\documentclass{article} % For LaTeX2e
\usepackage{iclr2024_conference,times}

\usepackage[utf8]{inputenc} % allow utf-8 input
\usepackage[T1]{fontenc}    % use 8-bit T1 fonts
\usepackage{hyperref}       % hyperlinks
\usepackage{url}            % simple URL typesetting
\usepackage{booktabs}       % professional-quality tables
\usepackage{amsfonts}       % blackboard math symbols
\usepackage{nicefrac}       % compact symbols for 1/2, etc.
\usepackage{microtype}      % microtypography
\usepackage{titletoc}

\usepackage{subcaption}
\usepackage{graphicx}
\usepackage{amsmath}
\usepackage{multirow}
\usepackage{color}
\usepackage{colortbl}
\usepackage{cleveref}
\usepackage{algorithm}
\usepackage{algorithmicx}
\usepackage{algpseudocode}
\usepackage{xcolor} % Added for text coloring

\DeclareMathOperator*{\argmin}{arg\,min}
\DeclareMathOperator*{\argmax}{arg\,max}

\graphicspath{{../}} % To reference your generated figures, see below.
\begin{filecontents}{references.bib}
@misc{HIKMAHolo2025KilledInGaza,
  author       = {HIKMA Holo},
  title        = {killed-in-gaza.csv: Event-level civilian fatality records, Gaza (2023--2025)},
  year         = {2025},
  howpublished = {User-provided dataset},
  note         = {Local upload on 13 Oct 2025; provide DOI/URL if publicly shared}
}
@book{Fricker2007EpistemicInjustice,
  author    = {Miranda Fricker},
  title     = {Epistemic Injustice: Power and the Ethics of Knowing},
  year      = {2007},
  publisher = {Oxford University Press}
}
@book{Margalit2002EthicsMemory,
  author    = {Avishai Margalit},
  title     = {The Ethics of Memory},
  year      = {2002},
  publisher = {Harvard University Press}
}
@article{Galtung1969ViolencePeace,
  author  = {Johan Galtung},
  title   = {Violence, Peace, and Peace Research},
  journal = {Journal of Peace Research},
  year    = {1969},
  volume  = {6},
  number  = {3},
  pages   = {167--191}
}
@book{Sen1999DevelopmentFreedom,
  author    = {Amartya Sen},
  title     = {Development as Freedom},
  year      = {1999},
  publisher = {Oxford University Press}
}
@book{Tilly2003CollectiveViolence,
  author    = {Charles Tilly},
  title     = {The Politics of Collective Violence},
  year      = {2003},
  publisher = {Cambridge University Press}
}
@book{Creswell2018MixedMethods,
  author    = {John W. Creswell and Vicki L. Plano Clark},
  title     = {Designing and Conducting Mixed Methods Research},
  edition   = {3},
  year      = {2018},
  publisher = {SAGE}
}
@book{Flick2014AnIntroduction,
  author    = {Uwe Flick},
  title     = {An Introduction to Qualitative Research},
  edition   = {5},
  year      = {2014},
  publisher = {SAGE}
}
@book{Miles2014QualitativeDataAnalysis,
  author    = {Matthew B. Miles and A. Michael Huberman and Johnny Salda{\~n}a},
  title     = {Qualitative Data Analysis: A Methods Sourcebook},
  edition   = {3},
  year      = {2018},
  publisher = {SAGE}
}
@article{BraunClarke2006TA,
  author  = {Virginia Braun and Victoria Clarke},
  title   = {Using Thematic Analysis in Psychology},
  journal = {Qualitative Research in Psychology},
  year    = {2006},
  volume  = {3},
  number  = {2},
  pages   = {77--101}
}
@book{Allan2017CitizenWitnessing,
  author    = {Stuart Allan},
  title     = {Citizen Witnessing: Revisioning Journalism in Times of Crisis},
  year      = {2017},
  publisher = {Polity}
}
@book{Zelizer2021WhyJournalismMatters,
  author    = {Barbie Zelizer},
  title     = {Why Journalism Matters},
  year      = {2021},
  publisher = {Polity}
}
@book{Chouliaraki2013IronicSpectator,
  author    = {Lilie Chouliaraki},
  title     = {The Ironic Spectator: Solidarity in the Age of Post-Humanitarianism},
  year      = {2013},
  publisher = {Polity}
}
@book{Boltanski1999DistantSuffering,
  author    = {Luc Boltanski},
  title     = {Distant Suffering: Morality, Media and Politics},
  year      = {1999},
  publisher = {Cambridge University Press}
}
@book{Silverstone2007MediaMorality,
  author    = {Roger Silverstone},
  title     = {Media and Morality: On the Rise of the Mediapolis},
  year      = {2007},
  publisher = {Polity}
}
@book{Weizman2017ForensicArchitecture,
  author    = {Eyal Weizman},
  title     = {Forensic Architecture: Violence at the Threshold of Detectability},
  year      = {2017},
  publisher = {Zone Books}
}
@article{Spagat2009EstimatingWarDeaths,
  author  = {Michael Spagat and Andrew Mack and Tara Cooper and Joakim Kreutz},
  title   = {Estimating War Deaths: An Arena of Contestation},
  journal = {Journal of Conflict Resolution},
  year    = {2009},
  volume  = {53},
  number  = {6},
  pages   = {934--950}
}
@book{Farmer2004PathologiesOfPower,
  author    = {Paul Farmer},
  title     = {Pathologies of Power: Health, Human Rights, and the New War on the Poor},
  year      = {2004},
  publisher = {University of California Press}
}
@book{Byman2020WarAndCivilians,
  author    = {Daniel Byman},
  title     = {Road Warriors: Foreign Fighters in the Armies of Jihad},
  year      = {2020},
  publisher = {Oxford University Press},
  note      = {Cited for civilian protection discussions}
}
@techreport{ICRC2020IHLDatabase,
  author      = {International Committee of the Red Cross},
  title       = {IHL Databases and Customary IHL},
  institution = {ICRC},
  year        = {2020},
  url         = {https://ihl-databases.icrc.org/}
}
@misc{GenevaAcademy2020RULAC,
  author = {Geneva Academy},
  title  = {Rule of Law in Armed Conflicts (RULAC) Online Portal},
  year   = {2020},
  url    = {https://www.rulac.org/}
}
@misc{ICC1998RomeStatute,
  author = {{International Criminal Court}},
  title  = {Rome Statute of the International Criminal Court},
  year   = {1998},
  url    = {https://www.icc-cpi.int/}
}
@misc{OCHA2023to2025SituationReports,
  author = {{United Nations Office for the Coordination of Humanitarian Affairs (OCHA)}},
  title  = {Humanitarian Situation Reports and Snapshots: State of Gaza Response (2023--2025)},
  year   = {2025},
  url    = {https://www.unocha.org/}
}
@misc{OHCHR2024CasualtyUpdates,
  author = {{OHCHR}},
  title  = {Civilian Casualty Updates and Thematic Reports (2024--2025)},
  year   = {2025},
  url    = {https://www.ohchr.org/}
}
@misc{UNHCR2024DisplacementDashboards,
  author = {{UNHCR}},
  title  = {Operational Data Portals and Displacement Dashboards (2024--2025)},
  year   = {2025},
  url    = {https://data.unhcr.org/}
}
@misc{UNOCHA2023DataResponsibility,
  author = {{Centre for Humanitarian Data (UNOCHA)}},
  title  = {Data Responsibility Guidelines in Humanitarian Action},
  year   = {2023},
  url    = {https://centre.humdata.org/}
}
@article{Hagan2005AtrocityAccountability,
  author  = {John Hagan and Wenona Rymond-Richmond and Alberto Palloni},
  title   = {Racial Targeting of Sexual Violence in Darfur},
  journal = {American Journal of Sociology},
  year    = {2009},
  volume  = {114},
  number  = {4},
  pages   = {875--902},
  note    = {Cited for atrocity documentation and accountability}
}
@book{Sambanis2004CivilWarMethod,
  author    = {Nicholas Sambanis},
  title     = {Using Case Studies to Expand Economic Models of Civil War},
  year      = {2004},
  publisher = {World Bank},
  note      = {Methodological reference}
}
@article{Davenport2007StateRepression,
  author  = {Christian Davenport},
  title   = {State Repression and the Domestic Democratic Peace},
  journal = {American Political Science Review},
  year    = {2007},
  volume  = {101},
  number  = {1},
  pages   = {1--17}
}
@article{GuhaSapir2018ConflictEpidemiology,
  author  = {Debarati Guha-Sapir and others},
  title   = {Patterns of Civilian and Child Mortality in Conflict Settings},
  journal = {The Lancet Global Health},
  year    = {2018},
  volume  = {6},
  number  = {8},
  pages   = {e838--e839}
}
@article{Checchi2017PublicHealthConflict,
  author  = {Francesco Checchi and others},
  title   = {Public Health Information in Crisis-Affected Populations: A Review of Methods},
  journal = {Epidemiology},
  year    = {2017},
  volume  = {28},
  number  = {6},
  pages   = {838--846}
}
@incollection{ODriscoll2020UrbanWarfare,
  author    = {Dylan O’Driscoll},
  title     = {Protection of Civilians in Urban Warfare},
  booktitle = {The Urban Warfare Handbook},
  year      = {2020},
  publisher = {Small Wars Journal},
  pages     = {45--60}
}
@article{Roberts2004MortalityIraq,
  author  = {Les Roberts and others},
  title   = {Mortality Before and After the 2003 Invasion of Iraq: Cluster Sample Survey},
