University of Southern California Dissertations and Theses

Assessing modern conflict to monitor human rights with remote sensing: Russia's war in Ukraine

(USC Thesis Other)

Assessing modern conflict to monitor human rights with remote sensing: Russia's war in Ukraine

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

ASSESSING MODERN CONFLICT TO MONITOR HUMAN RIGHTS WITH REMOTE
SENSING:

RUSSIA’S WAR IN UKRAINE

by

Rebecca Bosworth

A Thesis Presented to the
FACULTY OF THE USC DORNSIFE COLLEGE OF LETTERS, ARTS AND SCIENCES
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
MASTER OF SCIENCE
(GEOGRAPHIC INFORMATION SCIENCE AND TECHNOLOGY)

August 2023

Copyright 2023 Rebecca Bosworth
ii

This work is dedicated to the innocent civilians of Russia’s war in Ukraine
iii
Acknowledgements
Thank you to my parents who have given me limitless opportunities to see the world and taught
me the value of freedom. You are my biggest inspiration. I am grateful to my family, friends,
mentors, and colleagues for encouraging my pursuits at USC. Thank you to the American
Society of Photogrammetry and Remote Sensing for supporting my education. This work was
made possible with your support.
iv
Table of Contents
Acknowledgements ........................................................................................................................ iii
List of Tables ................................................................................................................................. vi
List of Figures ............................................................................................................................... vii
Abbreviations ................................................................................................................................. ix
Abstract ........................................................................................................................................... x
Chapter 1 Introduction .................................................................................................................... 1
1.1 Motivation ........................................................................................................................... 1
1.1.1 History of the Russia-Ukraine Conflict ..................................................................... 2
1.1.2 Human Rights Violations and International Humanitarian Law ............................... 5
1.1.3 Human Rights Violations in Ukraine ......................................................................... 6
1.2 Study Area .......................................................................................................................... 9
1.2.1 Significance of Study Area ...................................................................................... 10
1.3 Project Overview .............................................................................................................. 12
1.3.1 Constraints ............................................................................................................... 12
1.3.2 Data and Methods .................................................................................................... 13
1.4 Thesis Overview ............................................................................................................... 14
Chapter 2 Literature Review ......................................................................................................... 15
2.1 Remote Sensing of Building Damage ............................................................................... 15
2.1.1 Satellite Remote Sensing Overview......................................................................... 15
2.1.2 Synthetic Aperture Radar Overview ........................................................................ 16
2.1.3 Building Damage Detection ..................................................................................... 17
2.2 Remote Sensing for Human Rights................................................................................... 21
2.2.1 Satellite Remote Sensing Applications for Human Rights Monitoring ................... 21
2.2.2 Synthetic Aperture Radar Applications for Human Rights Monitoring .................. 26
Chapter 3 Methods ........................................................................................................................ 31
3.1 Data ................................................................................................................................... 31
3.1.1 Sentinel-1 SAR Imagery .......................................................................................... 33
3.1.2 UNOSAT Rapid Damage Assessment..................................................................... 35
3.2 Workflow .......................................................................................................................... 37
3.2.1 Imagery Preparation ................................................................................................. 41
3.2.2 Calculate Log Difference ......................................................................................... 42
3.2.3 Manual Threshold .................................................................................................... 43
3.2.4 Zonal Statistics as Table .......................................................................................... 44
3.2.5 Sentinel-1 SAR Damage and UNOSAT Damage Evaluation ................................. 44
3.2.6 Near Tool Analysis .................................................................................................. 44
Chapter 4 Results .......................................................................................................................... 46
v
4.1 Sentinel-1 Damage Assessment ........................................................................................ 46
4.1.1 Spatiotemporal Results ............................................................................................ 46
4.1.2 Damage Statistics ..................................................................................................... 67
4.2 UNOSAT Comparison ...................................................................................................... 68
4.2.1 Summarize Within ................................................................................................... 69
4.2.2 Near Distance ........................................................................................................... 69
4.2.3 Near Angle ............................................................................................................... 70
Chapter 5 Discussion .................................................................................................................... 73
5.1 Limitations and Challenges............................................................................................... 73
5.1.1 Sentinel-1 ................................................................................................................. 73
5.1.2 UNOSAT ................................................................................................................. 76
5.2 Future Research ................................................................................................................ 78
5.2.1 Backscatter Intensity Threshold Selection ............................................................... 78
5.2.2 Alternative Analysis Tools ...................................................................................... 79
5.3 Conclusion ........................................................................................................................ 79
References ..................................................................................................................................... 81

vi
List of Tables
Table 1. Examples of remotely sensed human rights violations ................................................... 24
Table 2. Analysis methods for conflict-induced effects ............................................................... 26
Table 3. Sentinel-1 SAR imagery ................................................................................................. 32
Table 4. UNOSAT rapid damage assessment ............................................................................... 32
Table 5. Mean percentage statistics of undamaged and damaged buildings ................................ 68
Table 6. Near distance of UN damage points to SAR damage polygons ..................................... 69

vii
List of Figures
Figure 1. Ukraine’s ethnic composition .......................................................................................... 3
Figure 2. Russian language affinity in Ukraine .............................................................................. 4
Figure 3. Map of Ukraine highlighting Crimea and Donbas regions ............................................. 9
Figure 4. Study area of Mariupol, Ukraine ................................................................................... 10
Figure 5. Mariupol’s significance during Russia’s war in Ukraine .............................................. 11
Figure 6. Satellite observation and backscattering of ground objects) ......................................... 18
Figure 7. Double bounce effect from intact buildings .................................................................. 19
Figure 8. Increased (L) and decreased (R) backscatter intensity after building damage .............. 20
Figure 9. Examples of SAR image intensity ................................................................................. 20
Figure 10. Sentinel-1 SAR image from February 28, 2022 .......................................................... 34
Figure 11. UNOSAT rapid damage assessment ........................................................................... 36
Figure 12. Distance rings for UN points representing damaged buildings ................................... 37
Figure 13. Workflow for Sentienl-1 SAR damage assessment..................................................... 39
Figure 14. Workflow for Sentinel-1 and UNOSAT damage evaluation ...................................... 41
Figure 15. Intensity change between February 16, 2022 and May 23, 2022 ................................ 47
Figure 16. Damage assessment using SAR for February 16 – 28, 2022 ...................................... 48
Figure 17. Zhovtnevyi: February 16 – 28, 2022 and February 28 – March 12, 2022 .................. 49
Figure 18. Mariupol maternity hospital attack imagery................................................................ 50
Figure 19. Livoberezhnyi: February 16 – 28, 2022 and February 28 – March 12, 2022 .............. 51
Figure 20. Zhovtnevyi: February 28 – March 12, 2022 and March 12 – 24, 2022 ...................... 52
Figure 21. Mariupol drama theatre imagery before attack ........................................................... 53
Figure 22. Mariupol drama theatre imagery after attack .............................................................. 54
viii
Figure 23. Livoberezhnyi: February 28 – March 12, 2022 and March 12 – 24, 2022 ................. 55
Figure 24. Zhovtnevyi: March 12 – 24, 2022 and March 24 – April 5, 2022 .............................. 56
Figure 25. Livoberezhnyi: March 12 – 24, 2022 and March 24 – April 5, 2022 .......................... 57
Figure 26. Zhovtnevyi: March 24 – April 5, 2022 and April 5 – 17, 2022 .................................. 58
Figure 27. Livoberezhnyi: March 24 – April 5, 2022 and April 5 – 17, 2022.............................. 59
Figure 28. Zhovtnevyi: April 5 – 17, 2022 and April 17 – 29, 2022 ............................................ 60
Figure 29. Livoberezhnyi: April 5 – 17, 2022 and April 17 – 29, 2022 ....................................... 61
Figure 30. Zhovtnevyi: April 17 – 29, 2022 and April 29 – May 11, 2022.................................. 62
Figure 31. Livoberezhnyi: April 17 – 29, 2022 and April 29 – May 11, 2022 ............................. 63
Figure 32. Zhovtnevyi: April 29 – May 11, 2022 and May 11 – 23, 2022 ................................... 64
Figure 33. Livoberezhnyi: April 29 – May 11, 2022 and May 11 – 23, 2022 .............................. 65
Figure 34. Damage assessment using SAR for February 16 – May 23, 2022 .............................. 66
Figure 35. Distribution of near distance ....................................................................................... 70
Figure 36. Near angle of UN damage points to SAR polygons .................................................... 71
Figure 37. Distribution of near angle ............................................................................................ 71
Figure 38. SAR damage pixels overlaying ground features in Zhovtnevyi District ..................... 74
Figure 39. SAR damage polygons and UNOSAT damage points over Mariupol drama theatre . 75
Figure 40. UN damage point representing Mariupol Drama Theatre ........................................... 77

ix
Abbreviations
AOI Area of Interest
ASF Alaska Satellite Facility
DN Digital Number
ESA European Space Agency
GIS Geographic Information System
MARS Mass Atrocity Remote Sensing
NGO Non-Governmental Organization
RTC Radiometric Terrain Correction
SAR Synthetic Aperture Radar
UN United Nations
UN OHCHR United Nations Office of the High Commissioner for Human Rights
UNOSAT United Nations Satellite Centre
VHR Very High-Resolution

x
Abstract
Russia’s unprovoked attack on Ukraine on February 24, 2022, sparked the largest armed conflict
in Europe since World War II. As war in Ukraine continues, widespread reports of violations of
human rights and international humanitarian law accompany extensive civilian casualties.
Satellite imagery has provided unprecedented awareness of Russia’s war to corroborate
testimonial evidence of human rights violations. While the use of satellite imagery is now
commonplace to aid such efforts, human rights groups need improved remote sensing methods in
active war zones. The objective of this study is to evaluate the suitability of freely accessible
medium-resolution synthetic aperture radar (SAR) imagery from the European Space Agency’s
(ESA) Sentinel-1 satellite versus expensive very high-resolution (VHR) optical imagery for the
purpose of detecting war-induced building damage. The study area is the Ukrainian city of
Mariupol, which was seized by Russia in May 2022. The study assesses building damage using
backscatter intensity changes between images over time. Detected damage in conjunction with
reports of civilian casualties may indicate potential violations of international humanitarian law.
This study’s results indicate cumulative building damage in both extent and magnitude
comparable to a United Nations damage assessment that relied on VHR optical imagery.
Statistics estimate 27% damage from February 2022 to May 2022, which is lower than the 32%
damage estimate by the UN for the same study area. While SAR imagery may provide less
accurate results compared to VHR optical imagery, the increased timeliness, accessibility, and
adaptability it offers may render SAR imagery analysis as a more feasible option for some
human rights practitioners.
1
Chapter 1 Introduction
The international human rights community needs improved remote sensing methods to detect
violations of human rights and international humanitarian law in conflict situations promptly and
safely. Russia’s full-scale invasion of Ukraine on February 24, 2022 sparked the largest armed
conflict in Europe since World War II (RAND 2022). Since the war’s onset, multiple sources
report Russia is conducting indiscriminate attacks on civilian areas in violation of international
humanitarian laws and human rights and with concomitant loss of life and damage to civilian
infrastructure. As the war continues, the need for timely detection and assessment of these
violations is vital. This thesis investigates the use of imagery from Synthetic Aperture Radar
(SAR) satellites to detect damage from Russian attacks. SAR imagery can be taken day or night
and is all-weather capable and cloud-penetrable, so it offers the possibility for change detection
in a greater range of conditions than optical imagery. The purpose of this study is to investigate
the feasibility of using SAR imagery for mass atrocity monitoring by the international human
rights community. This chapter introduces the research objective, motivation, study area, and
research constraints.
1.1 Motivation
On February 24, 2022, Russia conducted an unprovoked full-scale invasion of Ukraine
from multiple fronts to overthrow Ukraine’s Western-aligned government and bring it under
Russian control (Bowen 2023). The ongoing war has created a humanitarian crisis affecting
millions of civilians, spurring global food and energy crises, and shocking the global economy
(Levi and Molnar 2022; Margesson and Mix 2022; Torkington 2023; UN SDG 2022). Evidence
shows Russian violations of international human rights law and international humanitarian law.
2
Many of these actions amount to war crimes, including willful killings and attacks on civilians
(UN OHCHR 2023b). Shortly after Russia’s invasion, the UN Human Rights Council established
an Independent International Commission of Inquiry to examine allegations of human rights
abuses, international humanitarian law violations, and crimes relating to Russia’s aggression
against Ukraine (UN OHCHR 2023b).
Identifying and verifying violations of humanitarian law is extremely challenging and
often requires on-the-ground situational awareness not always possible in active war zones.
Remote sensing data can aid in damage and destruction assessments to better evaluate Russian
attacks in Ukraine. These assessments can corroborate other evidence of possible war crimes in
International Criminal Court. The goal of this thesis is to evaluate the use of Sentinel-1 SAR
imagery as an alternative to costly high-fidelity imagery to detect war-induced infrastructure
damage from Russian attacks. This analysis will promote work in pursuit of justice and
accountability of civilian targeting and international law violations.
1.1.1 History of the Russia-Ukraine Conflict
Tensions between Russia and Ukraine are rooted in deep historic and ethnic ties fueling
Russia’s reluctance to accept Ukraine’s independence. Since Ukraine’s independence from the
Soviet Union in 1991, Russia has negatively perceived Ukraine’s motions to align with Western-
style democracy, including the European Union and the North Atlantic Treaty Organization
(Masters 2022). These tensions are especially evident in Ukraine’s southern region of Crimea
and in the eastern Donbas. The Donbas region in eastern Ukraine consists of the Donetsk and
Luhansk regions called oblasts and is home to the highest proportion of ethnic Russians and
Russian-speaking population of any Ukrainian region except Crimea (Welt 2021; Yekelchyk
2015). Figure 1 illustrates the ethnic affinity in Crimea and Donbas.
3

Figure 1. Ukraine’s ethnic composition (Source: Inton 2014)
While the majority Russian speakers are located in eastern Ukraine, the majority
population of Donbas identifies as ethnic Ukrainian (Yekelchyk 2015). The ethnic divide is
further juxtaposed by the language composition. According to the same 2001 census, the
majority of the population in Donbas claimed Russian as their native language (Yekelchyk
2015). Figure 2 illustrates the linguistic affinity in Crimea and Donbas.

