to the school of Graduate Studies in partialfulfillment of the requirements for the degree ofMaster of EngineeringElectrical Engineering ProgramMemorial UniversityOctober 2019St. John’s Newfoundland & Labrador
THIS PAGE LEFT INTENTIONALLY BLANK
AbstractIcebergs are considered a threat to marine operations. Satellite monitoring of icebergs isone option to aid in the development of iceberg hazard maps. Satellite synthetic apertureradar (SAR) is an obvious choice because of its relative weather independence, day andnight operation. Nonetheless, the detection of icebergs in SAR can be a challenge,particularly with high iceberg areal density, heterogeneous background clutter and thepresence of sea ice.This thesis investigates and compares polarimetric signatures of icebergs embedded in seaice and icebergs in open water. In this thesis, RADARSAT-2 images have been used foranalysis, which was acquired over locations near the coastline (approximately 3-35 km) ofthe islands of Newfoundland and Greenland. All icebergs considered here are in the lowerincident angle range (below 30 degrees) of the SAR acquisition geometry. For analysis,polarimetry parameters such as co- (HH) and cross- (HV) polarization and severaldecomposition techniques, specifically Pauli, Freeman-Durden, Yamaguchi, Cloud-Pottierand van Zyl classification, have been used to determine the polarimetric signatures oficebergs and sea ice. Statistical hypothesis tests were used to determine the differencesamong backscatters from different icebergs. Statistical results tend to show a dominantsurface scattering mechanism for icebergs. Moreover, icebergs in open water producelarger volume scatter than icebergs in sea ice, while icebergs in sea ice produce largersurface scatter than icebergs in open water. In addition, there appear to be minor observabledifferences between icebergs in Greenland and icebergs in Newfoundland.i
THIS PAGE LEFT INTENTIONALLY BLANKii
AcknowledgementsThis thesis would not be possible without support from several people and several sources.Foremost, I would like to express my sincere thanks and gratitude to my supervisors, Dr.Peter McGuire and Desmond Power for their continuous support of my M.Eng. study andresearch. Besides teaching me scientific research, they are icons of enormous motivationand patience who helped me keep on the right track and overcome many obstaclesthroughout my research. I am truly blessed to have such knowledgeable and nicesupervisors during my program.Besides my supervisors, I thank Centre for Cold Oceans Resources Engineering (C-CORE)and the Natural Sciences and Engineering Research Council of Canada (NSERC) forproviding me sufficient funds, a wonderful office environment and all the equipmentneeded for my research.I would also like to thank C-CORE, Defence Research & Development Canada (DRDC),Canadian Space Agency and the Canadian Ice Service for the provision of satellite imageryfor this study. The satellite imagery analysed herein was made available through acollaboration agreement established for my NSERC grant. These data were collected andground truthed through funding provided from several different research programs, withiii
funds made available through DRDC, Cairn Energy, and the Research and DevelopmentCorporation of Newfoundland and Labrador (RDC)1.My sincere thanks also to Dr. Bahram Salehi, who was my instructor in the ‘AppliedRemote Sensing’ course. It was a well-organized course that taught me all the basics ofremote sensing. I also thank C-CORE employees, Carl Howell and Pamela Burke forhelping me to collect data for my research which was very time consuming.Finally, I thank specially to my Husband, Saimoom, who was always beside me and myparents for their selfless support and encouragement for my scientific discoveriesthroughout my education.1RDC has since been renamed InnovateNL.iv
Table of Contents1. Introduction . 11.1 Purpose of the Study . 42. Literature Review . 62.1 Fundamentals of SAR . 62.1.1 Synthetic Aperture Radar (SAR) . 72.1.2 SAR imaging . 82.1.3 SAR Polarization . 122.1.4 Ocean Response to SAR . 152.1.5 Icebergs and Sea Ice Response to SAR . 162.2 Polarimetric Decomposition. 182.2.1 Pauli Decomposition. 202.2.2 van Zyl Decomposition . 222.2.3 Freeman-Durden Decomposition . 232.2.4 Yamaguchi Decomposition . 252.2.5 Cloud-Pottier Decomposition . 26v