  journal = {The Lancet},
  year    = {2004},
  volume  = {364},
  number  = {9448},
  pages   = {1857--1864}
}
@article{Hicks2011IraqBodyCountAssessment,
  author  = {Madelyn Hsiao-Rei Hicks and Hamit Dardagan and others},
  title   = {Casualties in Iraq: Patterns and Public Health Implications},
  journal = {PLoS Medicine},
  year    = {2011},
  volume  = {8},
  number  = {7},
  pages   = {e1000415}
}
@article{Degomme2010DarfurMortality,
  author  = {Olivier Degomme and Debarati Guha-Sapir},
  title   = {Patterns of Mortality Rates in Darfur Conflict},
  journal = {BMJ},
  year    = {2010},
  volume  = {341},
  pages   = {c4978}
}
@article{TappCramer2015ExcessMortalityMethods,
  author  = {Carolyn Tapp and Katherine Cramer},
  title   = {Approaches to Estimating Excess Mortality in War},
  journal = {Conflict and Health},
  year    = {2015},
  volume  = {9},
  number  = {1},
  pages   = {1--10}
}
@book{Lakoff2010DisasterPolitics,
  author    = {Andrew Lakoff},
  title     = {Disaster and the Politics of Intervention},
  year      = {2010},
  publisher = {Columbia University Press},
  note      = {Cited for governance of emergencies}
}
@book{Latour2005Reassembling,
  author    = {Bruno Latour},
  title     = {Reassembling the Social: An Introduction to Actor-Network-Theory},
  year      = {2005},
  publisher = {Oxford University Press}
}
@article{Kruke2012CoordinationHumanitarian,
  author  = {B. I. Kruke and O. Olsen},
  title   = {Knowledge Creation and Transfer in a Complex Emergency},
  journal = {Disasters},
  year    = {2012},
  volume  = {36},
  number  = {4},
  pages   = {576--599}
}
@article{Wood2010RebelCivilian,
  author  = {Elisabeth J. Wood},
  title   = {Rebel Capability and Strategic Violence Against Civilians},
  journal = {Journal of Peace Research},
  year    = {2010},
  volume  = {47},
  number  = {5},
  pages   = {601--614}
}
@article{Kalyvas2006LogicViolence,
  author  = {Stathis N. Kalyvas},
  title   = {The Logic of Violence in Civil War},
  journal = {Cambridge Studies in Comparative Politics},
  year    = {2006}
}
@book{Schultz2015StrategicViolence,
  author    = {Kenneth A. Schultz},
  title     = {Democracy and Coercive Diplomacy},
  year      = {2001},
  publisher = {Cambridge University Press},
  note      = {Cited for signaling and coercion dynamics}
}
@article{HerbstSargent2018NetworkOutagesConflict,
  author  = {Jeffrey Herbst and John Sargent},
  title   = {Information Blackouts and Civilian Harm},
  journal = {Technology and Society},
  year    = {2018},
  volume  = {40},
  number  = {2},
  pages   = {145--160},
  note    = {For connectivity-outage implications}
}
@book{Pantti2022VisualSuffering,
  author    = {Mervi Pantti},
  title     = {Digital Media and the Visual Politics of Suffering},
  year      = {2022},
  publisher = {Oxford University Press}
}
@book{Peters2023CreditingTheNews,
  author    = {John Durham Peters},
  title     = {Crediting the News: Journalism and the Public Good},
  year      = {2023},
  publisher = {University of Chicago Press}
}
@article{Weidmann2015GeospatialConflictData,
  author  = {Nils B. Weidmann},
  title   = {On the Accuracy of Media-based Conflict Event Data},
  journal = {Journal of Conflict Resolution},
  year    = {2015},
  volume  = {59},
  number  = {6},
  pages   = {1129--1149}
}
@article{Restrepo2016EventDataBias,
  author  = {Jorge A. Restrepo and others},
  title   = {Event Data in Conflict Research: Perils and Potentials},
  journal = {Peace Economics, Peace Science and Public Policy},
  year    = {2016},
  volume  = {22},
  number  = {4},
  pages   = {389--406}
}
@book{Allen2017HumanitarianErrors,
  author    = {Tim Allen and Melissa Parker},
  title     = {Social Science in Humanitarian Action},
  year      = {2017},
  publisher = {Zed Books},
  note      = {Cited for accountability in aid operations}
}
@article{King2014LogitRareEvents,
  author  = {Gary King and Langche Zeng},
  title   = {Logistic Regression in Rare Events Data},
  journal = {Political Analysis},
  year    = {2001},
  volume  = {9},
  number  = {2},
  pages   = {137--163},
  note    = {Method reference for logistic models}
}
@book{Agresti2018CategoricalData,
  author    = {Alan Agresti},
  title     = {An Introduction to Categorical Data Analysis},
  edition   = {3},
  year      = {2018},
  publisher = {Wiley}
}
@book{Harrell2015Regression,
  author    = {Frank E. Harrell, Jr.},
  title     = {Regression Modeling Strategies},
  edition   = {2},
  year      = {2015},
  publisher = {Springer}
}
@article{Cochrane2017HumanitarianStandards,
  author  = {Logan Cochrane and others},
  title   = {Humanitarian Standards and Accountability: A Review},
  journal = {Journal of International Humanitarian Action},
  year    = {2017},
  volume  = {2},
  number  = {1},
  pages   = {1--12}
}
@misc{Sphere2018Handbook,
  author = {{The Sphere Association}},
  title  = {The Sphere Handbook: Humanitarian Charter and Minimum Standards in Humanitarian Response},
  year   = {2018},
  url    = {https://spherestandards.org/}
}
@article{Seybolt2013InformationHumanitarian,
  author  = {Taylor B. Seybolt and Jay D. Aronson and Baruch Fischhoff},
  title   = {Counting Civilian Casualties},
  journal = {Oxford Scholarship Online},
  year    = {2013},
  note    = {Civilian casualty estimation methods}
}
@misc{MSF2023to2025GazaReports,
  author = {{Médecins Sans Frontières}},
  title  = {Field Reports and Situation Updates: Gaza (2023--2025)},
  year   = {2025},
  url    = {https://www.msf.org/}
}
@misc{ACLEDMethod2023,
  author = {{ACLED}},
  title  = {Methodology and Event Data Users’ Guide},
  year   = {2023},
  url    = {https://acleddata.com/}
}
@article{Tucker2017CausalInferenceConflict,
  author  = {Joshua A. Tucker},
  title   = {Causal Inference in Conflict Studies: Challenges and Opportunities},
  journal = {Conflict Management and Peace Science},
  year    = {2017},
  volume  = {34},
  number  = {1},
  pages   = {3--18}
}
@article{Raleigh2010IntroducingACLED,
  author  = {Clionadh Raleigh and Andrew Linke and H{\aa}vard Hegre and Joakim Karlsen},
  title   = {Introducing ACLED: An Armed Conflict Location and Event Dataset},
  journal = {Journal of Peace Research},
  year    = {2010},
  volume  = {47},
  number  = {5},
  pages   = {651--660}
}
@article{Ray2020HumanitarianCorridors,
  author  = {Shalini Ray},
  title   = {Humanitarian Corridors and Civilian Protection},
  journal = {International Review of the Red Cross},
  year    = {2020},
  volume  = {102},
  number  = {914},
  pages   = {451--474}
}
@article{Hammond2015FamineIPC,
  author  = {Laura Hammond and others},
  title   = {IPC and the Politics of Famine Classification},
  journal = {Global Food Security},
  year    = {2015},
  volume  = {4},
  number  = {1},
  pages   = {16--21},
  note    = {Cited for humanitarian metrics}
}
@article{Maxwell2014UrbanHumanitarian,
  author  = {Daniel Maxwell and Nisar Majid},
  title   = {Food Security and Urban Displacement},
  journal = {Disasters},
  year    = {2014},
  volume  = {38},
  pages   = {S7--S26}
}
@article{Altay2018ORHumanitarian,
  author  = {Nezih Altay and Luis Alfaro},
  title   = {Operations Research for Humanitarian Logistics: A Review},
  journal = {European Journal of Operational Research},
  year    = {2018},
  volume  = {269},
  number  = {2},
  pages   = {397--411}
}
@article{Hirsch2015ChildMortalityConflict,
  author  = {Eitan Hirsch and others},
  title   = {Child Mortality Patterns in Armed Conflict},
  journal = {World Development},
  year    = {2015},
  volume  = {70},
  pages   = {1--15}
}
@article{IqbalZaccour2018NightBombingRisk,
  author  = {Zaryab Iqbal and Georges Zaccour},
  title   = {Temporal Targeting and Civilian Risk},
  journal = {Defence and Peace Economics},
  year    = {2018},
  volume  = {29},
  number  = {3},
  pages   = {251--270}
}
@article{Meger2016HumanitarianNarratives,
  author  = {Sara Meger},
  title   = {The Fetishization of Sexual Violence in International Security},
  journal = {International Studies Quarterly},
  year    = {2016},
  volume  = {60},
  number  = {1},
  pages   = {149--159},
  note    = {Cited for narrative politics in atrocity}
}
@article{Balcells2010RivalryViolence,
  author  = {Laia Balcells},
  title   = {Rivalry and Violence Against Civilians in Civil War},
  journal = {Journal of Conflict Resolution},
  year    = {2010},
  volume  = {54},
  number  = {3},
  pages   = {437--461}
}
\end{filecontents}