4

Figure 2. Russian language affinity in Ukraine (Source: Washington Post 2022)
While most of the population in Donbas claim Russian as their native language, the
majority also identifies as ethnic Ukrainians. This incongruity between ethnicity and language is
symbolic of the cultural assimilation of Ukrainians during the late Soviet period resulting in a
hybrid identity (Yekelchyk 2015).
Ukraine’s heavily industrialized eastern Donbas region maintained close economic ties
with Russia after independence from the Soviet Union. These ties fueled conflict in 2014, when
Russia annexed Crimea and armed Russian separatists in Donbas, justifying its actions with
claims of protecting Russian-speaking people in the east (Masters 2022). Donbas has been a hot
spot for civilian deaths, injuries, and infrastructure damage and destruction since 2014 hostilities.
Fighting culminated in Russia’s February 2022 “special military operation” claiming to protect
5
the civilian population, part of a long disinformation campaign reinforcing Russian false
narratives (U.S. State Department 2022). Since the invasion, devastation in Ukraine includes war
crimes, human rights abuses, and violations of international humanitarian law (UN OHCHR
2023a).
1.1.2 Human Rights Violations and International Humanitarian Law
The international community alleges Russia is guilty of violating international law and
committing war crimes and crimes against humanity (Mulligan 2023). Additionally, international
leaders including the UN Secretary General and the U.S. Secretary of State assert that the
conflict in Ukraine has led to human rights violations (Mulligan 2023). Evidence of these
atrocity crimes can assist prosecution of aggressors during international tribunals, including the
International Court of Justice, International Criminal Court, and European Court of Human
Rights (Mulligan 2023).
Law of war in the context of international law is often used interchangeably with the law
of armed conflict and international humanitarian law (Mulligan 2022). This paper uses these
terms interchangeably. Law of war regulates the initiation of use of force, conduct of conflict,
and protection of war victims (Mulligan 2022). The Hague Conventions of 1899 and 1907 and
the four Geneva Conventions of 1949 address methods of warfare regulation and protections for
non-combatants. Under these treaties, parties in conflict must adhere to engage legitimate
military targets and cannot direct attacks at civilians or protected objects (Mulligan 2022).
Evidence of breaches of the Geneva Conventions can constitute war crimes prosecuted in
International Criminal Court (Mulligan 2022). Human rights are universal laws protecting
individuals and groups against actions that impede fundamental freedoms and human dignity
(UN OHCHR 2001). International human rights law is distinct from international humanitarian
6
law but holds complimentary principles concerning protection of life, health, and dignity of all
human beings (ICRC 2010). The Universal Declaration of Human Rights adopted by the UN
General Assembly in 1948 is the main legal instrument of international human rights law (UN
n.d.) In armed conflict situations, human rights law reinforces International Humanitarian Law
(ICRC 2010).
Before the widely accepted use of satellite imagery, witness testimony, photography,
forensic evidence, and human rights researcher reporting were used as evidence in various
national and international courts (Hasian 2016; Herscher 2014). Reliance on witness testimony
came with various challenges, including reluctant observers or few surviving eyewitnesses.
International criminal court proceedings used satellite imagery for the first time following
genocidal massacre during the Bosnian War in the late 1990s (Kroker 2015; Lee et al. 1998).
During a UN Security Council meeting, Madeline Albright, in her role as U.S. ambassador to the
UN, presented photographic evidence of the Srebrenica and Zepa atrocities (Hasian 2016; Rohde
1995; Rotberg 2010). These “before” and “after” aerial and satellite photos revealed sites of
mass graves where an estimated 6,000 to 8,000 civilians were buried (Lee 1998; Rohde 1995).
This led Tribunal investigators to alleged massacre sites to collect evidence corroborating
witness accounts used for prosecution of war crimes (Rotberg 2010; United Nations International
Criminal Tribunal for the former Yugoslavia n.d.). This event marked a shift by legitimizing
remote sensing technologies used to investigate war crimes and human rights violations.
1.1.3 Human Rights Violations in Ukraine
Russia’s war in Ukraine has had devastating impacts on the civilian population. As of
February 15, 2023, the United Nations Office of the High Commissioner for Human Rights (UN
OHCHR) recorded 8,006 civilian deaths, 13,287 civilians injured, 8 million refugees, and 5.4
7
million internally displaced people. Widespread reports of violations of human rights and
international humanitarian law accompany extensive civilian casualties. Alleged crimes include
indiscriminate and mass killings, shelling of humanitarian corridors, and filtration operations
(forced interrogation and separation) of civilians and noncombatants from Russian-controlled
areas (Bowen 2023; UN 2023a).
The UN estimates nearly 18 million people in Ukraine need humanitarian assistance and
demands continue to rise rapidly. Massive damage and destruction to human infrastructure have
left hundreds of thousands of Ukrainians homeless while many are living in damaged homes or
in buildings ill-suited to provide protection during winter season in life-threatening sub-zero
temperatures (UNHCR 2022). Shelling from heavy artillery strikes, launch rocket systems, and
missile and air strikes are the cause of most of the civilian casualties reported by the UN
OHCHR (2023b). UN OHCHR (2023a) estimates over 90% of civilian casualties are caused by
explosive weapons with wide area effects used in populated areas. These attacks have damaged
or destroyed thousands of residential buildings, over 3,000 educational institutions, and more
than 600 medical facilities. Casualties are likely underestimated due to delayed reporting and
pending verification. Most attacks likely initiated by Russian armed forces have been determined
as indiscriminate, lacking specific military objective therefore violating international
humanitarian law (UN OHCHR 2023a).
Many countries have condemned Russia’s invasion of Ukraine as a violation of
international law governing the use of force and have identified examples of potential Russian
war crimes and human rights violations (Bowen 2023; Mulligan 2022). Evidence of violations of
international humanitarian law include indiscriminate attacks in densely populated areas
(Amnesty International 2022), attacks and mining of humanitarian corridors (Lister 2022), and
8
airstrikes on hospitals (Cullison 2022). However, the process of identifying, gathering
information, and proving international humanitarian law violations requires detailed fact-finding
for on-the-ground truth and can be extremely challenging to prove (Mulligan 2022). To address
some of these challenges, satellite remote sensing offers a means to document necessary
evidence within inaccessible active war zones. The scope of this study does not investigate
individual violation claims. Rather, it utilizes remote sensing methods to identify potential areas
of human rights violations. As war continues, remote sensing applications and geospatial
analysis can provide a more compressive understanding of the evolving ground situation to assist
the international human rights community.
9
1.2 Study Area
The study area is the coastal city of Mariupol, Ukraine located in the southern Donetsk
oblast. Figure 3 shows a map of Ukraine and highlights the Donbas region, comprised of
Luhansk and Donetsk, bordering western Russia.

Figure 3. Map of Ukraine highlighting Crimea and Donbas regions (Source: CRS 2021)
Figure 4 depicts the area of interest (AOI) of Mariupol, Ukraine comprised of two
districts. The two outlined zones, the Zhovtnevyi District (left side of AOI) and Livoberezhnyi
District (right side of AOI), make up some of the most heavily damaged civilian residential
areas. The study timeframe is February 16, 2022 (pre-Russian invasion) to May 23, 2022 (post-
capture of Mariupol). This study uses imagery collected on a 12-day revisit rate corresponding to
the temporal resolution of the Sentinel-1 satellite. Imagery provides a time-series of the same
10
satellite footprint of a consistent area of interest, including satellite imagery acquisitions with the
same path and frame.

Figure 4. Study area of Mariupol, Ukraine
1.2.1 Significance of Study Area
Mariupol was a key Russian military objective since the early stages of Russia’s invasion
(Bowen 2023). Mariupol is strategically important because of its location between Russian-
annexed Crimea in the south and separatist-controlled areas in Donbas. Analysts assessed
capture of the city could create a corridor between Russia and forcefully occupied Ukrainian
territories including Donbas and Crimea in addition to control of the Sea of Azov (Gardner 2022;
Ghaedi 2022; Parker et al. 2022; Vohra 2022). Figure 5 illustrates Mariupol’s geographic
strategic importance.
11

Figure 5. Mariupol’s significance during Russia’s war in Ukraine (Source: DW)
As a result of Russia’s objectives, Mariupol was one of the most devastated cities
suffering thousands of causalities and significant destruction (UN 2023a). Reports of constant
shelling and explosive weapons striking civilian buildings from Mariupol are included in the
Independent International Commission of Inquiry on Ukraine (UN OHCHR 2023a). Investigated
examples of human rights violations and international law violations include the indiscriminate
attacks on the Mariupol drama theatre that killed and injured many civilians, and the attack on
the Mariupol Maternity Ward No. 3 that resulted in at least two deaths (UN OHCHR 2023a).
After weeks of fighting, Russia announced seizure of Mariupol in late April 2022, followed by
Mariupol’s surrender in mid-May 2022 (Bowen 2023). Without access to the Donetsk region,
12
including Mariupol, the Commission has yet to make a sufficient determination of whether the
attacks and seizure of Mariupol constitute crimes against humanity. Remote sensing offers a
means to corroborate imagery with other sources, such as eyewitness testimony, in the absence
of direct access to Mariupol for investigative purposes. This research will map where potential
damage has occurred to corroborate human rights violations allegations.
1.3 Project Overview
The objective of this research is to evaluate the use of medium-resolution SAR imagery
to detect damage in the civilian residential areas of Mariupol, Ukraine due to Russian attacks
from February through May 2022. Final analysis is compared to the United Nations Satellite
Centre (UNOSAT) Rapid Damage Assessment results based on very high-resolution (VHR)
optical imagery. The overall goal is to assess the feasibility of using medium-resolution SAR
imagery in human rights contexts where use of more expensive, high-resolution optical imagery
may be less accessible. This research will recommend practical methods to aid human rights
efforts during Russia’s war in Ukraine. Data includes Sentinel-1 SAR imagery from the Alaska
Satellite Facility (ASF), geographic boundary data from UNOSAT, and building damage points
from UNOSAT. SAR backscattering intensity analysis is used to determine changes between
images representing potential war-induced damage. This research is tailored to human rights
practitioners in need of timely detection during conflict to record, assess, and prosecute potential
violations of human rights and international humanitarian law.
1.3.1 Constraints
Human rights practitioners investigating human rights violations in Ukraine require
timeliness, sufficient accuracy, low cost, and simplicity. These needs are evident during natural
or anthropogenic disasters where timely detection for emergency response is critical. Human
13
rights practitioners need access to reliable data in denied territories such as active war zones and
methods for quick detection of potential human rights and international humanitarian law
violations to document incidents promptly. While the highest fidelity imagery and most robust
methods are desirable, they are not always practical due to constraints which vary depending on
crisis. The data and methods of this project are chosen with this real-world context in mind.
Timely detection of potential human rights violations is prioritized over 100% accuracy
for the purposes of this research. Timely detection – defined herein as detection within hours or
days – allows researchers to identify focus areas, determine impacted populations, and work with
other organizations such as private research groups with access to higher fidelity imagery and
methods to refine analyses. Another constraint is sufficient accuracy. The cost of highly accurate
data can render it inaccessible to many human rights organizations, delaying detection and
assessment. For example, a review of the current state of satellite monitoring of armed conflicts
determines that commercial sub-meter WorldView-4 imagery from Maxar costs $22.50/km,
totaling US $13.6 million for the country of Ukraine (Bennet et al. 2022). Therefore, coarse yet
publicly accessible data is used in this project to test its suitability for providing information to
the international human rights community. Finally, this project prioritizes simplicity in its
methodology as its workflow should be reproducible by non-imagery experts in the field.
1.3.2 Data and Methods
This research employs the principle of SAR backscatter intensity changes to assess war-
induced damages and human rights violations. A SAR log intensity change method is adapted
from similar methods applied to studies of natural and anthropogenic disasters, including
earthquake and war-induced destruction (Aimaiti et. al 2022, Braun 2018, Matsuoka and
Yamazaki 2004). Data includes freely accessible Sentinel-1 SAR imagery through ASF, AOI
14
boundary data from UNOSAT, and geolocated building damage points derived from VHR
optical imagery from UNOSAT. All analysis is run in ArcGIS Pro using an imported Python
toolbox and various ArcGIS Pro geospatial analysis tools. Resulting cumulative damage
assessments from Sentinel-1 SAR imagery are compared to the UN damage assessment using
VHR optical imagery. Finally, relationships between UN damage points and Sentinel-1 SAR
damage pixels are examined.
1.4 Thesis Overview
This thesis includes the literature review, methodology, results, and discussion informing
the use of SAR imagery to assess potential human rights violations in active war zones. Chapter
2 provides a literature review expanding on the benefits and applications of SAR imagery as well
as methods for detecting building damage in the environmental studies and humanitarian fields.
Chapter 3 provides a description of the data and employed methodology, including imagery
preparation and analysis. Results are presented in Chapter 4, followed by a discussion in Chapter
5 on limitations, challenges, and proposed improvements for future studies.