2.3 Statistical Tests. 282.3.1 Two Tail T-test . 302.3.2 One Tail T-test . 322.4 Previous Work . 322.4.1 Properties of Sea Ice and Icebergs . 322.4.2 C-CORE’s Research on Iceberg Detection with SAR. 332.4.3 Detection of Icebergs in Sea Ice using Polarimetric RADARSAT-2 Data . 352.4.4 Iceberg Detection Using Full Polarimetric RADARSAT-2 Data in WestAntarctica. 362.4.5 Automatic Iceberg Detection in Open Water and Sea Ice . 373. RADARSAT-2 Data . 383.1 Data . 393.2 Target Detection . 443.3 Iceberg Detection Using Ground Truth information . 464. Methodology . 494.1 SAR Image Processing . 50vi
4.1.1 Polarimetric Feature Extraction . 504.1.2 Image Clipping . 504.1.3 Masking . 514.1.4 Filtering . 534.2 Polarimetric Decompositions . 544.3 Hypothesis Test . 564.3.1 Two tail T-test . 574.3.2 One tail T-test . 595. Results . 615.1 Radar Backscatter Plots. 615.2 Decomposition Results. 635.2.1 Icebergs in Open Water Versus in Sea Ice (NLSI Versus NLOW) . 645.2.2 Greenland Versus Newfoundland icebergs (GLOW versus NLOW) . 725.2.3 Statistical Comparison of Iceberg Groups . 775.3 Hypothesis Test . 825.3.1 Two Tail T-test . 82vii
5.3.2 One Tail T-Test . 855.4 Summary . 875.5 Discussion . 886. Conclusion . 946.1 Limitations . 956.2 Future Work . 967. References . 98Appendix I . 103Appendix II . 105Appendix III . 107Appendix IV . 109viii
List of TablesTable 3.1Image acquisition details42Table 4.1Normality test, Pauli Volume (Iceberg in Sea Ice)58Table 4.2Two tail T-test (assuming unequal samples) for Pauli 60decomposition volume scattering (sea ice versus open water)Table 4.3One tail T-test of Freeman-Durden surface scattering in sea ice61and open waterTable 5.1Two tail T-test of NLSI and NLOW group84Table 5.2Two tail T-test of NLOW and GLOW group85Table 5.3One tail T-test of NLSI and NLOW group (surface)86Table 5.4One tail T-test of NLSI and NLOW group (double bounce, 87volume, helix)Table 5.5One tail T-test of NLOW and GLOW groupix88
List of FiguresFigure 1.1Iceberg’s journey from Greenland to the Grand Banks of1Newfoundland and Labrador (figure courtesy of C-CORE)Figure 2.1Basic radar imaging7Figure 2.2SAR imaging geometry9Figure 2.3Horizontal and vertical polarization12Figure 2.4Backscatter Mechanism of iceberg and sea ice17Figure 2.5Types of decomposition19Figure 2.6Flow chart of Freeman-Durden decomposition (adapted from24Lee and Pottier (2009))Figure 2.7𝐻/𝛼 plane27Figure 2.8Visual illustration of the one tail and two tail tests; Left: one-31tailed T-test, although the picture is shaded on the right, it’s amirror image (critical region in the left will also be a one tailedtest, the condition of rejection of null hypothesis then be, t-stat -t-critical ); Right: two tailed T-testFigure 3.1Grouping of data sets for analysisx39
Figure 3.2Study area41Figure 3.3Top: iceberg identifying using pixel/line information from47ground truth data in PCI Geomatica; Bottom: Several icebergsthat have been clipped from HV channel from a sceneFigure 3.4Iceberg backscatter in SAR in different polarization channels48Figure 4.1Flow chart of the processing chain51Figure 4.2Clutter masking54Figure 4.3Iceberg completely merged with sea ice clutter in both HV (left)55and HH (right) channel; though it has the ground truthinformation (From field survey).Figure 5.1Comparison of polarimetric parameters (intensity) Left: HH vs63HV; Right: HH vs VVFigure 64Newfoundland and Greenland icebergs; Left: HH vs HV; Right:HH vs VVFigure 5.3Pauli decomposition of NLSI and NLOW group66Figure 5.4Freeman-Durden decomposition of NLSI and NLOW group67xi
Figure 5.5van Zyl decomposition of NLSI and NLOW groups68Figure 5.6van Zyl decomposition of NLSI and NLOW groups69Figure 5.7(a) Backscatter comparison using Cloude-Pottier decomposition71(b) 𝐻/𝛼 planeFigure 5.8Cloud-Pottier decomposition of a sea ice chip72Figure 5.9Freeman-Durden decomposition of NLOW and GLOW groups74Figure 5.10van Zyl decomposition results; Top: NLOW, Bottom: GLOW75Figure 5.11Yamaguchi decomposition results; Top: NLOW; Bottom:77GLOWFigure 5.12(a) Scatter plot of Cloude-Pottier decomposition; (b) 𝐻/𝛼 plane78Figure 5.13Box Plot representation of Pauli decomposition results81Figure 5.14Box Plot representation of Freeman-Durden decomposition82resultsFigure 5.15Box Plot representation of van Zyl decomposition results82Figure 5.16Box Plot representation of Yamaguchi decomposition results83xii