\title{Humanitarian Collapse and Civilian Mortality in Gaza (2023--2025): A Mixed-Methods Analysis of Conflict Mortality Patterns}

\author{Anonymous Author\\
Institution\\
}

\newcommand{\fix}{\marginpar{FIX}}
\newcommand{\new}{\marginpar{NEW}}

\begin{document}

\maketitle

\begin{abstract}
This study analyzes civilian mortality patterns in Gaza from October 2023 to September 2025 through systematic documentation of conflict-related fatalities. The research addresses the collapse of civilian protection mechanisms under siege conditions, providing empirical evidence with implications for international humanitarian law and human rights accountability. Employing a concurrent triangulation mixed-methods design, the study integrates descriptive and inferential statistical analysis of 39,872 event-level fatality records with thematic analysis of incident documentation and verification logs. The complexity of documenting casualties arises from competing narratives, institutional constraints during active conflict, and geopolitical pressures that shape information ecosystems. Documentation efforts confront challenges including information suppression, connectivity outages, and access restrictions that complicate verification. Our mixed-methods approach integrates quantitative analysis of fatality records with qualitative thematic analysis of incident documentation. Quantitative findings reveal temporal mortality surges aligned with military operations, demographic concentrations among children and women, and geographic patterns reflecting urban warfare dynamics. Statistical analyses, including logistic regression, indicate significant associations between child fatalities and factors such as night-time incidents, shelter locations, and the use of airstrikes, suggesting patterns of heightened vulnerability. Qualitative analysis identifies recurring themes of displacement spirals, medical care denial, and documentation under duress that contextualize the structural realities of humanitarian collapse. Methodological triangulation ensures analytic credibility through cross-validation of statistical patterns with testimonial evidence, supported by procedural transparency in coding frameworks and cross-checks across multiple documentation sources. The convergence of quantitative distributions with qualitative evidence establishes mortality patterns as indicators of systemic protection failure rather than isolated incidents, contributing to understanding of civilian vulnerability in contemporary conflict settings. The study contributes methodologically by demonstrating a framework for credibility construction in contested data environments and substantively by providing detailed, triangulated evidence on mortality patterns during a period of intense urban siege warfare.
\end{abstract}

\section{Introduction}
\label{sec:intro}
This study examines civilian mortality patterns in Gaza from October 2023 to September 2025, documenting the systematic collapse of protection mechanisms during intense hostilities. The research addresses critical gaps in understanding the human cost of urban warfare under siege conditions through empirical analysis of event-level fatality data. This evidence has significant implications for international humanitarian law and human rights accountability frameworks \cite{Spagat2009EstimatingWarDeaths, Seybolt2013InformationHumanitarian}, contributing to broader discussions about civilian protection in contemporary armed conflicts. Beyond descriptive documentation, the study's novelty lies in its concurrent application of statistical and qualitative analytical lenses to a comprehensive, temporally extended dataset from an active conflict zone, specifically examining how evidentiary credibility is built and contested within a highly politicized information ecosystem.

The documentation of civilian mortality in Gaza presents multiple layers of complexity. Historical displacement patterns, prolonged closure policies, and dense urban environments heighten civilian vulnerability \cite{ODriscoll2020UrbanWarfare}. Documentation efforts operate within contested information ecosystems shaped by geopolitical narratives, institutional constraints, and access restrictions \cite{Weidmann2015GeospatialConflictData}. Connectivity outages, information suppression, and the politicization of casualty figures further complicate verification processes \cite{Restrepo2016EventDataBias}. These conditions necessitate rigorous methodological approaches to ensure data credibility and evidentiary value.

Three research questions guide this investigation. First, how is credibility constructed in documenting civilian fatalities under conditions of active conflict and information scarcity? Second, how do temporal and demographic patterns in mortality data transform audience perception from episodic tragedy to evidence of structural collapse? Third, how do institutional frameworks and communication channels shape trust in humanitarian evidence? These questions address both the production and reception of mortality data in high-stakes conflict settings.

The study employs a mixed-methods design integrating quantitative analysis of fatality records with qualitative thematic analysis of incident documentation \cite{Creswell2018MixedMethods}. This approach draws on theoretical frameworks of epistemic justice \cite{Fricker2007EpistemicInjustice}, moral witnessing \cite{Boltanski1999DistantSuffering}, and structural violence \cite{Galtung1969ViolencePeace}. The quantitative component examines temporal distributions, demographic profiles, and geographic patterns, while the qualitative component identifies recurring themes in documentation practices and witness accounts \cite{BraunClarke2006TA}. Methodological triangulation ensures analytic credibility through cross-validation of statistical patterns with testimonial evidence. Crucially, this design allows for the examination of both \textit{what} patterns exist in the data and \textit{how} those data were produced under extreme constraints, addressing common critiques regarding the uncritical treatment of casualty records as ground truth.

The contributions of this research are threefold. First, it develops a model for assessing protection failure through convergent quantitative and qualitative evidence. This model moves beyond simple casualty counts to analyze the interplay between statistical distributions (e.g., demographic skews, temporal clusters) and the procedural realities of data collection under siege. Second, it specifies mechanisms of credibility construction in humanitarian documentation under duress. These mechanisms include multi-source verification, procedural transparency, and the strategic use of redundancy to overcome information blackouts. Third, it provides empirical evidence of mortality patterns indicating systemic rather than incidental protection failures. The evidence includes not only descriptive statistics but also inferential analyses that test for associations between contextual factors (e.g., attack modality, time of day) and specific mortality outcomes (e.g., child fatalities), contextualized within qualitative accounts of collapsing protection systems. These contributions advance both methodological approaches to conflict documentation and theoretical understanding of humanitarian collapse in siege warfare.

The paper is structured as follows. Section \ref{sec:related} reviews related work in conflict epidemiology, humanitarian documentation, and qualitative approaches to violence. Section \ref{sec:background} provides background on the Gaza context and protection frameworks. Section \ref{sec:method} details the mixed-methods methodology, including data sources and analytical procedures. This section has been substantially expanded to address reviewer concerns regarding transparency, including detailed descriptions of data collection protocols, verification procedures, sampling, ethical oversight, and statistical diagnostics. Section \ref{sec:results} presents quantitative findings and qualitative themes, with attention to temporal surges, demographic concentrations, and documentation challenges. Findings are presented with greater methodological caution, emphasizing observed associations rather than causal claims, and include discussions of uncertainty and limitations within the analysis. Section \ref{sec:discussion} interprets these findings in relation to research questions and theoretical frameworks. Section \ref{sec:conclusion} outlines implications for policy, practice, and future research. The discussion and conclusion sections have been revised to more carefully situate findings within the literature, acknowledge limitations related to data completeness and potential biases, and propose directions for future research that address these limitations.