15
Chapter 2 Literature Review
The purpose of this thesis is to evaluate the use of satellite remote sensing data to assess war-
induced building damage. This chapter introduces satellite remote sensing techniques used to
detect building damage followed by satellite remote sensing applications used for human rights
violations investigations. This literature review provides insights on existing methods and
challenges in the field of remote sensing for the detection of human rights violations.
2.1 Remote Sensing of Building Damage
Satellite remote sensing is used extensively for damage mapping and damage
assessments after natural disasters, where in-person data collection is often dangerous or
impossible. Such satellite-derived assessments assist damage extent surveys, search and rescue
operations, and reconstruction planning. A common imagery analysis method following natural
disasters utilizes pre- and post- event images for change detection (Dong and Shan 2013;
Korkmaz and Abualkibash 2018; Romaniello et al. 2017). While new methods for building
damage assessments in natural or anthropogenic disaster contexts continue to develop, extensive
limitations persist (Bennet et al. 2022; Dong and Shan 2013). The following sections provide an
overview of satellite remote sensing, SAR, and remote sensing science principles used for
building damage detection.
2.1.1 Satellite Remote Sensing Overview
Remote sensing refers to any technology that provides detection of physical phenomena
on Earth’s surface, such as destroyed buildings due to natural or anthropogenic disasters.
Platforms include, but are not limited to, manned aircraft, unmanned aerial systems, and
satellites. Satellite remote sensing involves measuring reflected and emitted electromagnetic
16
energy from a surface at a distance to detect physical characteristics (NASA n.d.a). There are
two kinds of satellite remote sensors: active and passive. An active sensor is a radar instrument
that transmits signals and measures reflected, refracted, or scattered signals from a surface
(NASA n.d.a). A passive sensor uses optical instruments and records electromagnetic waves
emitted by the sun and reflected from the Earth. A passive sensor example is an optical satellite,
which provides imagery easily interpreted by the human eye. Both active and passive sensors can
make valuable observations in inaccessible environments such as active war zones. This thesis
utilizes imagery from the European Space Agency (ESA) Sentinel-1 SAR satellite, an active
sensor. The benefits of low-risk acquisition, wide coverage area, and high temporal resolution
are discussed further in the following section.
2.1.2 Synthetic Aperture Radar Overview
SAR is a unique type of remote sensing technology that provides all-weather, all-day
imagery used to detect changes in the Earth’s surface after natural or human disturbance. SAR
satellites have active sensors that transmit electromagnetic energy and record reflected energy
called backscatter (ASF n.d.b). SAR uses microwave wavelengths, with most radar applications
operating within the 3mm to 30 cm range. These longer wavelengths give radar sensors the
unique capability to penetrate clouds, making SAR imagery an optimal choice during unreliable
weather conditions. The name is derived from the practice of combining a sequence of imagery
acquisitions from a shorter satellite antenna to provide higher resolution imagery (NASA n.d.b).
A unique feature of SAR instruments is a side-looking sensor, which differs from other satellites
that look straight down (nadir). This feature enables the SAR sensor to identify the location of
received waves on the ground and differentiate between features equidistant from the sensor on
opposite sides (ArcGIS Pro n.d.a). SAR satellites record backscatter in phase and amplitude to
17
render a 2D image. Phase provides information on distance between a sensor and target, and
amplitude indicates amount of sent signal that returns to the sensor. The digital number (DN) for
the amplitude of a SAR imagery pixel represents backscatter. A high DN corresponds to a strong
backscatter, while a low DN represents a weak backscatter. The amplitude strength of the
measured backscatter is used to discern features on the ground (ArcGIS Pro n.d.a). Many factors
influence backscatter returned to a SAR sensor, including sensor wavelength, surface roughness,
and a phenomenon known as the double bounce effect, explained in the next section (ASF n.d.b).
Because SAR is an active type of satellite sensor, it is not dependent on time of day because it
does not require sunlight to illuminate a target. SAR sensors offer the benefit of 24-hour, all-
weather capability. SAR offers benefits over its optical counterpart, which can be ineffective at
night or in the presence of clouds or smoke (Brown and Hogan 2020). This research utilizes SAR
satellite imagery over optical imagery due to these benefits, greatly reducing dependency on
optimal imaging conditions.
2.1.3 Building Damage Detection
SAR technology can be used as a powerful remote sensing tool to detect changes in the
Earth’s surface, including natural or human disturbance (ASF n.d.e). Changes in the Earth’s
surface, such as war-induced infrastructure damage, can be detected by radar reflections called
backscatter. The principle of SAR intensity (backscattering) change detection for building
damage assessment is based on weak reflection from collapsed buildings. Pioneers in the field of
SAR imagery, Matsuoka and Yamazaki (2004) investigate use of SAR intensity to detect
building damages after the devastating 1995 earthquake in Kobe, Japan. Their research reveals
significantly lower backscattering coefficient values and intensity correlation in pre- and post-
event images in severely damaged areas (Figure 6).
18

Figure 6. Satellite observation and backscattering of ground objects (Source: Matsuoka and
Yamazaki 2004)

Using coarse 30 m spatial resolution imagery, their study demonstrates the challenge of
identifying backscattering characteristics of individual buildings. Instead, the study proves the
possibility of detecting groups of damaged buildings. Previous studies evaluating backscatter in
the 1995 Kobe earthquake by Aoki et al. (1998) and Matsuoka and Yamazaki (2004) show that
man-made structures such as urban buildings exhibit high reflection. This is due to multiple
reflections, known as the cardinal effect. Normally, a non-collapsed building exhibits strong
backscatter effect caused by corner reflectors between intact structures and the ground. This
phenomenon is commonly referred to in literature as the double bounce effect, illustrated in
Figure 7. In contrast, open spaces and damaged buildings exhibit low reflectance when
microwaves are scattered in multiple directions (Figure 6). Changes in backscattering can be
indicative of changes due to destruction.
19

Figure 7. Double bounce effect from intact buildings (Source: Ge et al. 2020)
A key assumption for this thesis is that a damaged building will result in a significant
backscattering intensity change. Intensity change values can be positive or negative depending
on the geometry of the building and the nature of collapse (van Heyningen 2018). Figure 8
demonstrates this dependency. The left-hand side of Figure 8 demonstrates the following
scenario. A radar signal hitting the corner of an intact building returns a strong backscatter
signal. However, after a wall collapse, rubble forms a corner reflector and causes an increase in
backscatter signal. On the right-hand side of Figure 8, a radar signal hits an intact wall and
ground resulting in a double bounce (strong signal). Debris resulting from damage disperses
subsequent radar signals, resulting in decreased backscatter intensity.

20

Figure 8. Increased (L) and decreased (R) backscatter intensity after building damage (Source:
van Heyningen 2018)

While SAR imagery is not as easily interpreted by the human eye as its optical
counterpart, SAR backscatter intensity analysis offers valuable insights for potential
infrastructure damage. Figure 9 shows examples of intensity characteristics of high-resolution
SAR images in intact and collapsed building areas. Image a) shows post-event optical image of
intact buildings; b) post-event SAR image of intact buildings; c) pre-event optical image of
collapsed buildings; d) post-event optical image of collapsed building; e) post-event SAR image
of collapsed buildings (Cui et. al 2018).

Figure 9. Examples of SAR image intensity (Source: Cui et al. 2018)
21
The SAR intensity image of an intact building shows regular shadow and layover zones
and building features can be coarsely interpreted (b). In contrast, the SAR intensity image of a
damaged building shows random pixel distribution and identification of physical features is
extremely challenging (e). These “before” and “after” SAR images demonstrate the need for
SAR backscatter intensity change analysis to detect damages.
2.2 Remote Sensing for Human Rights
The increased use of satellite remote sensing came after the Gulf War in the early 1990s
when satellite technology introduced the ‘first space war’ (Anson and Cummings 1991; Witmer
2015) offering first-ever on-demand war coverage to the public (Datta 2022). This trend
expanded within the humanitarian realm, where satellite remote sensing was used to monitor the
2003 Darfur conflict in Sudan (Amnesty International 2004; Prins 2008; Witmer 2015). Satellite
imagery is a valuable alternative to other remote sensing platforms for situations demanding
large study areas, short acquisition and analysis timelines, or in remote areas or dangerous
conflict zones. Despite these benefits, limitations may include variable spatial resolution,
weather-related constraints, and high costs. Furthermore, while the number of commercial
satellite providers increases, privatization of satellite imagery may limit data accessibility for
humanitarian actors. The following sections review satellite remote sensing methods used for
human rights violation monitoring and SAR methods to detect war-induced damage.
2.2.1 Satellite Remote Sensing Applications for Human Rights Monitoring
Satellite imagery is increasingly used to identify crimes against humanity by
documenting the scale and method of human rights abuses and affected areas (Rotberg 2010).
This is primarily done with very high spatial resolution optical images (≤1 m), which enables
individual building scale analysis (Witmer 2015). Human rights abuse in conflict settings
22
analyses are conducted by non-government organizations such as Human Rights Watch and
Amnesty International, and intergovernmental organizations, such as the UN. Example satellite
imagery applications include detection of troop activity and village destruction in Sudan and
South Sudan (Harvard Humanitarian Initiative 2012), unlawful airstrike evaluation in Libya
(Human Rights Watch 2012b), identification of mass graves (Amnesty International 2016), and
fire detection of destroyed villages from conflict (UNOSAT 2011). While these organizations
can use satellite imagery as complimentary evidence to corroborate eyewitness testimony, lack
of new methods has contributed to slow progress in the field.
A standardized, universal forensic approach using satellite imagery to detect and
document human rights atrocities does not exist and is likely impractical (Raymond et al. 2014).
Every conflict varies in study area, affected populations, type of warfare, research objectives,
and imagery requirements. While this study cannot address the needs of every conflict, it applies
existing methods suited for constraints of the war in Ukraine. Spatiotemporal analysis will help
international aid workers identify humanitarian needs and assist human rights groups with
documenting impacts of violent conflict.
Although the number of organizations engaged in the use of remote sensing within the
humanitarian space is growing, efforts to professionalize and standardize the practice for mass
atrocities monitoring lags behind other fields (Marx 2013; Raymond et al. 2014; Witmer 2015).
Furthermore, lack of technical knowledge and training required to analyze remote sensing
imagery presents a challenge for non-imagery experts in the conflict research field (Witmer
2015). As a result of scarce documented practice, humanitarian practitioners are operating
without accepted forensic standards specific to confirming mass atrocity events. Aware of this
shortfall, Raymond et al. (2014) identifies the need for a standard forensic approach for high-
23
resolution satellite imagery used to document mass atrocities as its own discipline, referred to as
Mass Atrocity Remote Sensing (MARS). An object-based remote sensing method is proposed in
which activity patterns are categorized by observable phenomena to identify activity consistent
with mass atrocities. An example indicator of interest is intentional targeting of civilian
populations and forced displacement. This alleged action is observable by destroyed structures
consistent with civilian dwellings and facilities (observable object indicators). This observable
object indicator framework is applicable to human rights violations investigations for Russia’s
ongoing war in Ukraine.
MARS research differs from other disciplines using remote sensing due to the unique
operational challenges and requirements of monitoring conflict, including data availability and
technology (Raymond et al. 2014). Human rights remote sensing researchers adopt a general
approach using a standard sequence of steps. Researchers 1) identify desired violation, 2) select
remotely sensed phenomenon associated with the violation, and 3) select an appropriate sensor
that will detect the phenomenon (Marouf 2016; Marx and Goward 2013). Several studies by
Marx investigate damage detection methods using this framework for village burnings in
Myanmar and in Darfur (Marx and Loboda 2013; Marx et. al 2019) and bombings and missile
attacks against civilian neighborhoods in Syria (Marx 2016). These studies address shortfalls of
costly and labor-intensive methods with Earth-observing satellites to detect potential human
rights violations (Marx and Goward 2013; Marx and Loboda 2013). Table 1 summarizes
potential human rights violations that can be identified indirectly by specific signals
characteristics detected by various satellite sensors. Following the approach by Marx and
Goward (2013), this thesis aims to investigate human rights violations in the form of
24
indiscriminate Russian attacks on Ukrainian residential areas evidenced by building damage
detected by changes in SAR backscatter intensity from Sentinel-1 SAR imagery.
Table 1. Examples of remotely sensed human rights violations
Violation Phenomenon Signal Analysis Sensor Revisit Source
Artillery near
civilians
Artillery,
bomb craters
Identification of
craters near
civilians
WorldView 1
(0.5 meters)
2 weeks UNOSAT
2009

Mass
execution
Creation of
mass graves
Detection of
disturbed earth,
earthmovers
U-2 (unknown) n/a

NYT 1995

Homes
targeted by
ethnicity

Destruction of
individual
houses

Destruction of
destroyed roofs

DigitalGlobe (2
meters)

6
months

AAAS 2008

Civilian
buildings
targeted

Damage to
public buildings

Identification of
destroyed
buildings

WorldView 1
(0.5 meters)

n/a

UNOSAT
2008

Political prison
camps

Expansion of
prisons

Detection of
changes in the
size of prisons

Digital Globe (2
meters)

10 years

AI 2011

Civilian
buildings
targeted

Destruction of
forests, fields,
and villages

Detection of
changes in land
cover
classification

Landsat 5 (30
meters)

4 years

De Vos 2008
Civilian
population
removed
Abandonment
of agricultural
land
Detection of
changes in
agricultural fields
Landsat 5 (30
meters)
~4 years Witmer 2008

Villages
attacked

Burning of arid
villages

Detection of fires
by Moderate
Resolution
Imaging
Spectroradiometer

MODIS (250
meters)

Annual

Bromley
2010

Source: Marx and Goward 2013
MARS phenomena detection is dependent on factors including spatial resolution,
temporal resolution, and scale (localized or regional). Witmer (2015) summarizes the current
state of remote sensing conflict research and organizes observable phenomena based on detection
25
time. War-induced structural damage from bombs or fires is generally detectable from satellites
within minutes to hours; environmental damage takes hours to days, population movement takes
days to months, and land-cover and land-use change takes months to years. Applying this
observation timeline to the war in Ukraine, imagery capable of detecting war-induced structural
damage within minutes to hours is needed. Based on the current capabilities of satellite sensors,
Witmer (2015) suggests fine resolution imagery (1-10 m) for detecting destroyed buildings and
structures, with visual photointerpretation offering the easiest and most accurate identification
technique for small study areas. Manual inspection of 1-10 m resolution imagery is one option
for analysis in Ukraine but requires accessible data. The majority of MARS related imagery
analysis employ optical (visible spectrum) and near-infrared sensors (Error! Reference source
not found.). The current state of MARS research does not include extensive use of SAR
imagery.
26
Table 2. Analysis methods for conflict-induced effects
Category/effect Analysis method Sensors(s) Citation
Bomb impacts,
destroyed bridges, oil
spills, destroyed oil
tanks
Visual ID IRS, Landsat UNEP 1999

Bomb damage

Visual ID

QuickBird, IKONOS

UNEP 2003

Damaged structures

Visual ID, OO, MM &
PCD

IKONOS

Al-Khudhairy,
Caravaggi, and
Giada 2005

Destroyed and rebuilt
structures
Support Vector Machine
classification
IKONOS Pagot and Peraresi
2008

Village burned

Drop in village albedo

Landsat

Prins 2008

Village burned Detection of fires MODIS Bromley 2010

Huts burned Classification & MM to
identify huts
QuickBird Sulik and Edwards
2010