List of Abbreviations and SymbolsdBNormalized unitless intensity measureCISCanadian Ice ServiceRS-2RADARSAT-2RS-1RADARSAT-1SARSynthetic Aperture RadarHHHorizontal transmit and Horizontal receive polarizationHVHorizontal transmit and Vertical receive polarizationVHVertical transmit and Horizontal receive polarizationVVVertical transmit and Vertical receive polarization𝐻Entropy𝐴Anisotropy𝛼Alpha angle𝛿Significance level𝑆𝐻𝐻Scattering intensity value in HH channel𝑆𝐻𝑉Scattering intensity value in HV channel𝜎0Radar cross sectionSLCSingle look complexCFARConstant False Alarm Ratexiii
THIS PAGE LEFT INTENTIONALLY BLANKxiv
1. IntroductionCalving of icebergs at the tidewater glacier fronts is a component of the regular mass lossfrom glaciers and ice sheets in Arctic regions (Dierking & Wesche, 2014). TheNewfoundland and Labrador region can have hazardous environmental conditions thatthreaten exposed human-made structures due to extreme ice conditions. This areaexperiences thousands of icebergs and a large amount of sea ice every year.Figure 1.1: Iceberg’s journey from Greenland to the Grand Banks of Newfoundland andLabrador (figure courtesy of C-CORE)1
Greenland is the birthplace of most of the icebergs that frequent the coastline of easternCanada. Icebergs, unlike sea ice, originate from the land, specifically from glaciers. Whensnowfall exceeds ablation, over time the snow accumulates and compresses to form ice.This ice then flows from higher altitude regions to sea level. The glacier interface at theocean is constantly exposed to tides and sea level fluctuation that cause large pieces of iceto calve (Ulaby et al., 2014). These icebergs make their way to the Grand Banks over manyfreezes and thaw seasons (shown in Figure 1.1). Over the winter, sea ice and cold weatherconditions protect icebergs from degradation as they drift south through the Baffin Bay andthe Davis Strait. When the sea ice starts to melt in the spring, icebergs break away and it isthen that they become a hazard to shipping and marine operations. However, once freefrom the sea ice, icebergs degrade quickly due to weathering from the ocean, rains, windand temperature.A means of surveilling icebergs while they are drifting south with the sea ice pack iscompelling because it would allow marine operators to determine the number of icebergsthat may become threats once they break away from the pack. This study is equallyimportant to weather forecasters to set the initial conditions of icebergs in drift models toestimate the size and spatial distribution of icebergs breaking up. Satellite syntheticaperture radar (SAR) is one such solution, due to its wide areal coverage, day and nightand all-weather acquisition capabilities. SAR is also the de facto standard for themonitoring of ice in extreme northern and southern regions. SAR has become a standardfor monitoring of icebergs (see, for example Power et al. 2001; Lane et al. 2002; Howellet al. 2008; Wesche & Dierking 2012) and is presently used by the Canadian Ice Service2
(CIS), the International Ice Patrol and the Danish Meteorological Institute for icebergsurveillance.Icebergs manifest themselves as bright echoes against the darker ocean backscatter and thedetection of icebergs in SAR can generally be accomplished using an adaptive thresholdtechnique known as the constant false alarm rate (CFAR) (Skolnik, 2008). Nonetheless,the detection of icebergs within the pack is difficult because SAR backscatter from sea icecan be similar to that of icebergs. Standard CFAR techniques can also produce largenumbers of false alarms when applied to sea ice regions. Therefore, SAR detection oficebergs in sea ice is fairly challenging.New features in recent and upcoming satellite SAR missions promise to lead to improvedcapabilities for iceberg and sea ice monitoring both on the Grand Banks and in Arctic andsubarctic regions. The type of sea ice in which icebergs are present play a role in SARbackscatter and in the context of analysis and detection of icebergs (Dierking & Wesche,2014). The ice surface characteristics have a strong influence on radar intensity. There aremany different types of sea ice, based on their thermodynamic and geophysical evolutionthroughout the seasons. Dierking & Wesche (2014) state that the potential of SAR remotesensing for iceberg detection depends on several factors: i) physical properties of icebergssuch as size, shape and structure ii) SAR sensor specific properties such as incident angle,frequency band, resolution and polarization; iii) geophysical parameters such as winds, seastate, surface currents and season and iv) the backscatter of the surrounding sea ice or openwater. Since space borne SAR systems can image the ocean and their sea ice regions3