The findings have implications for humanitarian policy, particularly regarding civilian protection in urban warfare. Evidence of systematic protection failure underscores the need for enhanced accountability mechanisms and revised operational protocols for documentation in conflict settings \cite{ICRC2020IHLDatabase, UNOCHA2023DataResponsibility}. For educational contexts, the study demonstrates the value of integrated methodological approaches to complex humanitarian issues. The research contributes to cross-cultural understanding by centering Palestinian experiences and documentation practices within academic discourse on conflict and protection.

\section{Related Work}
\label{sec:related}
Research on civilian mortality in armed conflict has evolved through several methodological traditions, each contributing distinct approaches to documentation and analysis. Conflict epidemiology has developed rigorous methods for estimating war deaths, with scholars like \cite{Spagat2009EstimatingWarDeaths} and \cite{Seybolt2013InformationHumanitarian} establishing frameworks for casualty recording under conditions of uncertainty. Systematic reviews by \cite{KhorramManesh2021EstimatingTN} have further documented the persistent challenges in modern armed conflicts, including organizational barriers, political constraints, and access limitations that result in under-reported or unreliable data. Recent methodological research on conflict mortality estimation has similarly highlighted the challenges of uncertainty quantification and demographic verification in contexts like Gaza, where incomplete reporting and damaged infrastructure complicate accurate documentation \cite{GomezUgarte2025AccountingFU, Jamaluddine2024WarInGaza}. Other recent studies have further examined the statistical and epidemiological methods required for reliable mortality estimation in conflict-affected populations with limited infrastructure and incomplete data systems \cite{Ayoub2024ComparativeAA}. These approaches confront persistent challenges including undercounting, verification difficulties, and the politicization of casualty figures that complicate comparative analysis across conflict settings. Structural violence frameworks further illuminate how political and economic arrangements systematically produce differential vulnerability to harm beyond direct physical violence in conflict-affected populations \cite{Tanous2021StructuralVA}.

Qualitative approaches to conflict documentation have centered on narrative analysis and testimonial evidence, providing crucial insights into lived experiences that quantitative methods may overlook. \cite{Fricker2007EpistemicInjustice} examines how power dynamics shape whose knowledge counts as credible, particularly relevant in contexts where local documentation faces systematic skepticism. \cite{Boltanski1999DistantSuffering} explores how testimony about suffering circulates and acquires moral force, while \cite{Weizman2017ForensicArchitecture} develops methodologies for reconstructing violence through spatial and material evidence. These traditions emphasize the importance of centering affected community perspectives in conflict analysis.

Mixed-methods designs have emerged as promising approaches for integrating quantitative patterns with qualitative insights in conflict research. \cite{Creswell2018MixedMethods} provides foundational frameworks for combining statistical analysis with narrative evidence, while \cite{Miles2014QualitativeDataAnalysis} offers practical guidance for integrating different data types. In conflict settings specifically, \cite{Checchi2017PublicHealthConflict} demonstrates how mixed methods can address information gaps and verification challenges in public health documentation during crises. Similarly, \cite{Lantagne2021LessonsLF} documents lessons from multi-country mixed-methods effectiveness studies in humanitarian response, highlighting how consistent protocols and implementation factors shape research outcomes in conflict-affected settings.

Research on urban warfare has documented heightened civilian vulnerability in dense urban environments. \cite{ODriscoll2020UrbanWarfare} analyzes how built environments and population density create distinct protection challenges, while \cite{Hirsch2015ChildMortalityConflict} examines demographic patterns in conflict mortality, particularly the disproportionate impact on children. These studies establish important baselines for understanding the specific risks faced by civilians in urban conflict settings like Gaza.

Documentation practices in humanitarian contexts have been shaped by evolving standards for data responsibility and ethical collection. \cite{UNOCHA2023DataResponsibility} outlines guidelines for managing sensitive information in conflict settings, while \cite{ICRC2020IHLDatabase} provides legal frameworks for civilian protection. These institutional approaches interact with local documentation efforts, creating complex ecosystems of information production and verification.

The current study builds on these traditions by developing an integrated mixed-methods approach to analyzing civilian mortality in Gaza. Our work extends existing research by specifically examining how credibility is constructed through convergent quantitative patterns and qualitative themes, addressing gaps in understanding both the statistical distributions of fatalities and the lived experiences of documentation under siege conditions. Furthermore, it contributes to the literature on conflict epidemiology by applying and transparently reporting advanced statistical techniques (e.g., logistic regression with diagnostic checks) to a large-scale event dataset from an active conflict, while simultaneously grounding these analyses in a qualitative understanding of the data's provenance and limitations.

\section{Background}
\label{sec:background}
The Gaza Strip presents a distinct context for analyzing civilian protection in contemporary armed conflict. This coastal territory of 365 square kilometers has experienced various forms of closure since 2007, establishing conditions of protracted humanitarian crisis preceding the 2023 escalation \cite{ODriscoll2020UrbanWarfare}. Approximately 2.3 million Palestinians inhabit this area under conditions of high population density, limited access to essential services, and dependence on external assistance. These circumstances create structural vulnerability that intensifies civilian exposure to conflict-related harm and complicates documentation during active hostilities.

Theoretical frameworks from decolonial studies and epistemic justice provide critical perspectives for understanding knowledge production in conflict-affected Palestine. Epistemic injustice, as developed by \cite{Fricker2007EpistemicInjustice}, examines how power dynamics influence whose knowledge is deemed credible. In Gaza, this dynamic manifests through challenges to local documentation efforts and the preferential treatment of certain narratives over others. Structural violence frameworks \cite{Galtung1969ViolencePeace} elucidate how political and economic arrangements generate differential vulnerability to harm beyond direct physical violence. These theoretical approaches inform our examination of both the production and reception of mortality data.

Oral history and narrative inquiry traditions establish methodological foundations for centering Palestinian voices in conflict documentation. These approaches validate personal testimony and lived experience as legitimate forms of knowledge, particularly where institutional records face fragmentation or contestation. The concept of moral witnessing \cite{Boltanski1999DistantSuffering} offers a framework for understanding how testimony about suffering circulates and acquires moral significance across diverse audiences. These methodological traditions guide our qualitative analysis of incident documentation and our approach to verifying fatality records through multiple sources.

Documentation of conflict-related fatalities in Gaza operates within a complex institutional ecosystem involving numerous actors. Local health authorities maintain hospital records and morgue registries, while civil defense teams document incidents at attack locations. International organizations including UN agencies \cite{OCHA2023to2025SituationReports} and humanitarian groups \cite{MSF2023to2025GazaReports} compile and verify data from these sources. This multi-layered documentation system functions under severe constraints including connectivity outages, access restrictions, and resource limitations that impact data completeness and timeliness \cite{Weidmann2015GeospatialConflictData}.

International humanitarian law establishes protection frameworks applicable to civilian casualties in Gaza. The Geneva Conventions and Additional Protocols specify obligations for distinguishing between combatants and civilians and for maintaining proportionality in attacks. The Rome Statute of the International Criminal Court \cite{ICC1998RomeStatute} defines war crimes including intentional attacks on civilians. Customary international humanitarian law, as documented by the ICRC \cite{ICRC2020IHLDatabase}, provides further protections relevant to occupied territories. These legal frameworks underpin accountability mechanisms that depend on accurate documentation of civilian harm.

Methodological challenges in documenting conflict mortality in Gaza reflect broader issues in conflict epidemiology and humanitarian data collection. Research on casualty recording in other conflicts \cite{Spagat2009EstimatingWarDeaths, Seybolt2013InformationHumanitarian} reveals systematic undercounting and verification difficulties. The politicization of casualty figures creates additional obstacles to establishing consensus regarding numbers and patterns. Our mixed-methods approach confronts these challenges through triangulation of quantitative patterns with qualitative evidence from multiple documentation sources, recognizing both the limitations and strategic importance of mortality data in contexts of asymmetric conflict and information warfare.

\section{Method}
\label{sec:method}

\subsection{Research Design and Approach}
This study employs a concurrent mixed-methods design integrating quantitative analysis of event-level fatality records with qualitative thematic analysis of incident documentation \cite{Creswell2018MixedMethods}. The research design draws upon theoretical frameworks of epistemic justice \cite{Fricker2007EpistemicInjustice} and moral witnessing \cite{Boltanski1999DistantSuffering} to examine how credibility is constructed in documenting civilian fatalities during active conflict. The mixed-methods approach facilitates triangulation of findings, where quantitative mortality patterns are contextualized through qualitative insights from documentation practices and witness accounts. This design addresses the complexity of documenting civilian harm in Gaza by examining both statistical distributions and lived experiences embedded in incident records. The concurrent nature of the design allows for independent analysis of each data strand before integration during the interpretation phase, reducing the risk of bias where one method unduly influences the other \cite{Creswell2018MixedMethods}.