Village burned Near-infrared
reflectance decrease
Landsat Marx and Loboda
2013
Source: Witmer 2015
2.2.2 Synthetic Aperture Radar Applications for Human Rights Monitoring
Two types of remote sensing technologies applied to assess disaster-induced building
damage are optical and SAR sensors. Optical sensors provide images that can be easily
interpreted by the human eye. High spatial resolution optical satellite imagery is the most
frequently used Earth observation medium for post-natural disaster mapping. However, optical
satellite sensors require sun illumination and cannot image through clouds, greatly limiting use
as an emergency response tool (Ge et al. 2020). In contrast, SAR is not dependent on sun
illumination or impacted by clouds but is challenging to interpret and limited by speckle noise.
Due to its day/night and all-weather capabilities, SAR imagery is usually available earlier than
optical imagery. The flexibility and reliability of SAR appeals to conflict researchers seeking
27
assessment of building damage available for detection within minutes to hours (Ge et al. 2020;
Witmer 2015). SAR offers advantages over its optical counterpart making it a suitable sensor
choice for study of Russia’s war in Ukraine.
Various approaches must be considered for SAR-based building analysis, including
change detection approach, change detection method, and spatial scale. These options make it
difficult if not impossible to recommend a single approach for SAR-based building damage
assessment (Joyce et al. 2009). This research addresses the need for remote sensing solutions to
produce timely and accurate human rights violations by proposing a SAR framework for the war
in Ukraine. While scientific literature covering SAR-based building damage detection due to
environmental disasters offer insights (Dong and Shan 2013; Matsuoka and Yamazaki 2004),
few publications specifically address the remote sensing needs of conflict researchers and
organizations monitoring human rights violations. This thesis addresses these gaps by adapting
SAR-based building damage assessments for natural disasters and applying them to assess war-
induced damage.
Two types of SAR-based building damage detection approaches are change detection
(using both pre- and post-event data) and assessment (using only post-event data). Since imagery
from pre-Russian invasion and post-Russian invasion of Ukraine are available, this study will use
a change detection approach comparing pre- and post- event imagery.
Damage detection studies use two different scales of analysis: block unit or single
building level. Block unit change detection was first developed during the 1990s in the SAR-
based building damage assessment field due to image resolution limitations (Dong and Shan
2013; Ge et. al 2020). There are three types of block level building damage assessment: pixel-
based (or grid) analysis (Matsuoka and Yamazaki 2004), irregular blocks separated by urban
28
boundaries (Zhai and Huang 2016), and irregular blocks based on homogenous features (Gokon
et al. 2017). Other similar studies using block level SAR-based building damage assessments
from natural disasters use medium-resolution SAR imagery ranging from 8-30m and coarser
(Chini et al. 2009; Chini et al. 2013; Matsuoka and Yamazaki 1999). This study uses the 10 m
pixel size of Sentinel-1 imagery as the scale.
Three methods of change detection used in building damage assessment are intensity-
based, coherence-based, or polarimetry-based analysis. Coherence-based analysis and
polarimetry-based analysis are not considered for the scope of this study. Intensity-based
analysis can be used in any SAR satellite operating mode and exploits the amplitude information
of the backscattering from ground targets received by a SAR sensor. Intensity changes can
indicate ground changes caused by a disaster event. Generally, built-up structures exhibit high
backscatter values due to double bounce effects (building wall to ground). Some of the first
published investigations of SAR amplitude data analysis used for building damage assessments
found relationships between backscatter changes from ground targets using pre- and post-event
imagery (Matsuoka and Yamazaki 1999; Shinozuka and Loh 2004). In Matsuoka and
Yamazaki’s (1999) SAR-based study of the 1995 Hyogoken-Nanbu earthquake in Japan, SAR
backscatter values decrease with increasing damage. This principle is utilized extensively in
remote sensing for disaster management literature, and more recently by researchers applying
methods for detection of conflict-related damage due to the Syrian Civil War (Braun 2018) and
Russia’s War in Ukraine (Aimaiti et. al 2022).
While SAR imagery is used more commonly to assess building damages from natural
disasters, studies also use SAR to explore damages from armed conflict. Braun (2018) uses time-
series of Sentinel-1 radar imagery to identify building damage resulting from civil war from
29
2014-2017 in the city of Raqqa, Syria. Scattering mechanisms of built-up structures including
corner reflection, building materials, and orientation toward the sensor influence radar
amplitudes in urban areas. The study uses these principles to identify building presence and
identify changes probably caused by war. Results show that Sentinel-1 data can indicate heavy
damage, but is limited due to low spatial resolution inhibiting detection of moderate damages.
Finally, a UNOSAT dataset consisting of points representing damaged structures is used for
validation and shows Sentinel-1 analysis strongly underestimates changes indicative of damage.
Despite this shortfall, Braun (2018) concludes Sentinel-1 data is a highly suitable indicator for
severe damage in urban areas. With these limitations in mind, this thesis utilizes SAR imagery
and UNOSAT data to investigate potential for damage assessment due to war in Ukraine.
To the author’s knowledge, research by Aimaiti et. al (2022) is the only publication as of
March 2023 using SAR backscattering intensity change analysis as part of a building damage
assessment due to war in Ukraine. While the strong backscatter from damaged buildings usually
decreases or disappears when a building collapses due to a disaster, an overall increased
backscattering intensity can also result from a strong double bounce effect formed from partially
collapsed buildings and resulting corner reflectors (Matsuoka and Yamazaki 2004; Matsuoka and
Nobuoto 2010). Using these principles captured in a SAR log ratio of intensity for Sentinel-1
imagery, Aimaiti’s results classify 58% of damaged buildings correctly when compared with
UNOSAT damage assessment derived from very high-resolution optical imagery.
This study utilizes the SAR log ratio of intensity between images to detect and assess
war-induced damage in Mariupol, Ukraine (1):
𝐼 𝑟𝑎𝑡𝑖𝑜 = 10𝑙𝑜𝑔 10
(
𝐼 𝑁 𝐼 𝑁 +1
) (1)
30
where 𝐼 𝑁 is pre-event image and 𝐼 𝑁 +1
is post-event image. Calculating the log difference
between two images can identify areas of significant changes in backscatter over time (ASF
2020).
Building damage analysis evaluates the humanitarian cost of Russia’s war in Ukraine and
offers further insights of impacted civilian communities based on temporal and spatial
characteristics of damage. The next chapter describes methods used to assess war-induced
building damage due to Russia’s war in Ukraine applying concepts, approaches, and
methodology gaps described in this literature review.

31
Chapter 3 Methods
The goal of this study is to assess the use of medium-resolution, publicly available SAR imagery
to detect war-induced building damage in Mariupol, Ukraine due to the ongoing Russia-Ukraine
conflict. This chapter provides a methods overview for SAR imagery analysis attributing
backscatter intensity change to Russian attacks. Data and methods were selected to address the
needs of the international human rights and humanitarian law communities for timely and
accurate detection of mass atrocities. Results were compared to the UNOSAT Rapid Damage
Assessment and data derived from very VHR optical imagery. Final analysis provides spatial
insights for the war in Ukraine and expands research in SAR imagery used to support human
rights violations monitoring efforts.
3.1 Data
This study used Sentinel-1 SAR imagery from ASF and geospatial data from the UN
Rapid Damage Assessment. These sources are described in Tables 3 and 4. One of the key
benefits of Sentinel-1 satellite SAR imagery data is its accessibility from the ASF. However, its
large file size (65 GB) demanded several hours for data downloading and imagery pre-
processing. UN data is also freely available and easily accessible from UNOSAT and requires
less storage space and download times compared to the Sentinel-1 imagery. While Sentinel-1
SAR images and UNOSAT geolocated point data derived from VHR imagery do not exactly
align, both cover appropriate temporal timescales for fair comparison for this study. Table 3 and
Table 4 highlight key data attributes.

32
Table 3. Sentinel-1 SAR imagery
Date Event Data Type Purpose Spatial
Resolution
Temporal
Resolution
Size on
Disk
Availability
02/16/2022 Pre-Russian invasion GeoTiff Change 10 m 12-day 65 GB Freely available at
02/28/2022 Post-Russian invasion raster detection revisit https://search.asf.alaska.edu
03/12/2022 Conflict rate
03/24/2022 Conflict
04/05/2022 Conflict
04/17/2022 Conflict
04/29/2022 Conflict
05/11/2022 Conflict
05/23/2022 Post-Russian
seizure of Mariupol

Table 4. UNOSAT rapid damage assessment
Data Date Data Type Purpose Spatial
Resolution
Size on Disk Availability
AOI of
Mariupol
residential area
2022 Polygon
vector
Study area
boundary
N/A 23 MB, 13.3
MB
Freely available at
https://unosat.org/products/3300
Rapid Damage
Assessment
Map
06/21/2021,
03/14/2022,
05/7/2022,
05/08/2022,
05/12/2022
Point vector
with damage
scale attribute
information
Validation 30 cm
WorldView-3
imagery, 50 cm
WorldView-2
imagery

33
3.1.1 Sentinel-1 SAR Imagery
ESA’s Sentinel-1 satellite mission is to provide continuous radar mapping of the Earth by
providing enhanced revisit frequency, coverage, and timeliness for Earth science and emergency
response applications (Sentinel n.d.b). ASF provides an accessible graphical interface for
creating imagery searches and downloading remote sensing data from its archive. Imagery is
available within three days of acquisition to any user free of charge, making it suitable for NGOs
and human rights organizations lacking special licenses or funding (ASF n.d.c). An alternative
source, ESA, was also considered. ESA is the owning agency of the Sentinel-1 satellite and
delivers Sentinel data within 24 hours or within one hour of reception for near real-time
emergency monitoring (Sentinel n.d.a). While both ASF and ESA offer accessible imagery, ASF
was chosen as the source for this study because of its online Radiometrically Terrain Corrected
(RTC) imagery conversion tool providing RTC data for download. ASF was selected for its user-
friendly platform and ArcGIS Pro compatible geoprocessing tools, making it a user-friendly data
source for non-imagery experts. 10 m resolution, C-band, ground range detected (GRD) SAR
imagery was downloaded from the ASF Data Search Vertex with the creation of an account.
ASF provides new imagery every 12 days corresponding to the Sentinel-1 satellite’s 12-
day imaging rate (Kristenson 2020). Images between February 16, 2022, and May 23, 2022, are
used for this study. These dates align with pre-Russian invasion and post-seizure of Mariupol.
Figure 10 is an example of RTC SAR imagery.
34

Figure 10. Sentinel-1 SAR image from February 28, 2022
SAR images are not always easily interpreted due to the non-intuitive, side-looking
geometry of the sensor (ASF n.d.a). Surfaces, slope, and man-made structures can affect
backscattering and therefore brightness in an image (ASF n.d.a). These white and black images
at coarse 10 m resolution do not reveal clearly discernable buildings to determine damage and
therefore identify potential violations of human rights. Methods for using this imagery for
damage analysis are explained in the workflow section.
This study considered other potential imagery sources but identified limiting factors for
consideration. Cost and accessibility are the two greatest limitations when choosing an
appropriate SAR data source. At the time of this study, very few civil and commercial companies
capture SAR data, limiting sources for researchers. Many commercial companies require
contracts for access, contributing to the low prevalence of SAR research in the human rights
community. ESA’s TerraSAR-X and TanDEM-X Earth observation SAR imagery are only
35
offered to users located in the territory of ESA member states in the European Commission
Member States and in Africa. These accessibility restrictions make it an unreliable source for
human rights practitioners working worldwide in often unpredictable locations. In the
commercial sector, Airbus imagery resellers quoted an academic price for high-resolution SAR
imagery at $156.25 - $568.75 for 25 sq km of coverage. Furthermore, at the time of this study,
Airbus was restricting sale of SAR data collected over Ukraine during ongoing war for security
purposes.
Optical imagery was also investigated as a potential validation source but was not
selected accessibility restrictions. At the time of this study, Planet and Maxar are two leading
commercial imagery companies offering VHR optical imagery. While both offer some of the
best available imagery to date, Planet’s Education and Research Program for students only offers
free optical imagery up to 3 m resolution, unsuitable for building-level damage assessment.
Maxar only offers select archive imagery from natural disaster events which did not include the
war in Ukraine at the time of this study. Medium-resolution Sentinel-1 SAR imagery does not
have these accessibility restrictions. Rather, it offers the most suitable coverage, accessibility,
cost, and practicality for this study. These characteristics make Sentinel-1 SAR imagery a
suitable source for human rights practitioners.
3.1.2 UNOSAT Rapid Damage Assessment
UN data was used for geospatial analysis and compared to Sentinel-1 SAR damage
analysis. UNOSAT produces damage assessment maps using VHR satellite data of areas affected
by disaster, complex emergencies, and conflict. The UNOSAT shapefile included a study area
polygon of the AOI in the residential area of Mariupol. UN geolocated point data representing
damaged buildings identified from 30 cm optical WorldView-3 and 50 cm optical WorldView-2
36
imagery were used to compare imagery analysis results from 10 m SAR Sentintel-1 imagery.
Figure 11 displays the Mariupol AOI boundary and damaged building points from UNOSAT.

Figure 11. UNOSAT rapid damage assessment
While UN point data are classified into four degrees of damage (destroyed, severe
damage, moderate damage, and medium damage), UNOSAT warns results have not been
validated in the field due to lack of access to an ongoing war zone. Although UNOSAT point
data derived from VHR Worldview imagery has not been validated in the field due to denied
access during ongoing war, it is the best validation source at the time of this study due to the
data’s accessibility, usability, and high fidelity. Due to the coarse spatial resolution of the
Sentinel-1 SAR imagery used in this study, classifying individual building degrees of damage is
beyond the scope of this study and all UN point data used for verification are assumed to indicate
damaged or destroyed buildings.
37
UN point data representing damaged buildings generally align with buildings from
ArcGIS World Imagery base map. Distance rings in Figure 12 show buffers at 5, 10, 20, and 30
meters around each UN damage point to visualize proximity of damaged buildings to other sites.