independent of light and cloud conditions, their potential for iceberg monitoring has beenextensively investigated.1.1 Purpose of the StudyThe purpose of this research is to investigate the comparative characteristics of icebergssurrounded by sea ice and open water in SAR images. The SAR response from icebergs isgenerally complicated compared to other natural surfaces due to the geometric shapes oficebergs, the composition of the iceberg glacial ice and the presence or absence ofmeltwater on the surface of the iceberg. In addition, minor variations of temperature, windspeed and direction, SAR incident angle and the line of sight direction of the satellite canaffect the results. Thus, iceberg polarimetric response varies by a significant number ofparameters. Therefore, further study of iceberg polarimetric response in differentconditions is important to optimize the surveillance of icebergs and enhance icebergdetection in sea ice. The main objective of this study is to determine the variability of theSAR response of icebergs in various aspects. In particular, the va
Figure 2.1 Basic radar imaging 7 Figure 2.2 SAR imaging geometry 9 Figure 2.3 Horizontal and vertical polarization 12 Figure 2.4 Backscatter Mechanism of iceberg and sea ice 17 Figure 2.5 Types of d
7705 SAR-8 7705 SAR-X 7705 SAR-A 7705 SAR-M 7705 SAR-H 7705 SAR-Hc 7705 SAR-W 7705 SAR-Ax 7705 SAR-Wx. 2 Daa okia 7705 ic ggraion oer Reliable service delivery SR 0S software provides superior QoS on the 7705 SAR. The same level of deep buffering and support for ingress and egress shaping that isFile Size: 696KBPage Count: 16
7705 SAR-8 or 7705 SAR-H Substation Control center 5620 SAM 7705 SAR-18 s Switching device Collector WiMAX/LTE CPE Voltage controller WiMAX/LTE base station Microwave packet transport (IP/MPLS) 7705 SAR-Hc or SAR-W 7705 SAR-Hc or SAR-W 7705 SAR-Hc 7705 SAR-Hc 7705 CabinetSAR-Hc Figure 3. Alcatel-Lucent IP/MPLS products in the field area .
Nov 30, 2000 · SUBJECT: Suspicious Activity Report (SAR) ENCL: (1) SAR Bulletin, Issue 3, June 2001 (2) Identity Theft Highlighted in SAR Activity Review, June 18, 2001 (3) SAR Activity Review – Trends, Tips & Issues,
7705 SAR-M 7705 SAR-H 7705 SAR-Hc 7705 SAR-Wx 7705 SAR-Ax. 2 Data sheet Nokia 7705 Service Aggregation Router Routing and resiliency The depth of its IP/MPLS/segment routing protocol implementation and robust OAM capabilities
networking [1, 43, 48, 56, 87]. Backscatter sensors can wirelessly communicate at near-zero power by simply reflecting radio signals in the environment. In this paper, we investigate the ability to take backscatter networking to underwater environments. In particular, since wireless communication is the largest source of energy con-
underwater backscatter localization poses new challenges that are different from prior work in RF backscatter localization (e.g., RFID localization [14, 25, 26, 41]). To answer this question, in this section, we provide background on underwater acoustic channels, then explain how these channels pose interesting new challenges for
devices. The Intel WISP [5] and UMass Moo [27] are examples of backscatter-enabled sensor platforms. Despite the energy benefits of backscatter radios, ex-isting network stacks achieve only short communication range and low throughput. We make an empiric argu-ment these limitations are, in part, due to the design of the network stack.
The Element Encyclopedia of Secret Signs and Symbols The Ultimate A-Z Guide from Alchemy to the Zodiac Adele Nozedar. For Adam and for the seven secrets ‘In every grain of sand there lies Hidden the soil of a star’ Arthur Machen ‘I do not need a leash or a tie To lead me astray In the land where dreams lie’ Yoav In Nature’s temple, living pillars rise Speaking sometimes in words of .