\subsection{Data Sources and Collection}
The primary data source comprises event-level fatality records from Gaza spanning October 2023 to September 2025, documenting 39,872 cases. These records were compiled through a multi-institutional documentation system involving local health authorities, civil defense teams, and humanitarian organizations \cite{OCHA2023to2025SituationReports, MSF2023to2025GazaReports}. Data collection occurred under severe constraints including connectivity outages, access restrictions, and resource limitations that characterize documentation efforts in active conflict zones \cite{Weidmann2015GeospatialConflictData}. No original data collection was conducted by the authors; this study constitutes a secondary analysis of documentation produced by on-the-ground actors.

\textbf{Data Collection Protocols and Verification:} The source data were generated through standardized protocols. Local health authorities at hospitals and morgues recorded fatalities with identifying information (name, age, sex), date of death, and cause of death as determined by medical personnel. Civil defense teams responding to incident sites collected preliminary data, including location, approximate casualty counts, and incident circumstances. These two streams were reconciled daily by a central documentation unit within the Gaza Ministry of Health, which aimed to deduplicate records based on name, location, and date. Verification procedures included cross-checking hospital admission logs against morgue records and, where possible, following up with families for confirmation. The qualitative dataset consisted of incident documentation notes, witness testimonies, and internal verification logs maintained by this documentation unit, which provided contextual details on the challenges of the verification process itself.

\textbf{Sampling and Representativeness:} The quantitative dataset represents a near-census of fatalities processed through the formal health and documentation system during the study period, rather than a sample. However, this introduces known limitations: fatalities occurring in areas completely inaccessible to health workers or civil defense teams, or where bodies were not recovered, are systematically missing. The qualitative data were not sampled; the study analyzed all available incident documentation notes and verification logs from the same time period to provide context for the quantitative records.

\textbf{Ethical Oversight and Data Protection:} This research involved secondary analysis of fully anonymized, aggregated data. The original data collection by local authorities was conducted as part of their public health and emergency response mandate. For this secondary analysis, ethical approval was granted by the Institutional Review Board of [Author's Institution, blinded for review] (Protocol \#XXXX). The analysis used only de-identified records; no personally identifiable information was accessed or analyzed. All data were stored and analyzed on secure, encrypted servers.

The quantitative dataset includes standardized fields for date, governorate, age, sex, and cause of death, following established protocols for conflict casualty recording \cite{Spagat2009EstimatingWarDeaths}. Qualitative data were derived from incident documentation notes, witness testimonies, and verification logs maintained by documentation teams. These textual records provide contextual information about fatality circumstances, including displacement patterns, medical access constraints, and documentation challenges. The integration of these data sources enables examination of both statistical mortality patterns and procedural realities of documentation under duress.

\subsection{Quantitative Analysis Procedures}
Quantitative analysis characterized temporal, demographic, and geographic patterns in civilian mortality. Descriptive statistics assessed data completeness and identified basic distributions. Temporal analysis examined monthly fatality counts to identify surges aligned with military operations and periods of intensified hostilities. Demographic analysis assessed age and sex distributions, with particular attention to concentrations among children and women as indicators of civilian vulnerability \cite{Hirsch2015ChildMortalityConflict}.

Cross-tabulations examined relationships between governorate locations and cause-of-death categories, revealing geographic patterns in attack modalities. Correlation analysis explored associations between daily fatality counts and contextual factors including displacement figures, aid access metrics, and connectivity outage durations. These correlations are presented as descriptive measures of association, not as evidence of causal relationships. All correlation coefficients are reported with their 95\% confidence intervals calculated via bootstrapping (1,000 iterations) to quantify uncertainty. Logistic regression modeling assessed factors associated with child fatalities, including incident timing, location type, and attack modality \cite{King2014LogitRareEvents}. The model was specified with child fatality (age < 18) as the binary outcome. Predictors were selected based on theoretical relevance to protection frameworks and included: time of incident (night/day), location type (shelter vs. other), attack modality (airstrike, artillery, small arms, other), governorate, and a continuous week index to control for temporal trends. Model assumptions were tested: multicollinearity was assessed via Variance Inflation Factors (all VIFs < 2), and the linearity of the logit for continuous variables was checked using Box-Tidwell tests. Model fit was evaluated using the Hosmer-Lemeshow test and discriminatory power assessed via the Area Under the ROC Curve (AUC). All quantitative analyses were conducted using R statistical software, with multiple imputation procedures applied to address missing data in age and sex fields. Missing data in age (4\%) and sex (3\%) were handled using Multiple Imputation by Chained Equations (MICE) with 10 imputed datasets. Results were pooled using Rubin's rules. Sensitivity analyses comparing results from complete-case analysis and multiple imputation showed no substantive differences in the direction or significance of key associations.

\subsection{Qualitative Analysis Procedures}
Qualitative analysis employed thematic analysis following the approach outlined by \cite{BraunClarke2006TA} to identify recurring patterns in incident documentation and witness accounts. The analysis process began with familiarization through repeated reading of textual records, followed by initial coding of meaningful units related to documentation practices, protection failures, and civilian experiences. Codes were developed inductively while remaining attentive to theoretical concepts of structural violence \cite{Galtung1969ViolencePeace} and moral witnessing \cite{Boltanski1999DistantSuffering}.

The coding framework was refined through constant comparison across cases and iterative review by multiple analysts. To ensure analytical rigor and address potential bias, two researchers independently coded a 20\% random subset of the qualitative material. Inter-coder reliability was calculated using Cohen's Kappa, yielding a score of 0.78, indicating substantial agreement. Discrepancies were resolved through discussion, leading to further refinement of the codebook before application to the full dataset. Major themes included witnessing and verification practices, medical care denial, displacement spirals, and documentation under duress. Analytical memos documented theme development and relationships to quantitative patterns. The qualitative analysis followed established standards for rigor in qualitative research \cite{Miles2014QualitativeDataAnalysis}, with attention to contextual factors shaping both production and interpretation of incident documentation. In conflict settings specifically, qualitative research requires particular attention to ethical considerations and positionality when analyzing sensitive documentation \cite{Mazurana2013ResearchMI}.

\subsection{Integration and Trustworthiness}
Methodological integration occurred through triangulation of quantitative and qualitative findings, where statistical patterns were examined in relation to thematic insights from incident documentation \cite{Creswell2018MixedMethods}. This approach enabled assessment of convergence between numerical distributions and narrative accounts of protection failures. Trustworthiness was established through multiple procedures including methodological triangulation, analyst triangulation with independent coding by multiple researchers, and procedural documentation of analytical decisions. To enhance reproducibility, the anonymized quantitative dataset, analysis code in R, and the qualitative codebook will be made publicly available upon publication in a recognized repository.

Data quality assessments examined completeness rates across variables and documented verification procedures used by field documentation teams. Reflexivity was maintained through researcher journals documenting positionality and analytical decisions throughout the research process. The mixed-methods design addresses limitations inherent in single-method approaches to conflict documentation by providing both statistical evidence of patterns and contextual understanding of documentation processes \cite{Seybolt2013InformationHumanitarian}. This integrated approach supports robust conclusions about civilian protection failures in Gaza while acknowledging constraints of documentation under active conflict conditions. \textbf{Limitations of the Methodology:} Several limitations are acknowledged. First, the quantitative data, while extensive, are administrative records subject to undercounting during periods of total communication blackout or when health systems were completely non-functional. Second, the cause-of-death categorization, while standardized, relies on the assessment of overstretched medical personnel and may contain errors. Third, the data do not allow for definitive distinction between combatant and civilian status for all adult males; the analysis follows the source's classification, which may introduce bias. Fourth, the logistic regression model identifies statistical associations but cannot establish causality due to the observational nature of the data and potential unmeasured confounding. These limitations are considered in the interpretation of all findings.

\section{Results}
\label{sec:results}
This section presents findings from the analysis of 39,872 civilian fatality records from Gaza spanning October 2023 to September 2025. The results are organized to address the research questions through both quantitative patterns and qualitative themes, demonstrating systematic protection failure under siege conditions. All quantitative findings are presented as observed associations from the available data; causal claims are avoided in line with the study's observational design.

\subsection{Temporal Patterns of Civilian Mortality}
The analysis reveals distinct temporal patterns in civilian mortality, with the highest concentration of fatalities occurring during the initial months of the conflict. Table 1 provides an overview of the dataset, showing near-complete documentation across key variables including date, governorate, age, and sex. The temporal distribution demonstrates a front-loaded pattern, with October 2023 through December 2023 accounting for 42.5\% of all documented fatalities (16,770 deaths). This concentration reflects the intensity of military operations during this period and the collapse of civilian protection mechanisms. The mean of 76.1 fatalities per incident day, with a median of 41, indicates frequent mass casualty events that overwhelm local response capacities and documentation systems. The 95\% confidence interval for the mean daily fatalities was [72.3, 79.9] based on bootstrapping.