Figure 12. Distance rings for UN points representing damaged buildings
Distance rings in Figure 12 show that damage points can possibly be associated with
damages detected by SAR imagery in this project. Chapter 4.2.2 includes further details
comparing UNOSAT rapid damage assessment points with SAR damage detection.
3.2 Workflow
Figure 13 provides the SAR imagery analysis and spatial statistics workflow used in this
study. This workflow generated damage assessment raster images every 12 days using SAR
38
imagery pairs between February 16, 2022, and May 23, 2022. Light blue boxes are data from
Table 3 and Table 4, yellow boxes are ArcGIS Pro geoprocessing tools, and green boxes are
products. The Calculate Log Difference tool used in this study is not an ArcGIS Pro native tool.
The ArcGIS Python Toolbox designed by ASF includes the Calculate Log Difference tool
designed for Sentinel-1 RTC SAR datasets necessary to complete this workflow. The resulting
products of this analysis are the damage assessment map clipped to the Mariupol AOI and
corresponding damage statistics.
39

Figure 13. Workflow for Sentienl-1 SAR damage assessment
40
Figure 14 provides the geospatial analysis and accuracy assessment between Sentinel-1
SAR and UNOSAT. This workflow generated a summation of UNOSAT damage points that lie
within SAR damage polygons, providing an accuracy evaluation. Blue boxes include the
Sentinel-1 SAR cumulative damage raster resulting from analysis in Figure 13 and UNOSAT
damage points from Table 4. Yellow boxes are ArcGIS Pro geoprocessing tools and green boxes
are results.
41

Figure 14. Workflow for Sentinel-1 and UNOSAT damage evaluation
3.2.1 Imagery Preparation
One of the most significant challenges of working with SAR data is distortions resulting
from the satellite’s side-looking sensor. A process called radiometric terrain correction (RTC)
addresses these concerns and stabilizes backscatter values to reduce geometric distortions that
may lead to geolocation errors (ASF n.d.d). SAR datasets must be RTC processed to align well
42
with other geospatial data before working in GIS applications for time-series analysis. This study
used the On-Demand RTC Processing tool in the ASF Data Search Vertex portal with the
Copernicus Digital Elevation Model to adjust for distortions. All images were RTC processed in
amplitude scale to accommodate follow-on optimization. The On-Demand RTC Processing tool
replaces the lengthy process of manual image-preprocessing and drastically reduces the overall
workflow completion time. This benefit is crucial for human rights practitioners in need of
timely information.
3.2.2 Calculate Log Difference
A simple and practical way to detect change between two SAR images is the log
difference calculation detailed in Chapter 2 (Aimaiti 2022; Matsuoka and Yamazaki 2004; ASF
2020). This study used the calculation of Log10 (Date2/Date1) to identify significant changes in
backscatter between two images (1). Negative values indicate a decrease in radar backscatter
between two images and positive values indicate an increase in backscatter. Areas with little to
no change in backscatter will indicate no change in value. This study followed the simple and
robust SAR log equation used in a similar study by Aimaiti et al. (2022) based on the hypothesis
that building backscattering characteristics will change after natural or anthropogenic disasters
resulting in building damage (Equation 1).
After downloading the ArcGIS Python Toolbox, the Calculate Log Difference Tool was
used to calculate the log of the ratio of pixel values from two images of the same area taken at
different times. A total of 9 images were used to create 9 backscatter intensity change rasters
indicating damage due to Russian attacks on the AOI in Mariupol.
43
3.2.3 Manual Threshold
The SAR intensity-based change detection method applied in the previous step produced
a backscatter intensity change raster. The study used an amplitude scale for SAR images which is
optimal for calculating log difference ratios (ASF n.d.d). Values in the amplitude scale are the
square root of the power scale values, which brightens darker pixels and darkens brighter values,
reducing the range of the image (ASF n.d.d). Positive values in the difference raster indicate
increased backscatter over time, whereas negative values indicate decreased backscatter over
time (ASF 2020). Since these values are not easily interpreted by non-imagery experts, pixels
were classified to visualize damaged and undamaged areas. A binary classification draws
attention to significant change areas rather than displaying the full spectrum of backscatter
change values (ASF n.d.d). This study adapted the simple histogram thresholding method
outlined in Chapter 2 to achieve a binary classification scheme (ASF 2020; Braun 2018; Kim
2023). Using the method provided by the ASF Log Difference Tool tutorial, values less than and
greater than one standard deviation are used to indicate damage, since both positive and negative
backscatter values can represent change (ASF 2020; van Heyningen 2018). The histogram
statistics of the log difference raster were used to set class break points of one and two standard
deviations for both positive and negative values. For simplicity, this study does not attribute
positive or negative values to certain types of damage. Any change in backscatter, positive or
negative values, was classified as potential damage. The Raster Calculator geoprocessing tool
was used to create a binary classification using pixel values from the difference raster to produce
a map of undamaged (no change in backscatter) and damaged areas (change in backscatter).
Finally, the raster was clipped to the Mariupol AOI to create a damage map.
44
3.2.4 Zonal Statistics as Table
The Zonal Statistics as Table geoprocessing tool was used to calculate mean percentage
of damage. This tool calculated mean values of undamaged area in the AOI represented as a
fraction. Multiplying these values by one hundred produced mean percentage of undamaged
area. Mean percentage of damaged pixels were calculated using Equation 2:
𝑀𝑒𝑎𝑛 %
𝐷𝑎𝑚𝑎𝑔𝑒𝑑 = 100 − 𝑀𝑒𝑎𝑛 %
𝑈𝑛𝑑𝑎𝑚𝑎𝑔𝑒𝑑
(2)

The calculation was repeated for each timestep every 12 days from the February – May 2022
timeframe of Russian attacks. Table 5 summarizes results.
3.2.5 Sentinel-1 SAR Damage and UNOSAT Damage Evaluation
To evaluate SAR damage accuracy, the SAR damage assessment was compared to UN
damage points representing geolocated damaged buildings derived from VHR optical imagery.
To compare SAR damage assessment to UNOSAT damage assessment, SAR damage pixels
were converted to SAR polygons using the Raster to Polygon tool. Summarize Within tool was
used to summate UN points representing damaged buildings that fell within SAR damage
polygons.
3.2.6 Near Tool Analysis
The Near Tool was used to calculate distance and angle information between SAR
damage polygons and UN damage points. Near Distance calculated the number of UN points
within various radius distances (0 m, 5 m, 10 m, 20 m, and 30 m) of SAR damage polygons.
Radius distances were selected based on reasonable damage radius estimates resulting from four
types of weapons identified by the Independent International Commission of Inquiry on
Ukraine’s report: unguided bombs from aircraft, long-range anti-ship missiles, cluster munitions,
45
and multiple launch rocket systems (UN 2023a). Near Angle measured the direction of the line
connecting UN damage points to the nearest SAR damage polygon. The range spans from -180°
to 180°, with 0° to the east, 90° to the north, 180° (or -180°) to the west, and -90° to the south
(ArcGIS Pro n.d.b).

46
Chapter 4 Results
This chapter describes results of SAR imagery analysis and compares damage estimates with
UNOSAT damage assessment derived from VHR optical imagery. The following sections
present results in the form of maps and spatial statistics.
4.1 Sentinel-1 Damage Assessment
Sentinel-1 SAR damage analysis using methods described in Chapter 3 exhibited
increasing damage prevalence with time. The following sections exhibit damage results from
Russian attacks culminating in the seizure of Mariupol in May 2022. The results of this study
revealed that the Sentinel-1 SAR imagery underestimated cumulative damage compared to
damage estimated from UNOSAT VHR optical imagery over the same AOI.
4.1.1 Spatiotemporal Results
A key assumption used in this project is that any significant change in backscatter
intensity, positive or negative, is indicative of war-induced damage, an assumption shared by van
Heyningen in a study on Sentinel-1 damage detection (2018). Figure 15 shows an example of
backscatter intensity change detection between two images using the log of the ratio of the pixel
values over the same AOI (ASF 2020).
47

Figure 15. Intensity change between February 16, 2022 and May 23, 2022
Areas in yellow represent values associated with no backscatter change. Red and blue
represent the greatest changes in intensity (negative and positive values), indicating damages.
These rasters are considered intermediary products since they do not provide easily discernable
or meaningful information to non-imagery experts.
Using methods described in Chapter 3, re-classified rasters resulted in maps identifying
damaged and undamaged areas. Results show detected changed areas as “damaged” in red, and
unchanged areas as “undamaged” in yellow. Old damage detected from each previous 12-day
period is shown in transparent red. SAR damage change results for individual districts of the
AOI are provided beginning with Figure 17 to easily note changes. Figure 16 shows the first
damage assessment using a pre-invasion Sentinel-1 SAR image and a post-invasion image 12-
days later.

48

Figure 16. Damage assessment using SAR for February 16 – 28, 2022
Very little detected SAR damage in the AOI is indicative of low levels of conflict during
the early stages of the war. Change detection over this period includes war-induced damages
inflicted over four days (since Russia’s invasion on February 24, 2022). Low levels of damage
detection suggest Mariupol had not yet experienced heavy attacks.
49

Figure 17. Zhovtnevyi District damage assessment using SAR for February 16 – 28, 2022 (old
damage), and February 28 – March 12, 2022 (new damage)

Figure 17 shows SAR damage changes in the Zhovtnevyi District. Red pixels indicate
SAR damage detected from images between February 28, 2022 – March 12, 2022. Transparent
red pixels indicate old SAR damage detected from images between February 16 – February 28,
2022. Results show an increase in damage extent and magnitude which corresponds with reports
of devastating attacks in Mariupol during this time period, including the deadly attack on a
Mariupol maternity hospital on March 9, 2022 (Cullison 2022; OHCHR 2023a).
50

Figure 18. Mariupol maternity hospital attack imagery
Commercial optical satellite imagery from Maxar shows the site of the Mariupol
maternity hospital attack, approximately located in areas of increased SAR damage in
corresponding analysis (Figure 18). This timeline and damage comparison demonstrate how
SAR damage analysis can corroborate alleged war crimes such as unlawful attacks and also
challenge misinformation attempts to uphold aggressors accountable (Hinnant and Chernov
2022b). When available, complementary VHR optical imagery can be used together with
medium-resolution SAR imagery to detect, monitor, and attribute attacks causing damage to
civilian infrastructure.
51

Figure 19. Livoberezhnyi District damage assessment using SAR for February 16 – 28, 2022 (old
damage) and February 28 – March 12, 2022 (new damage)

Figure 19 shows SAR damage changes in the Livoberezhnyi District. Red pixels indicate
SAR damage detected from images between February 28, 2022 and March 12, 2022. Transparent
red pixels indicate old SAR damage detected from images between February 16 – February 28,
2022. Similar to estimates in the Zhovtnevyi District, results show an increase in damage extent
and magnitude.
52

Figure 20. Zhovtnevyi District damage assessment using SAR for February 28 – March
12, 2022 (old damage) and March 12 – 24, 2022 (new damage)

Increasing levels of new SAR damage were detected in the Zhovtnevyi District for the
period of March 12 – March 24, 2022. Damaged pixels are prevalent throughout and
concentrated in the east. New SAR damage results agree with reports of missile and air strikes in
Mariupol, including the deadly bombing of the Mariupol drama theatre located in the Zhovtnevyi
District on March 16 (Hinnant and Chernov 2022a; OHCHR 2023a). Increased SAR damages
during this period correspond to the attack violating international humanitarian law (Benedek et
al. 2022). Figure 21 and Figure 22 display pre- and post- drama theatre attack VHR optical
images corresponding to SAR damage detected from March 12 – 24, 2022.

53

Figure 21. Mariupol drama theatre imagery before attack
Figure 21 shows how buildings and ground features are easily discernable with VHR
optical imagery compared to medium-resolution SAR imagery. Commercial VHR optical
imagery from CNES/Airbus shows the undamaged Mariupol drama theatre on March 14, two
days before the attack.
54

Figure 22. Mariupol drama theatre imagery after attack
Sixty cm resolution optical imagery from CNES/Airbus on March 16, 2022 shows
Mariupol drama theatre damage, including a destroyed roof and two debris fields to the north and
the south of the building. Figure 22 shows how smoke in the lower and upper left corners of the
image can conceal ground features. This demonstrates how dependence on optical images alone
can be unpredictable and unreliable for monitoring war-induced damages. Since SAR is not
affected by smoke, clouds, weather, or time of day, it provides consistent imaging capability
regardless of imaging conditions, ideal for human rights practitioners (Brown and Hogan 2020).
55

Figure 23. Livoberezhnyi District damage assessment using SAR for February 28 – March 12,
2022 (old damage) and March 12 – 24, 2022 (new damage)

Increasing levels of new SAR damage were detected in the Livoberezhnyi District. Less
damage was detected in the north and south. Damage extent was scattered and distributed
throughout the area.
56

Figure 24. Zhovtnevyi District damage assessment using SAR for March 12 – 24, 2022 (old
damage) and March 24 – April 5, 2022 (new damage)

The third consecutive increase in SAR damage occurred from March 24 – April 5, 2022,
represented in red. The size of new SAR damage pixels appeared larger than old SAR damage
pixels from analysis for March 12 – 24, 2022 represented in transparent red. Higher levels of
SAR damage appeared in the south and east. Increased damages suggest intensifying attacks
which could include severe shelling, airstrikes, and bombing responsible for civilian casualties
(UN OHCHR 2023a).

57

Figure 25. Livoberezhnyi District damage assessment using SAR for March 12 – 24, 2022 (old
damage) and March 24 – April 5, 2022 (new damage)

Figure 25 shows a mix of new and old SAR damage. Red shows new damage detected
from March 24 – April 5, 2022, and transparent red shows old damage detected from March 12 –
24, 2022. Results in the Livoberezhnyi District indicate extensive damages. Damage was
concentrated mostly in the central area and in the southwest.
58

Figure 26. Zhovtnevyi District damage assessment using SAR for March 24 – April 5, 2022 (old
damage) and April 5 – 17, 2022 (new damage)

Detected SAR damage significantly decreased from April 5, 2022 – April 17, 2022, for
the first time since February 28, 2022. Figure 26 shows a majority of old SAR damage
corresponding to significant levels of damage from March 24 – April 5, 2022. Results indicate a
possible decrease in Russian attacks or reduction in overall conflict.
59

Figure 27. Livoberezhnyi District damage assessment using SAR for March 24 – April 5, 2022
(old damage) and April 5 – 17, 2022 (new damage)

Figure 27 shows a majority of old SAR damage corresponding to significant levels of
damage from March 24 – April 5, 2022. Low levels of new SAR damage were detected in the
Livoberezhnyi District for this period.
60

Figure 28. Zhovtnevyi District damage assessment using SAR for April 5 – 17, 2022 (old
damage) and April 17 – 29, 2022 (new damage)

Damage remained low from April 17, 2022 – April 29, 2022. This decrease in detected
damages corresponds with Russia’s announcement of Mariupol’s capture on April 21, 2022
(Bowen 2023). Low levels of dispersed red pixels representing new damage could depict
damages from intermittent fighting or result from a pause in fighting.
61

Figure 29. Livoberezhnyi District damage assessment using SAR for April 5 – 17, 2022 (old
damage) and April 17 – 29, 2022 (new damage)

Damages remained low for the second consecutive period in the Livoberezhnyi District
from April 17 – 29, 2022. No significant new damages suggest low levels of fighting. This
reduced fighting corresponds to Russian claims of Mariupol’s seizure (Bowen 2023).
62

Figure 30. Zhovtnevyi District damage assessment using SAR for April 17 – 29, 2022 (old
damage) and April 29 – May 11, 2022 (new damage)

Despite Russia’s announcement of Mariupol’s capture on April 21, 2022, Ukrainian
forces displayed continuous resistance against Russian forces (Bowen 2023). SAR damage
detection indicates extensive and severe damages from April 29 – May 11, 2023. This
widespread sudden significant increase in new damage suggests continued fighting took place
despite Russia’s claims of capture.
63

Figure 31. Livoberezhnyi District damage assessment using SAR for April 17 – 29, 2022 (old
damage) and April 29 – May 11, 2022 (new damage)

Damage was similar in severity and extent in the Livoberezhnyi District from April 29 –
May 11, 2022. New damages were dispersed throughout the district indicating widespread
attacks. While SAR damage pixels alone cannot attribute attacks to either Russian or Ukrainian
forces, the prevalence during this period suggests increased hostilities during the late stages of
Mariupol’s attacks.
64

Figure 32. Zhovtnevyi District damage assessment using SAR for April 29 – May 11, 2022 (old
damage) and May 11 – 23, 2022 (new damage)

From May 11 – 23, 2022, damages decreased dramatically to low levels similar to those
during early April 2022 (Figure 27 and Figure 28). This striking decrease corresponds with
reports of Ukrainian forces’ surrender of Mariupol in mid-May 2023 after ceasing combat at the
Azovstal iron and steel plant (Bowen 2023).
65

Figure 33. Livoberezhnyi District damage assessment using SAR for April 29 – May 11, 2022
(old damage) and May 11 – 23, 2022 (new damage)

Damages decreased to low levels in the Livoberezhnyi District from May 11 – 23, 2022.
This dramatic decrease in SAR damage agrees with Ukranian forces’ surrender in mid-May
2022. Transparent red pixels correspond with old damage from April 29 – May 11, 2022.

66

Figure 34. Damage assessment using SAR for February 16 – May 23, 2022
Figure 34 shows the cumulative damage reclassified from its backscatter intensity change
raster (Figure 15). Results show extensive damages throughout both districts comprising the
Mariupol AOI. A spatiotemporal trend is a clear consecutive increase in damage from February
28, 2022 – March 12, 2022, March 12, 2022 – March 24, 2022, March 24, 2022 – April 5, 2022,
and April 29, 2022 – May 11, 2022. Damage changes remained low from April 5, 2022 – April
17, 2022, and April 17, 2022 – April 29, 2022. The most severe damage in extent and magnitude
occurred from April 29, 2022 – May 11, 2022. Finally, damages appeared low from May 11,
2022 – May 23, 2022, during the final stages of Russia’s capture of the city of Mariupol.
Spatiotemporal changes in damage extent and magnitude correspond to statistical changes
highlighted in Table 5.
67
4.1.2 Damage Statistics
Damage statistics align with spatiotemporal trends of SAR imagery damage assessment.
Mean damage percentage values indicated increasing degrees of damage with Russia’s
progressing war. These results demonstrate the potential to corroborate open-source reports of
attacks and subsequent violations of human rights and international humanitarian law described
in Chapter 1.
The cumulative damage assessment from February 16, 2022 (pre-invasion) and May 23,
2022 (post-Russian seizure of Mariupol) estimates 27% of mean damage. Gradual increases in
mean percentage of damage occurred between February 28, 2022 – March 12, 2022 (1.25%),
March 12, 2022 – March 24, 2022 (2.5%), and March 24, 2022 – April 5, 2022 (3.5%). The
period from April 5, 2022 – April 17, 2022, and April 17, 2022 – April 29, 2022 showed low
mean percentages of damage. The greatest mean percentage damage of 11.2% occurred from
April 29, 2022 – May 11, 2022. This sudden increase in damage detection indicates a significant
surge in violence and active attacks corresponding to Russia’s capture of Mariupol. The sudden
decrease in detected damage over the next 12-day period from May 11, 2022 – May 23, 2022,
indicates little to no conflict-induced damage, signifying the end of conflict due to Russia’s
seizure of Mariupol. Table 5 summarizes mean statistics.

68
Table 5. Mean percentage statistics of undamaged and damaged buildings
Dates (pre-, post-) Mean % Undamaged Mean % Damaged
February 16, 2022
February 28, 2022
99.75% 0.25%
February 28, 2022
March 12, 2022
98.75% 1.25%
March 12, 2022
March 24, 2022
97.5% 2.5%
March 24, 2022
April 5, 2022
96.5% 3.5%
April 5, 2022
April 17, 2022
99.6% 0.4%
April 17, 2022
April 29, 2022
99.35% 0.65%
April 29, 2022
May 11, 2022
88.8% 11.2%
May 11, 2022
May 23, 2022
99.6% 0.4%
February 16, 2022
May 23, 2022

73% 27%
4.2 UNOSAT Comparison
Sentinel-1 SAR damage results were compared to the UNOSAT damage assessment to
assess accuracy. The overall damage statistics derived from Sentinel-1 SAR imagery
underestimated damage compared to the rapid damage assessment of UNOSAT. UNOSAT’s
manual building inspection using 30 cm WorldView-3 imagery and 50 cm WorldView-2
imagery resulted in 32% estimated total damage. This is 5% greater than the 27% mean
percentage of damage estimated from Sentinel-1 SAR analysis. Direct comparison of individual
damaged buildings represented by UNOSAT geolocated point data was not possible with
Sentinel-1 imagery due to the satellite sensor’s 10 m resolution constraints.
69
4.2.1 Summarize Within
Sentinel-1 SAR damage results were compared with UNOSAT’s damage assessment to
assess overall accuracy and feasibility of using SAR imagery to detect infrastructure damage.
The Summarize Within tool resulted in 37% of all UN damage points within SAR damage
polygons. Despite the low accuracy estimate, the magnitude and extent of damages are
comparable between SAR estimates and UNOSAT’s assessment. While SAR damage
underestimated total damage, the spatial distribution of damages was consistent with the
distribution of UNOSAT damage buildings. Figure 39 shows an example of UN damage points
within SAR damage polygons. Limitations of this estimate are discussed in Chapter 5.
4.2.2 Near Distance
Near Tool analysis indicates that the percentage of total UN damage points within SAR
damage polygons increases with an increasing distance radius. While 37% of all UN damage
points occur within SAR damage polygons (0 m radius), 94% of all UN damage points occur
within SAR damage polygons within a 30 meter radius. Results are summarized in Table 6.
Table 6. Near distance of UN damage points to SAR damage polygons
Near Distance (Meters)
of UN Points to SAR
Damage Polygon
Percentage of Total UN
Damage Points Within
SAR Damage Polygon
0 37%
5 54%
10 67%
20 85%
30 94%

Results suggest SAR imagery can offer damage detection from true locations within a
reasonable distance. Since the UN OHCHR reports most of civilian casualties were caused by
heavy artillery shelling, rockets, and missile and air strikes, this project assumes a reasonable
70
damage radius from such weapons up to 30 meters (UN OHCHR 2023a). Figure 35 displays
results from Near Distance analysis.

Figure 35. Distribution of near distance
79% of UN damage points were located within 10.8 meters of the nearest SAR polygon.
Results suggest UN damage points are reasonably distanced from corresponding damage pixels
from SAR analysis. These results complement Table 6.
4.2.3 Near Angle
Near angle results of UN damage points to SAR damage polygons appear randomly
distributed throughout the AOI, suggesting no correlation between near angles of UN damage
points to SAR polygons and location (Figure 36).

71

Figure 36. Near angle of UN damage points to SAR polygons
Most UN damage points have a near angle to a SAR damage polygon between -90 to 0
degrees, in the southeast direction (Figure 37).

Figure 37. Distribution of near angle
72
One possible explanation for this distribution is the incidence angle of the acquired
imagery, which is the angle between the radar sensor and a line perpendicular to the surface
(Marghany 2020). Another factor could be building orientation which could affect the
backscatter intensity.

73
Chapter 5 Discussion
The goal of this study was to assess the effectiveness of using medium-resolution publicly
available SAR imagery in place of expensive high-resolution optical imagery for human rights
violations monitoring purposes. Chapter 5 addresses the limitations and challenges of this project
and provides suggestions for future research.
5.1 Limitations and Challenges
Limitations of this project affect assessment accuracy and present implications for human
rights researchers. The following section describes limitations that create challenges for
conducting analysis using Sentinel-1 and UNOSAT data.
5.1.1 Sentinel-1
Coarse spatial resolution is a significant limitation of the Sentinel-1 SAR imagery used in
this study. Since 10 m resolution cannot identify individual building damages, determination of
degree of damage (possible, moderate, severe, destroyed) is not possible with medium-resolution
SAR imagery. This may be a limiting factor for human rights researchers aiming to identify
specific degree of damage. Coarse resolution may affect accuracy of assessment since multiple
pixels can represent damage spanning multiple buildings. Figure 38 shows an example of SAR
damage pixels covering partial or multiple buildings for a sample area in Mariupol adjacent to
the Mariupol drama theatre. Damage pixels result from analysis using images on February 16,
2022 and May 23, 2022.
74

Figure 38. SAR damage pixels overlaying ground features in Zhovtnevyi District
SAR pixels in Figure 38 illustrate the challenge of attributing SAR damage to individual
buildings. Moreover, SAR damage pixels do not exclusively represent building damage. Pixels
may represent any disturbance to ground surfaces, including roads, fields, and other
infrastructure. Human rights researchers using medium-resolution SAR imagery must be aware
of this limitation and subsequent challenges.
The Raster to Polygon method presents accuracy challenges affecting False Positive and
False Negative conclusions. A False Positive is defined as a SAR damage polygon that does not
contain a UNOSAT damage point, and a False Negative is defined as undamaged SAR area that
contains a UNOSAT damage point (Qian et al. 2020). When investigating individual events such
as the devastating attack on the Mariupol drama theatre (Hinnant and Chernov 2022a),
researchers must be aware that damage to a single point of interest may be represented by
multiple damage polygons rather than a single point. Since a SAR damage polygon may cover
75
multiple buildings or represent multiple types of damages, researchers may be challenged with
determining how to attribute damages to single locations of interest. Figure 39 demonstrates this
challenge with the Mariupol drama theatre. Damage pixels result from analysis using images on
March 12,2022 and March 24, 2022.

Figure 39. SAR damage polygons and UNOSAT damage points over Mariupol drama theatre
The infamous attack on the drama theatre is an example of an indiscriminate and
disproportionate attack which violates international humanitarian law (UN 2023a). Several
separate SAR damage pixels cover the drama theatre in Figure 39. However, the UNOSAT
damage point representing the drama theatre occurs near, but not within, a SAR damage
polygon. A possible explanation is that debris or rubble resulting from the theatre attack
dispersed radar signals and affected backscatter intensity, resulting in multiple detected damage
pixels over the same building and vicinity (van Heyningen 2018). Although the SAR damage
polygons do not contain UNOSAT damage points, suggesting multiple False Positives,
76
UNOSAT data included only one damage point representing associated drama theatre damage.
Researchers must be aware of the Raster to Polygon tool limitation when evaluating results and
comparing with other data derived from VHR optical imagery.
Temporal resolution is another limitation of this study. While 12 days may be sufficient
for monitoring general activity, it may not be sufficient for international criminal courts requiring
assurance with high quality evidence. Frequent revisit rates within a 24-hour period, or within 1-
2 days may be necessary depending on the application.
Despite these limitations and challenges, coarse resolution may offer sufficient
information for researchers aiming to track trends or identify potential violation areas and
prompt investigation using higher fidelity satellite imagery. Coarse results are sufficient for
human rights researchers seeking timely results during active conflict. Thus, medium-resolution
SAR imagery may still be desirable for some human rights practitioners due to accessibility,
sufficient accuracy, low cost, and simplicity. Satellite imagery alone is not sufficient to prove
war crimes or crimes against humanity in international tribunals (Hazan 2017; Kroker 2015).
However, results of this project encourage continued research to advance SAR imagery that will
compliment non-imagery evidence presented in international criminal courts.
5.1.2 UNOSAT
Limitations of UNOSAT data used in this study include lack of verification, potential
geolocation errors, and different imagery dates. While this project uses UNOSAT’s data to
validate the SAR damage assessment, UNOSAT damage assessment has not been verified in the
field due to unsafe conflict zones making it impossible for in-person confirmation of
infrastructure damage. A key assumption of this project is that UNOSAT’s data is accurate based
on increased spatial resolution of optical images used.
77
Geolocation errors may potentially affect analysis in this research. UNOSAT provided
geolocated point data derived from VHR optical images but did not specify geolocation
methodology. Depending on the base map used, UNOSAT data might not always exactly match
buildings or SAR damage polygons in this analysis. Even minor errors could result in unmatched
UNOSAT damage points within SAR damage polygons. Figure 40 illustrates position of
UNOSAT damage points in relation to SAR damage pixels.