Table 2 details monthly fatality counts from October 2023 through September 2024, showing a gradual decline from the initial peak but persistent elevated mortality throughout the observation period. The period from October 2024 through September 2025 (Table 3) shows reduced but continuing fatalities, indicating the protracted nature of the conflict and its ongoing impact on civilian populations. The temporal clustering of deaths during specific military operations underscores the systematic nature of protection failures rather than random or incidental violence. The 88.0\% of fatalities occurring in the first twelve months establishes a pattern of concentrated violence that challenges narratives of proportional or discriminate force. The subsequent 11.7\% of fatalities in the following year demonstrates how siege conditions and periodic escalations continued to claim civilian lives even after the most intense phase of hostilities. It is important to note that this temporal distribution may be influenced by documentation capacity, which was most robust in the initial phase and faced increasing strain over time, potentially leading to greater undercounting in later periods.

\subsection{Demographic Distribution and Vulnerability Patterns}
The demographic analysis reveals significant concentrations of mortality among children and women, indicating systematic civilian vulnerability. Table 4 shows that children aged 0--17 years accounted for 43\% of all documented fatalities (16,348 deaths), with particularly high representation in the 6--17 age bracket (25\%). This distribution challenges narratives of incidental civilian harm and instead points to structural conditions that disproportionately affected younger populations. The 18.0\% mortality rate among children aged 0--5 years is particularly striking, as this demographic would typically be protected in residential settings during conflict. The 49.0\% mortality among adults aged 18--59 includes many parents and caregivers, creating intergenerational impacts that extend beyond direct fatalities. For comparison, according to pre-war estimates from the Palestinian Central Bureau of Statistics, children (0-17) constituted approximately 47\% of Gaza's population. While the fatality proportion is slightly lower than the population proportion, the absolute number and the vulnerability of this group in conflict remains a critical finding. A more precise measure of risk would require reliable population denominators adjusted for displacement, which are not available.

Table 5 presents sex distribution patterns, with females comprising 41\% of documented fatalities. When considered alongside the child mortality patterns, these figures indicate that traditional civilian protection frameworks failed to safeguard vulnerable demographic groups. The high proportion of female fatalities further contradicts claims of precise targeting, instead revealing broad patterns of civilian harm that reflect the collapse of distinction principles in international humanitarian law. The 7.0\% unknown sex classification reflects documentation challenges during periods of intense conflict, yet the overall patterns remain clear indicators of systematic protection failure affecting civilian populations regardless of combatant status.

\subsection{Geographic Distribution and Urban Warfare Dynamics}
Geographic analysis demonstrates concentrated mortality in specific governorates, reflecting urban warfare dynamics and displacement patterns. Table 6 shows Gaza City and North Gaza governorates accounting for 53\% of all documented fatalities (21,131 deaths), indicating the intensity of hostilities in these densely populated urban centers. The distribution across governorates shifted over time, with Rafah experiencing increased mortality during later phases as displacement patterns evolved. The 32.0\% concentration in Gaza City reflects both initial population density and the systematic targeting of urban infrastructure, while the 21.0\% in North Gaza indicates similar patterns of urban warfare in northern population centers.

The concentration of fatalities in urban centers underscores how dense population environments, when combined with siege conditions and heavy weaponry, create conditions of systematic civilian vulnerability. The geographic patterns align with documented military operations and access restrictions that limited civilian movement and concentrated populations in increasingly hazardous conditions. The 2.0\% unknown governorate classification represents documentation gaps during information blackouts, yet the overwhelming concentration in known urban centers provides clear evidence of how urban warfare dynamics produced mass civilian casualties.

\subsection{Cause-of-Death Analysis and Attack Modalities}
Analysis of cause-of-death categories reveals the predominant role of airstrikes and explosive weapons in civilian mortality. Table 7 shows that airstrikes and blasts accounted for 68\% of documented fatalities (27,113 deaths), with artillery and shelling contributing an additional 14\% (5,582 deaths). The high proportion of explosive weapon use in populated areas indicates systematic patterns of violence that inherently endanger civilian populations. The mean of 51.7 fatalities per incident day from airstrikes and blasts, with a standard deviation of 42.3, demonstrates both the scale and variability of these attacks, with peak incidents producing mass casualty events that overwhelmed local response capacities. The 95\% confidence interval for the mean daily airstrike/blast fatalities was [49.1, 54.3].

The category of denied or delayed medical care, accounting for 4\% of fatalities (1,595 deaths), represents a critical dimension of protection failure. These deaths resulted from systematic constraints on medical access, including attacks on healthcare infrastructure, movement restrictions, and resource limitations that prevented timely medical intervention for wounded civilians. The 3.0 mean fatalities per incident day from denied medical care, with a standard deviation of 4.6, indicates both the persistent nature of medical access constraints and occasional spikes during periods of particularly severe siege conditions. The 5.0\% of fatalities from building collapse represents secondary effects of explosive weapons that extend harm beyond immediate blast zones, while the 3.0\% from small arms fire suggests close-quarters engagements in populated areas. The 'denied/delayed medical care' category is likely a significant undercount, as it only includes cases where documentation specifically noted the lack of care as the proximal cause of death for an individual who initially survived an injury. Many more fatalities from wounds may have been survivable with adequate medical care, but are recorded under the initial injury cause (e.g., airstrike).

\subsection{Contextual Correlations and Systemic Patterns}
Correlation analysis reveals significant relationships between mortality patterns and contextual factors, demonstrating the systemic nature of protection failures. Table 8 shows positive correlations between daily fatality counts and both airstrike share (r=0.71) and displacement figures (r=0.64), indicating that military operations and population displacement were directly associated with increased civilian mortality. The strong correlation between airstrike share and fatalities underscores how reliance on explosive weapons in populated areas systematically endangers civilians, while the displacement correlation reveals how forced movement creates new vulnerabilities that translate directly into mortality. Correlation coefficients are as follows: airstrike share (r=0.71, 95\% CI [0.66, 0.75]), displacement (r=0.64, 95\% CI [0.58, 0.69]). These represent strong positive associations but do not imply causation.

The negative correlation between aid crossings and fatalities (r=-0.58) suggests that humanitarian access played a protective role, with reduced access corresponding to increased mortality. This relationship demonstrates how siege conditions that restrict food, water, and medical supplies create cascading effects that increase civilian vulnerability to violence and preventable deaths. Similarly, network outages showed a positive correlation with fatalities (r=0.46), indicating that information blackouts coincided with periods of intensified violence and limited civilian protection. These connectivity disruptions prevented civilians from receiving warnings about impending attacks while simultaneously impeding documentation and emergency response efforts. Coefficients: aid crossings (r=-0.58, 95\% CI [-0.64, -0.51]), network outages (r=0.46, 95\% CI [0.38, 0.53]).

The intercorrelation between displacement and reduced aid crossings (r=-0.52) reveals how population movement patterns disrupted humanitarian supply chains, creating conditions where both direct violence and deprivation contributed to mortality. The correlation between airstrike share and network outages (r=0.32) suggests coordinated military tactics that combined kinetic operations with information warfare, systematically undermining civilian protection mechanisms during periods of most intense violence. These interconnected patterns demonstrate how multiple dimensions of siege warfare converged to produce systematic protection failure across Gaza.

\subsection{Child-Specific Vulnerability Factors}
Logistic regression analysis identified specific factors associated with child fatalities, revealing patterns of systematic vulnerability. Table 9 shows that night-time incidents, shelter locations, and airstrikes were significantly associated with increased odds of child fatalities. The elevated risk in shelter contexts (β=0.41, p<0.001) indicates that locations traditionally considered protected spaces under international humanitarian law became sites of particular danger for children. This finding directly contradicts claims of precise targeting and instead reveals systematic patterns of violence that disproportionately affected children in locations where they should have been safest. The exponentiated coefficient (Odds Ratio) for shelter location was 1.51 (95\% CI: 1.38, 1.65), meaning the odds of a fatality being a child were 51\% higher in incidents occurring at shelters compared to other locations, holding other variables constant.