Figure 40. UN damage point representing Mariupol Drama Theatre
Using multiple distances for the near tool analysis described in Chapter 3 addresses
possible geolocation errors. For example, an expanded radius of 30 meters results in 94% of
UNOSAT damage points within SAR damage pixels (polygons). Since manual inspection and
78
judgement of matching UNOSAT data points with SAR damage is impractical over large areas,
the near tool offers practical analysis with geolocation errors in mind.
While the majority of UNOSAT damage points classify building damage, data also
includes non-building damage, such as roads and fields. Less than 1% of UNOSAT building
point data includes non-building damage (17 out of 5,5660 points). This study did not filter out
other infrastructure since SAR damage pixels also include non-building damage. Future analysis
could consider including this distinction in analysis.
Another limitation is a non-exact comparison of different imagery dates from Sentinel-1
and UNOSAT’s WorldView-3 and WorldView-2 imagery. UNOSAT’s assessment used pre-
conflict imagery from June 21, 2021, whereas the first pre-conflict SAR image used for this
project was from February 16, 2022. The last image used for UNOSAT’s analysis was from May
12, 2022, whereas the last SAR post-conflict image used was from May 23, 2022. Although
UNOSAT and SAR analysis used different images acquired on different dates, these images are
considered adequate for this study.
5.2 Future Research
Future research can improve accuracy of SAR damage assessments and refine remote
sensing analysis methods for human rights practitioners. Two recommended research areas are
backscatter intensity threshold selection and imagery analysis tools.
5.2.1 Backscatter Intensity Threshold Selection
One area of future research to improve accuracy of SAR damage assessment is an
evaluation of True Positive, False Positive, True Negative, and False Negative rates. Comparing
SAR damage with UNOSAT damage results in these values. In another study looking at Russia’s
war-induced damage conducted by Aimaiti et al. (2020), researchers calculated the precision,
79
recall, and F1 score values under different threshold values for a backscatter intensity change
image. Precision is the true positive divided by all that was classified as positive; recall is the
true positive divided by all actual positives, and F1 score is the harmonic average of precision
and recall (Aimaiti et al. 2020). A method incorporating this approach could refine the approach
used in this project and potentially refine damage detection accuracy.
5.2.2 Alternative Analysis Tools
This study used the ASF Vertex portal and ArcGIS Pro as the primary imagery data
source and analysis tool. Another imagery data source option is ESA’s Copernicus Open Access
Hub, which offers freely accessible Sentinel-1 SAR imagery with the creation of an account.
ESA also offers freely available SNAP software. Aimaiti et. al. (2020) successfully conducted
change detection using SNAP software as an alternative to ArcGIS Pro. Data and software
access are key limiting factors for human rights practitioners and the specific restraints of
researchers should be considered when selecting data sources and software applications.
Future projects could also reframe data incorporated from multiple disciplines to enhance
MARS research. Reframing data from a human security perspective could offer additional
insights by fusing imagery with geospatial data, social media information, and witness
testimony. Multi-source analysis is critical for advancing the MARS field and developing new
methods for human rights researchers.
5.3 Conclusion
This project achieved its objective by assessing the feasibility of using medium-
resolution SAR imagery in place of expensive VHR optical imagery to detect war-induced
infrastructure damage. Mean damage percentage was assessed every 12 days from the beginning
of Russia’s invasion to the seizure of Mariupol and results indicated a general increase in
80
damage as the conflict continued. Results indicate that time of day, weather, and cloud cover did
not affect imagery used in this workflow and highlights the practical benefit for human rights
practitioners. A cumulative SAR damage assessment underestimated total damage compared to a
UNOSAT VHR optical imagery damage assessment. SAR analysis estimated 27% cumulative
damage while UNOSAT estimated 32% damage. Analysis of total UN damage points occurring
in SAR damage polygons showed low accuracy, but improved significantly with an expanded
damage radius of 30 m. Furthermore, damage results were consistent with open-source reports of
attacks and violations, including those from media, independent researchers, and the UN. These
coherent results from multiple sources reveal how the human rights community can use SAR
imagery to uncover violations and promote accountability.
Despite the lower fidelity SAR estimates, damages were spatiotemporally consistent with
UNOSAT data. Further inspection also shows challenges with comparing different spatial
resolutions as data points do not always align with SAR damage pixels representative of the
same damage. Future studies can investigate methods to improve backscatter intensity threshold
and damage classification. While medium-resolution SAR imagery underestimates total damage,
it may still offer critical insights to human rights practitioners who must set aside precision in
favor of timely assessment for wartime conditions.

81
References
Aimaiti, Y., Sanon, C., Koch, M., Baise, L.G., and Moaveni, B. 2022. “War Related Building
Damage Assessment in Kyiv, Ukraine, Using Sentinel-1 Radar and Sentinel-2 Optical
Images.” Remote Sensing 14, no. 24 (2022): 1-21. https://doi.org/10.3390/rs14246239
Alaska Satellite Facility. 2020. “Log Difference Tool.” Last modified November 17, 2020.
https://storymaps.arcgis.com/stories/743d1126594b46e0b43be36473dba0e3
— n.d.a. “How do I interpret SAR images? Accessed June 24, 2023.
https://asf.alaska.edu/information/sar-information/how-do-i-read-sar-images/
— n.d.b. “Introduction to SAR.” Accessed March 15, 2023. https://hyp3-
docs.asf.alaska.edu/guides/introduction_to_sar/
— n.d.c. “Sentinel-1 – Data and Imagery.” Accessed April 1, 2023. https://asf.alaska.edu/data-
sets/sar-data-sets/sentinel-1/sentinel-1-data-and-imagery/
— n.d.d. “Sentinel-1 RTC Product Guide.” Accessed April 1, 2023. https://hyp3-
docs.asf.alaska.edu/guides/rtc_product_guide/
— n.d.e. “What is SAR?” Accessed June 14, 2023. https://asf.alaska.edu/information/sar-
information/what-is-sar/
Amnesty International. 2004. “Sudan: At the Mercy of Killers – Destruction of Villages in
Darfur.” Last modified July 1, 2004.
https://www.amnesty.org/en/documents/afr54/072/2004/en/
— 2016. “Burundi: Suspected Mass Graves of Victims of 11 December Violence.” Last
modified January 29, 2016. https://www.amnesty.org/en/documents/afr16/3337/2016/en/
— 2022. “Ukraine and Russia: Amnesty International’s Key Recommendations for EU Leaders.
Reference: TIGO IOR 10/2022.2787.” Last modified March 22, 2022.
https://www.amnesty.org/en/wp-
content/uploads/2022/03/TIGO_IOR_10_2022_2787_Ukraine-and-Russia_Amnesty-
Internationals-Key-Recommendations-for-EU-leaders_22032022.pdf
Anson, P., Cummings, D. 1991. “The First Space War: The Contribution of Satellites to the Gulf
War.” RUSI Journal 136, no. 4 (1991): 45–53.
https://doi.org/10.1080/03071849108445553
ArcGIS Pro. n.d.a. “Introduction to SAR.” Accessed March 28, 2023.
https://pro.arcgis.com/en/pro-app/latest/help/analysis/image-analyst/introduction-to-
synthetic-aperture-radar.htm
— n.d.b. “Near (Analysis).” Accessed May 22, 2023. https://pro.arcgis.com/en/pro-
app/latest/tool-reference/analysis/near.htm
82
Benedek, W., Bilkova, V., and Sassoli M. 2022. “Report on Violations of International
Humanitarian and Human Rights Law, War Crimes, and Crimes Against Humanity
Committed in Ukraine Since 24 February 2022.” Last modified April 13, 2022.
https://www.osce.org/files/f/documents/f/a/515868.pdf
Bowen, A.S. 2023. CRS Report R47068, “Russia’s War in Ukraine: Military and Intelligence
Aspects.” Last modified February 13, 2023.
https://crsreports.congress.gov/product/pdf/R/R47068
Braun, A. 2018. “Assessment of Building Damage in Raqqa during the Syrian Civil War Using
Time-Series of Radar Satellite Imagery.” Journal for Geographic Information Science 6,
no. 1 (2018): 228–42. https://doi.org/10.1553/giscience2018_01_s228
Brown, J., and Hogan, D. 2020. “SAR 101: An Introduction to Synthetic Aperture Radar.”
Accessed April 25, 2023. https://www.capellaspace.com/sar-101-an-introduction-to-
synthetic-aperture-radar/
Chini, M., Pierdicca, N., and Emery, W. J. 2009. "Exploiting SAR and VHR Optical Images to
Quantify Damage Caused by the 2003 Bam Earthquake.” Institute of Electrical and
Electronics Engineers Geoscience and Remote Sensing 47, no. 1 (2009):145-152.
https://doi.org/10.1109/TGRS.2008.2002695
Chini, M., Piscini, A., Cinti, F.R., Amici, S., Nappi, R., and de Martini, P.M. 2013. “The 2011
Tohoku (Japan) Tsunami Inundation and Liquefaction Investigated Through Optical,
Thermal, and SAR data.” Institute of Electrical and Electronics Engineers Geoscience
and Remote Sensing 10, no. 2 (2013): 347–351. https://doi.org/
10.1109/LGRS.2012.2205661
Congressional Research Service. 2021. “Ukraine: Background, Conflict with Russia, and U.S.
Policy.” Last modified October 5, 2021.
https://crsreports.congress.gov/product/pdf/R/R45008
Copernicus Sentinel data 2022 and 2023. Retrieved from ASF DAAC 11 February 2023,
processed by ESA.
Cullison, A. 2022. “Russian Airstrikes Hits Maternity Hospital in Ukrainian City of Mariupol.”
Accessed January 29, 2023. https://www.wsj.com/articles/russia-presses-offensive-as-
ukrainians-try-to-evacuate-11646819525
Datta, A. 2022. “GEOINT, OSINT Comes off Age for Near Real Time Coverage of Ukraine
Conflict.” Last modified March 7, 2022. https://www.geospatialworld.net/blogs/geoint-
osint-comes-off-age-of-ukraine-conflict/
Dong, L. and Shan, J. 2013. “A Comprehensive Review of Earthquake-Induced Building
Damage Detection with Remote Sensing Techniques.” International Journal of
Photogrammetry and Remote Sensing 84, no. 1 (2013): 85–99.
https://doi.org/10.1016/j.isprsjprs.2013.06.011
83
DW. 2022. “Mariupol’s Strategic Importance Between Donetsk and Crimea.” Accessed on June
24, 2023. https://www.dw.com/en/what-do-we-know-about-the-situation-in-mariupol/a-
61125638
Gardner, F. “Mariupol: Why Mariupol is So Important to Russia’s Plan.” Last modified on
March 21, 2022. https://www.bbc.com/news/world-europe-60825226
Ge, P., Gokon, H., and Meguro, K. 2020. “A Review on Synthetic Aperture Radar-Based
Building Damage Assessment in Disasters.” Remote Sensing of Environment 240, no. 1
(2020): 1-19. https://doi.org/10.1016/j.rse.2020.111693
Ghaedi, Monir. 2022. “What is known about the situation in Mariupol?” Last modified March
15, 2022. https://www.dw.com/en/what-do-we-know-about-the-situation-in-mariupol/a-
61125638
Gokon, H., Koshimura, S., Meguro, K. 2017. “Verification of a Method for Estimating Building
Damage in Extensive Tsunami Affected Areas Using L-Band SAR Data.” Journal of
Disaster Research 12, no. 2 (2017): 251-258. https://doi.org/10.20965/jdr.2017.p0251
Hasian M. 2016. Forensic Rhetorics and Satellite Surveillance: The Visualization of War Crimes
and Human Rights Violations. Lanham: Lexington Books.
Hazan, P. 2017. “Can Satellite Imagery Still Prove War Crimes?” Last modified on March 31,
2017. https://www.justiceinfo.net/en/32694-can-satellite-imagery-still-prove-war-
crimes.html#:~:text=At%20the%20moment%2C%20international%20jurisdictions,eleme
nt%20to%20corroborate%20the%20accusations.
Herscher A. “Surveillant Witnessing: Satellite Imagery and the Visual Politics of Human
Rights.” Public Culture 26, no. 3 (2014):469-500. https://doi.org/10.1215/08992363-
2683639
Hinnant, L., and Chernov, M. 2022a. “AP Evidence Points to 600 Dead in Mariupol Theater
Airstrike.” Last modified May 4, 2022. https://apnews.com/article/Russia-ukraine-war-
mariupol-theater-c321a196fbd568899841b506afcac7a1
— 2022b. “Doctors, Crater Disprove Russia’s Hospital Airstrike Misinfo.” Last modified April
8, 2022. https://apnews.com/article/russia-ukraine-europe-
25204be08db8868e858f624f851d2ae3
ICRC. 2010. “IHL and Human Rights Law.” Last modified October 29, 2010.
https://www.icrc.org/en/document/ihl-human-rights-law
Inton, C. 2014. “Ukraine’s Ethnic Divide.” Last modified February 28, 2014.
https://www.cbc.ca/news2/interactives/ukraine-dashboard/
Kim, M., Park, S., and Lee, S. 2023. "Detection of Damaged Buildings Using Temporal SAR
Data with Different Observation Modes." Remote Sensing 15, no. 2 (2023): 1-15.
https://doi.org/10.3390/rs15020308
84
Korkmaz, K. A., and Abualkibash, M. 2018. “Earthquake Damage Detection Using Before and
After Earthquake Satellite Images.” 2018 IEEE International Conference on
Electro/Information Technology (2018): 0615–19.
https://doi.org/10.1109/EIT.2018.8500225
Kristenson, H. 2020. “RTC on Demand.” Last accessed April 1, 2023.
https://storymaps.arcgis.com/stories/2ead3222d2294d1fae1d11d3f98d7c35
Kroker. 2015. “Satellite Imagery as Evidence for International Crimes.” Accessed May 27, 2023.
https://www.coalitionfortheicc.org/news/20150423/satellite-imagery-evidence-
international-crimes
Lee, M., Perl, R., and Woehrel, S. 1998. CRS 96-404, “Bosnia War Crimes: The International
Criminal Tribunal for the Former Yugoslavia and U.S. Policy.”
https://sgp.fas.org/crs/row/96-404.pdf
Levi, P., and Molnar, G. 2022. “How the Energy Crisis is Exacerbating the Food Crisis.”
International Energy Agency.” Last modified June 14, 2022.
https://www.iea.org/commentaries/how-the-energy-crisis-is-exacerbating-the-food-crisis
Lister, T., Pennington, J., and Mcgee, L. 2022. “Civilians Killed as Russian Military Strike Hits
Evacuation Route in Kyiv Suburb.” Last modified March 7, 2022.
https://www.cnn.com/2022/03/06/europe/ukraine-russia-invasion-sunday-intl-
hnk/index.html
Margesson, R., and Mix, D.E. 2022. CRS Insight IN11882, “Humanitarian and Refugee Crisis in
Ukraine.” Last modified November 23, 2022.
https://crsreports.congress.gov/product/pdf/IN/IN11882
Marghany, M. 2020. “Principle Theories of Synthetic Aperture Radar.” In Synthetic Aperture
Radar Imaging Mechanism for Oil Spills, edited by Maged Marghany, 127–50. Houston:
Gulf Professional Publishing.
Masters, J. 2022. “Ukraine: Conflict at the Crossroads of Europe and Russia.” Last modified
October 11, 2022. https://www.cfr.org/backgrounder/ukraine-conflict-crossroads-europe-
and-russia
Matsuoka, M., and Nobuoto, N. 2010. "Building Damage Estimation by Integration of Seismic
Intensity Information and Satellite L-band SAR Imagery" Remote Sensing 2, no. 9
(2010): 2111-2126. https://doi.org/10.3390/rs2092111
Matsuoka, M., and Yamazaki, F. 1999. “Characteristics of Satellite Images of Damaged Areas
Due to the 1995 Kobe Earthquake.” Proceedings of 2nd Conference on the Applications
of Remote Sensing and GIS for Disaster Management.
https://www.researchgate.net/profile/F-
Yamazaki/publication/255597603_CHARACTERISTICS_OF_SATELLITE_IMAGES_
OF_DAMAGED_AREAS_DUE_TO_THE_1995_KOBE_EARTHQUAKE/links/0c9605
85
2ba212e9480c000000/CHARACTERISTICS-OF-SATELLITE-IMAGES-OF-
DAMAGED-AREAS-DUE-TO-THE-1995-KOBE-EARTHQUAKE.pdf.
— 2004. “Characteristics of Satellite SAR Images in the Areas Damaged by Earthquakes.”
Earthquake Spectra 20, no. 3 (2004): 975-994. https://doi.org/10.1193/1.1774182
Mulligan, S.P. 2022. CRS Legal Sidebar LSB10710, “The Law of War and the Russian Invasion
of Ukraine.” Last modified March 16, 2022.
https://crsreports.congress.gov/product/pdf/LSB/LSB10710
— 2023. CRS Legal Sidebar LSB10704, “The Role of International Tribunals in the Response to
the Invasion of Ukraine.” Last modified March 22, 2023.
“https://crsreports.congress.gov/product/pdf/LSB/LSB10704/6
NASA. n.d.a. “What is Remote Sensing?” Accessed May 30, 2023.
https://www.earthdata.nasa.gov/learn/backgrounders/remote-sensing
— n.d.b. “What is Synthetic Aperture Radar?” Accessed April 24, 2023.
https://www.earthdata.nasa.gov/learn/backgrounders/what-is-sar
Parker, C., Karklis, L., and Dormido, H. 2022. “Why Mariupol Matters to Russia in Three
Maps.” Last modified April 21, 2022.
https://www.washingtonpost.com/world/2022/04/21/ukraine-russia-mariupol-fall-maps/
Prins, E. 2008. “Use of Low Cost Landsat ETM+ to Spot Burnt Villages in Darfur,
Sudan.” International Journal of Remote Sensing 29, no. 4 (2008): 1207-
1214. https://doi.org/10.1080/01431160701730110
Qian S.S., Refsnider J.M, Moore J.A, Kramer G.R, Streby H.M. 2020. “All Tests are Imperfect:
Accounting for False Positives and False Negatives Using Bayesian Statistics.” Heliyon
6, no. 3 (2020): 1-6. https://doi.org/10.1016/j.heliyon.2020.e03571
RAND. 2022. “Russia’s War in Ukraine: Insights from Rand.” Accessed June 27, 2022.
https://www.rand.org/latest/russia-ukraine.html
Rohde, D. 1995. “Evidence Indicates Bosnia Massacre.” Accessed May 27, 2023.
https://www.csmonitor.com/1995/0818/18012.html
Romaniello, V., Piscini, A., Bignami, C., Anniballe, R., and Stramondo, S. 2017. “Earthquake
Damage Mapping by Using Remotely Sensed Data: The Haiti Case Study.” Journal of
Applied Remote Sensing 11, no. 1 (2017): 1-16. https://doi.org/10.1117/1.JRS.11.016042
Rotberg, R. 2010. Mass Atrocity Crimes: Preventing Future Outrages. Washington D.C.:
Brookings Institution Press.
Sentinel. n.d.a. “Data Products.” Accessed April 1, 2023.
https://sentinels.copernicus.eu/web/sentinel/missions/sentinel-1/data-products
86
— n.d.b. “Mission Objectives.” Accessed June 22, 2023.
https://sentinel.esa.int/web/sentinel/missions/sentinel-1/mission-objectives
Torkington, S. 2023. “1 year On – The Global Impact of the War in Ukraine.” Last modified
March 2, 2023. https://www.weforum.org/agenda/2023/03/ukraine-crisis-food-energy-
poverty/
UNHCR. 2022. “Ukraine Flash Appeal (March to December 2022).” Accessed February 3, 2023.
https://reliefweb.int/report/ukraine/ukraine-flash-appeal-march-december-2022
United Nations International Criminal Tribunal for the Former Yugoslavia. n.d. “Investigations.”
Accessed on May 27, 2023. https://www.icty.org/en/content/investigations-0
UNITAR. 2022. “Mariupol Updated Building Damage Assessment Overview Map –
Livoberezhnyi and Zhovtnevyi Districts.” Product ID: 3561.
https://www.unitar.org/maps/map/3561
UN OHCHR. 2001. Training Manual on Human Rights Monitoring. New York: United Nations
— n.d. “Universal Declaration of Human Rights.” Accessed on June 24, 2023.
https://www.un.org/en/about-us/universal-declaration-of-human-rights
— 2023a. “Report of the Independent International Commission of Inquiry on Ukraine.”
https://www.ohchr.org/sites/default/files/documents/hrbodies/hrcouncil/coiukraine/A_HR
C_52_62_AUV_EN.pdf
— 2023b. “Ukraine: Civilian Casualty Update 23 January 2023.” Accessed January 29, 2023.
https://www.ohchr.org/en/news/2023/01/ukraine-civilian-casualty-update-23-january-
2023
UNOSAT. 2022. “Mariupol Updated Building Damage Assessment Overview Map –
Livoberezhnyi and Zhovtnevyi Districts.” Accessed March 28, 2023.
https://unosat.org/products/3300
UN SDG. 2022. “Global Impact of War in Ukraine on Food, Energy, Finance systems – BRIEF
NO.2.” Last modified June 8. https://unsdg.un.org/resources/global-impact-war-ukraine-
food-energy-and-finance-systems-brief-no2.
U.S. State Department. 2022. “Russia’s War on Ukraine: Six Months of Lies, Implemented.
Accessed January 29, 2023. https://www.state.gov/disarming-disinformation/russias-war-
on-ukraine-six-months-of-lies-implemented/
van Heyningen, J. 2018. “Rapid Building Damage Detection Through SAR Timeseries Analysis
in the Google Earth Engine: Using Sentinel-1 GRD Imagery in the Google Earth Engine
to Detect Building Damage in Rapid Disaster Response Situations.” M.S. diss., Delft
University of Technology. https://repository.tudelft.nl/islandora/object/uuid:9bd94479-
fbdc-4796-97e2-899fe181277e.
87
Vohra, A. 2022. “Is Ukraine’s Endgame a Russian Land Bridge?” Last modified April 20, 2022.
https://foreignpolicy.com/2022/04/20/is-ukraines-endgame-a-russian-land-bridge/
Washington Post. 2022. “Four Maps that Explain the Russia-Ukraine conflict.” Last modified
February 26, 2022. https://www.washingtonpost.com/world/2022/01/21/ukraine-russia-
explain-maps/
Welt, C. 2021. CRS Report R45008, “Ukraine: Background, Conflict with Russia, and U.S.
Policy.” Last modified October 5, 2021.
https://crsreports.congress.gov/product/pdf/R/R45008/14
Witmer, F. D. W. 2015. “Remote Sensing of Violent Conflict: Eyes from above.” International
Journal of Remote Sensing 36, no. 9 (2015): 2326–52.
https://doi.org/10.1080/01431161.2015.1035412
Yekelchyk, S. 2015. The Conflict in Ukraine: What Everyone Needs to Know. New York:
Oxford University Press, Incorporated.