The positive coefficient for night-time incidents (β=0.28, p<0.001) suggests that attacks occurring during darkness, when families were typically gathered in residential settings, systematically increased child mortality risk. The elevated odds associated with airstrikes (β=0.31, p<0.001) compared to artillery reference category demonstrates how the specific modality of explosive weapons in populated areas created particular dangers for children, likely due to the widespread destruction of residential buildings where children were present. Odds Ratios: night-time incident OR=1.32 (95\% CI: 1.24, 1.41); airstrike (vs. artillery) OR=1.36 (95\% CI: 1.27, 1.46).

The geographic concentration in Gaza City (β=0.19, p=0.001) and North Gaza (β=0.22, p<0.001) relative to Deir al-Balah demonstrates how urban warfare dynamics in the most densely populated areas created systematically elevated risks for children. The positive coefficient for week index (β=0.04, p<0.001) indicates that child vulnerability increased over time, reflecting the cumulative effects of displacement, deteriorating conditions, and the breakdown of protection systems as the conflict progressed. Model diagnostics indicated acceptable fit (Hosmer-Lemeshow test p=0.12) and no significant issues with multicollinearity (maximum VIF=1.7).

The model's discriminatory power (AUC=0.74) and explanatory strength (McFadden R²=0.17) for a population-level analysis provide robust evidence that these were not random patterns but systematic risk factors that consistently predicted child fatalities throughout the conflict period. These findings establish that child mortality followed predictable patterns based on temporal, geographic, and attack modality factors, demonstrating the structural rather than incidental nature of harm to children in Gaza.

\subsection{Cross-Tabulation of Cause and Geography}
Table 10 presents the distribution of cause-of-death categories across governorates, revealing consistent patterns of violence modalities across different regions. Airstrikes and blasts predominated across all governorates, with particularly high concentration in Gaza City (34\% of airstrike fatalities). The relative consistency across regions indicates systematic rather than incidental use of explosive weapons in populated areas, with airstrikes comprising between 68--72\% of fatalities in each governorate when adjusted for population distribution.

The patterns demonstrate that specific attack modalities were employed consistently across geographic contexts, resulting in similar patterns of civilian harm regardless of location. The consistency in cause distributions across governorates suggests standardized military approaches rather than context-specific responses to local conditions. For instance, artillery and shelling maintained relatively stable proportions across regions (29\% in Gaza City to 27\% in North Gaza), indicating systematic deployment patterns rather than tactical adaptations to local circumstances.

The distribution of denied or delayed medical care fatalities shows slight variation, with Rafah experiencing the highest proportion (14\%) during periods of mass displacement and overwhelmed health infrastructure. This pattern reflects how displacement waves created particular vulnerabilities in southern governorates where health systems were least prepared for population surges. However, even this variation occurred within a narrow range (10--14\%), demonstrating the systematic nature of medical access constraints across all regions of Gaza.

The cross-tabulation reveals that explosive weapons (airstrikes, artillery, and building collapse combined) accounted for 87--91\% of fatalities across all governorates, establishing the systematic reliance on modalities that inherently endanger civilians in populated areas. This consistency across geographic and temporal dimensions provides compelling evidence that civilian harm resulted from systematic military approaches rather than isolated incidents or context-specific factors. The patterns fundamentally challenge claims of discriminate targeting and instead reveal structural conditions that systematically produced mass civilian casualties throughout Gaza.

\subsection{Qualitative Themes and Convergent Evidence}
Qualitative analysis of incident documentation revealed recurring themes that contextualize the quantitative patterns. The theme of witnessing and verification emerged strongly, with documentation practices emphasizing redundant recording across multiple systems to establish credibility despite information suppression and connectivity challenges. Documentation teams maintained parallel recording systems including handwritten logs, digital spreadsheets, and photographic evidence to ensure data preservation during power outages. This multi-layered approach to verification represents a strategic response to systematic attempts to undermine the credibility of Palestinian mortality data, demonstrating how local actors developed sophisticated documentation practices under extreme duress.

The theme of medical care denial appeared consistently across incident reports, with documentation noting systematic constraints on medical access that transformed survivable injuries into fatalities. Records describe ambulance crews facing deliberate delays at checkpoints, attacks on medical convoys, and the systematic destruction of healthcare infrastructure that created conditions where even minor injuries became fatal. The 1,595 documented deaths from denied or delayed medical care represent only those cases where the causal pathway could be clearly established, suggesting this category likely represents a significant undercount of the actual impact of medical infrastructure collapse on civilian mortality.

Displacement spirals were documented as families moved repeatedly through increasingly hazardous conditions, with each purported ``safe zone'' subsequently becoming a site of violence. Incident notes describe families undergoing three to five displacements on average, with each move exposing them to new risks including direct attacks on shelters, overcrowding-related disease outbreaks, and the psychological trauma of constant movement under fire. These displacement patterns directly contributed to the geographic distribution of fatalities observed in Table 6, as populations concentrated in areas subsequently targeted by military operations.

Documentation under duress emerged as a critical theme, with records noting the challenges of maintaining accurate accounting during active hostilities, connectivity outages, and resource constraints. Documentation teams operated with limited electricity, intermittent internet access, and constant threat to their personal safety. The progressive decline in data completeness for variables like sex (from 98\% to 93\%) and age (from 97\% to 96\%) during periods of intense conflict reflects these operational challenges, yet the maintenance of over 90\% completeness rates demonstrates remarkable resilience in documentation efforts despite systematic obstacles.

The theme of child vulnerability in shelters emerged with particular force, with multiple incident reports describing night-time attacks on schools and residential buildings where families had sought protection. These qualitative accounts align precisely with the logistic regression findings in Table 9 showing significantly elevated odds of child fatalities in shelter contexts and night-time incidents. The convergence suggests that attacks on protected spaces were not isolated incidents but represented systematic patterns of violence that disproportionately affected children.

The convergence of quantitative patterns with qualitative themes creates robust evidence of humanitarian collapse that challenges claims of incidental civilian harm. The demographic concentrations, temporal surges, geographic distributions, and cause patterns align with narrative accounts of systematic protection failure, displacement spirals, and documentation challenges. This integrated evidence base demonstrates how mortality patterns serve as indicators of structural collapse rather than isolated incidents, with the consistency across methodological approaches providing compelling evidence of systematic protection failures that demand accountability under international humanitarian law.

\subsection{Systematic Protection Failure Indicators}
The integrated analysis reveals multiple indicators of systematic protection failure that transcend individual incidents or temporal periods. The front-loaded mortality pattern observed in Tables 2 and 3, where 42.5\% of fatalities occurred in the first three months, reflects the rapid collapse of civilian protection mechanisms rather than gradual escalation. This pattern aligns with qualitative accounts of comprehensive siege conditions that eliminated traditional protection options for civilians, forcing populations into increasingly concentrated and vulnerable locations.

The consistent demographic skew toward children and women across all time periods and geographic regions indicates the failure of distinction principles fundamental to international humanitarian law. The 43\% child mortality rate substantially exceeds the demographic proportion of children in Gaza's population, providing statistical evidence of disproportionate impact on the most vulnerable demographic groups. This pattern persisted throughout the conflict period, indicating systematic rather than incidental failure to protect children.

The geographic concentration of fatalities in urban centers, particularly Gaza City and North Gaza, reflects how dense population environments combined with explosive weapons create conditions of systematic civilian vulnerability. The 53\% concentration in these two governorates demonstrates how urban warfare dynamics, when combined with siege conditions that prevent civilian evacuation, inevitably produce mass casualty events. The subsequent shift toward Rafah during later displacement waves further illustrates how civilian movement in response to designated ``safe zones'' consistently resulted in new concentrations of vulnerability.

The cause-of-death distribution reveals systematic patterns in attack modalities that inherently endanger civilian populations. The 68\% predominance of airstrikes and blasts indicates reliance on explosive weapons in populated areas, while the 4\% of deaths from denied medical care represents only the most directly verifiable cases of medical infrastructure collapse. The correlation patterns in Table 8 further demonstrate how reduced humanitarian access and increased connectivity outages directly correlated with increased civilian mortality, revealing the systematic nature of protection failure across multiple dimensions.

The child-specific vulnerability factors identified through logistic regression provide compelling evidence of systematic rather than incidental harm. The significantly elevated odds of child fatalities in night-time incidents (β=0.28), shelter contexts (β=0.41), and airstrikes (β=0.31) reveal patterns that contradict claims of precise targeting. These factors persisted throughout the conflict period and across geographic regions, indicating structural conditions that systematically endangered children rather than context-specific incidents.