Zhai, W., and Huang, C. 2016. “Fast Building Damage Mapping Using a Single Post-Earthquake
PolSAR Image: A Case Study of the 2010 Yushu Earthquake.” Earth, Planet, and
Space 68, no. 86 (2016): 1-12. https://doi.org/10.1186/s40623-016-0469-2

Abstract (if available)

Abstract Russia’s unprovoked attack on Ukraine on February 24, 2022, sparked the largest armed conflict in Europe since World War II. As war in Ukraine continues, widespread reports of violations of human rights and international humanitarian law accompany extensive civilian casualties. Satellite imagery has provided unprecedented awareness of Russia’s war to corroborate testimonial evidence of human rights violations. While the use of satellite imagery is now commonplace to aid such efforts, human rights groups need improved remote sensing methods in active war zones. The objective of this study is to evaluate the suitability of freely accessible medium-resolution synthetic aperture radar (SAR) imagery from the European Space Agency’s (ESA) Sentinel-1 satellite versus expensive very high-resolution (VHR) optical imagery for the purpose of detecting war-induced building damage. The study area is the Ukrainian city of Mariupol, which was seized by Russia in May 2022. The study assesses building damage using backscatter intensity changes between images over time. Detected damage in conjunction with reports of civilian casualties may indicate potential violations of international humanitarian law. This study’s results indicate cumulative building damage in both extent and magnitude comparable to a United Nations damage assessment that relied on VHR optical imagery. Statistics estimate 27% damage from February 2022 to May 2022, which is lower than the 32% damage estimate by the UN for the same study area. While SAR imagery may provide less accurate results compared to VHR optical imagery, the increased timeliness, accessibility, and adaptability it offers may render SAR imagery analysis as a more feasible option for some human rights practitioners.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Classifying ash cloud attributes of Eyjafjallajökull volcano, Iceland, using satellite remote sensing

PDF

From ruins to pixels: using remote sensing and GIS to analyze, document, and visualize archaeological sites and ancient roadways in Chaco Canyon, New Mexico

PDF

A comparison of pixel-based, object-based, deep learning, and data fusion image classification approaches in delineating urban tree canopy in City Terrace

PDF

A scenario based fire susceptibility approach for remote sensing platform comparison: Los Angeles county area, southern California, USA

PDF

Exploring remote sensing and geographic information systems technologies to understand vegetation changes in response to land management practices at Finke Gorge National Park, Australia Between ...

PDF

Assessing the use of normalized difference chlorophyll index to estimate chlorophyll-A concentrations using Landsat 5 TM and Landsat 8 OLI imagery in the Salton Sea, California

PDF

Satellite derived bathymetry in the Canadian Archipelago using multispectral and LiDAR space-based remote sensing

PDF

Impacts of vegetation management on wildfire severity: a study of the 2021 Caldor fire

PDF

Human suffering during wartime: a StoryMap of violations of international law during the Russo-Ukrainian War

PDF

Surface rupture detection with support vector machine classification: case study from Ridgecrest, CA

PDF

Utilizing GIS and remote sensing to determine sheep grazing patterns for best practices in land management protocols

PDF

Exploring the pernicious effects of redlining and discriminatory policies on an American city: a spatio-temporal case study of New York City

PDF

An exploratory spatial analysis of fire service and EMS accessibility in northeastern Illinois communities

PDF

Evaluating machine learning tools for humanitarian road network mapping

PDF

Detection and accuracy assessment of mountain pine beetle infestations using Landsat 8 OLI and WorldView02 satellite imagery: Lake Tahoe Basin-Nevada and California

PDF

A spatial and temporal exploration of how satellite communication devices impact mountain search and rescue missions in California’s Sierra Nevada mountain range

PDF

Enriching the Demographic Survey sampling for the Los Angeles County Annual Homeless Count with spatial statistics

PDF

Deep convolutional neural networks for remote sensing investigation of looting of the archeological site of Al-Lisht, Egypt

PDF

Using GIS to predict human movement patterns in complex humanitarian emergencies: a test case of the Syrian Conflict

PDF

Finding the green in greenspace: an examination of geospatial measures of greenspace for use in exposure studies

Asset Metadata

Creator Bosworth, Rebecca (author)

Core Title Assessing modern conflict to monitor human rights with remote sensing: Russia's war in Ukraine

School College of Letters, Arts and Sciences

Degree Master of Science

Degree Program Geographic Information Science and Technology

Degree Conferral Date 2023-08

Publication Date 08/09/2023

Defense Date 07/03/2023

Publisher University of Southern California. Libraries (digital)

Tag building damage,human rights,international humanitarian law,OAI-PMH Harvest,remote sensing,Russia,satellite,synthetic aperture radar,Ukraine

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Sedano, Elisabeth (committee chair), Fleming, Steven (committee member), Qi, Yi (committee member), Ruddell, Darren (committee member)

Creator Email bosworth@usc.edu,rebecca.e.bosworth@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC113297709

Unique identifier UC113297709

Identifier etd-BosworthRe-12233.pdf (filename)

Legacy Identifier etd-BosworthRe-12233

Document Type Thesis

Rights Bosworth, Rebecca

Internet Media Type application/pdf

Type texts

Source 20230809-usctheses-batch-1082 (batch), University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright. The original signature page accompanying the original submission of the work to the USC Libraries is retained by the USC Libraries and a copy of it may be obtained by authorized requesters contacting the repository e-mail address given.

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA

Repository Email cisadmin@lib.usc.edu