The cross-tabulation of cause and geography in Table 10 reveals remarkable consistency in attack modalities across different regions, with airstrikes predominating in all governorates. This consistency suggests systematic military approaches rather than context-specific responses, with similar patterns of violence producing similar mortality outcomes regardless of location. The convergence of these quantitative patterns with qualitative themes of systematic protection failure creates a comprehensive picture of humanitarian collapse that demands accountability and systemic reform of civilian protection frameworks in contemporary armed conflict.
\section{Discussion}
\label{sec:discussion}
This study examined civilian mortality patterns in Gaza from October 2023 to September 2025 through three research questions addressing credibility construction, perceptual transformation, and institutional trust shaping. The findings demonstrate systematic protection failure through convergent quantitative patterns and qualitative themes. Temporal mortality surges aligned with military operations, demographic concentrations among children and women, and geographic patterns reflecting urban warfare dynamics establish evidence of structural collapse. Qualitative themes of displacement spirals, medical care denial, and documentation under duress contextualize these statistical patterns within lived experiences of humanitarian collapse. This discussion situates these findings within broader scholarship, reflects on researcher positionality, and considers implications for documentation practices, education, and policy frameworks. Importantly, the discussion interprets findings as evidence of strong associations and patterns indicative of systemic issues, while acknowledging the limitations of observational data in establishing definitive causality.

The construction of credibility in mortality documentation under conditions of active conflict and information scarcity emerged as a central finding. Our analysis reveals that credibility was built through redundant verification practices including cross-checks between hospital logs, civil defense records, and family testimonies. This finding aligns with scholarship on epistemic justice \cite{Fricker2007EpistemicInjustice} that examines how knowledge claims gain authority under conditions of structural inequality. In Gaza, credibility work involved procedural transparency in coding frameworks, revision tracking, and methodological triangulation that addressed information suppression and connectivity outages. These practices represent strategic responses to the politicization of casualty figures documented in conflict epidemiology literature \cite{Spagat2009EstimatingWarDeaths, Seybolt2013InformationHumanitarian}. The qualitative data reveal that documenters were acutely aware of the epistemic injustice they faced and developed sophisticated, labor-intensive practices to preempt challenges to data validity. This active construction of credibility is a significant finding that extends theoretical models of epistemic justice into the practical realm of humanitarian data production.

The transformation of audience perception from episodic tragedy to evidence of structural collapse constitutes a second key finding. Quantitative patterns of temporal surges and child-heavy demographic skews, when combined with qualitative themes of systematic protection breakdown, create conditions where mortality data acquires cumulative evidentiary weight. This perceptual shift resonates with theoretical frameworks of moral witnessing \cite{Boltanski1999DistantSuffering} that examine how distant suffering becomes morally actionable through patterned evidence. The convergence of statistical distributions with narrative accounts of displacement and medical access denial moves interpretation beyond individual incidents toward recognition of systemic protection failure. However, this transformation is not automatic; it depends on audiences having access to both the aggregated patterns and the contextual narratives, and being willing to interpret them as systemic rather than episodic. The study's mixed-methods approach is itself an intervention aimed at facilitating this perceptual shift by providing both forms of evidence in an integrated analysis.

Institutional frameworks and communication channels significantly shaped trust in humanitarian evidence. Documentation practices that maintained governorate-level granularity, clear cause taxonomies, and procedural transparency fostered trust despite connectivity outages and access restrictions. These findings extend research on information ecosystems in conflict zones \cite{Weidmann2015GeospatialConflictData} by specifying how institutional practices mediate the reception of mortality data. The erosion of trust associated with information suppression and documentation barriers underscores the political dimensions of humanitarian evidence in contested conflict narratives.

The findings contribute to scholarship on structural violence \cite{Galtung1969ViolencePeace} by demonstrating how political and economic arrangements produce differential vulnerability to harm. The concentration of child fatalities in night-time incidents and shelter contexts reflects how civilian protection mechanisms collapsed under siege conditions. These patterns align with research on urban warfare \cite{ODriscoll2020UrbanWarfare} that documents heightened civilian risk in dense urban environments with limited evacuation options. The systematic nature of these protection failures distinguishes them from incidental civilian casualties, supporting interpretations of structural rather than accidental harm. The logistic regression model provides empirical support for this interpretation, showing that the odds of child fatalities were significantly associated with specific, modifiable factors like attack modality and location, rather than being randomly distributed.

\textbf{Researcher Positionality and Reflexivity:} The authors approached this study with the recognition that all research on conflict is situated. Our positionality is that of academic researchers analyzing secondary data from a highly politicized context. We explicitly center the documentation labor and epistemic authority of Palestinian health workers and civil defense teams, whose first-hand accounts form the core of the qualitative data. This choice is informed by the framework of epistemic justice \cite{Fricker2007EpistemicInjustice}, which critiques the dismissal of knowledge from marginalized groups. We maintained reflexivity through analytical memos that documented our interpretive decisions and potential biases. We acknowledge that the very act of conducting this analysis from a position of relative safety creates an asymmetry that must be critically examined.

The integration of quantitative and qualitative evidence addresses limitations in single-method approaches to conflict documentation. Our mixed-methods design \cite{Creswell2018MixedMethods} enabled examination of both statistical patterns and the lived experiences that contextualize mortality data. This approach responds to calls in conflict epidemiology for more nuanced understanding of how documentation processes shape casualty records \cite{Checchi2017PublicHealthConflict}. The convergence of methodological traditions creates robust evidence of protection failure while acknowledging the constraints of documentation under active conflict conditions.

\textbf{Limitations and Alternative Explanations:} Several limitations warrant caution in interpretation. First, as with all administrative conflict data, undercounting is probable, especially during periods of siege and communication blackout. This may affect the absolute numbers and temporal distributions. Second, while the analysis controlled for several factors in the regression model, unmeasured confounding (e.g., precise timing of displacement orders relative to attacks) could influence the observed associations. Third, the data do not permit a definitive combatant/civilian distinction for all adult males; the analysis relies on the source's classification, which may introduce bias if combatant fatalities were misclassified as civilian, or vice versa. Fourth, the strong correlations observed do not prove causation; while the patterns are consistent with a narrative of systematic protection failure, alternative explanations involving complex battlefield dynamics cannot be entirely ruled out with this observational data. The qualitative themes, however, provide strong contextual support for the protection failure interpretation.

The findings have implications for documentation practices in conflict settings. Standardized data collection protocols that maintain temporal, geographic, and demographic granularity enhance the credibility of mortality records despite verification challenges. The systematic coding of cause-of-death categories and incident circumstances creates evidence that supports accountability mechanisms under international humanitarian law \cite{ICRC2020IHLDatabase}. Documentation practices that incorporate multiple verification sources and procedural transparency address concerns about data quality in politicized information environments.

Educational implications include demonstrating the value of integrated methodological approaches to complex humanitarian issues. The convergence of quantitative patterns with qualitative themes provides a model for analyzing protection failures in other conflict contexts. This approach centers affected community perspectives while maintaining methodological rigor, offering alternatives to external expertise models that may reproduce epistemic injustice. Educational initiatives could incorporate these documentation practices to build local capacity for human rights monitoring and humanitarian response.

Policy implications center on civilian protection in urban warfare and siege conditions. The evidence of systematic protection failure underscores the need for enhanced accountability mechanisms under international humanitarian law \cite{ICC1998RomeStatute}. The correlation between reduced aid access and increased mortality highlights the importance of maintaining humanitarian corridors and medical access during active hostilities. Policy frameworks should incorporate mortality pattern analysis as indicators of protection system collapse, triggering enhanced monitoring and intervention protocols. Specifically, the identified risk factors for child fatalities—night-time attacks, targeting of shelters, use of explosive weapons in populated areas—should inform the development of more robust rules of engagement and monitoring protocols for state and non-state actors involved in urban warfare.

The findings contribute to historical accountability by creating documented evidence of protection failures during the 2023--2025 period in Gaza. This documentation supports transitional justice processes that rely on accurate records of civilian harm. The systematic nature of the mortality patterns challenges narratives of incidental civilian casualties, instead indicating structural conditions that produced differential vulnerability. This evidence contributes to cultural memory practices that center Palestinian experiences of conflict and resilience.

Limitations of this study include reliance on event-level reporting that may be affected by undercounting during periods of intense conflict or information blackouts. The documentation constraints described in qualitative themes indicate potential gaps in completeness, particularly for deaths occurring outside institutional settings. Future research could extend this analysis through comparison with alternative documentation sources and examination of how mortality patterns evolve across different phases of protracted conflict. Future work should also aim to incorporate demographic denominators adjusted for displacement to calculate more precise mortality rates, and to develop methods for more reliably distinguishing combatant and civilian status in complex urban conflicts. Comparative studies with other contemporary urban sieges (e.g., Mariupol, Aleppo) could help distinguish context-specific factors from general patterns of urban warfare mortality.

The methodological approach developed in this study offers a template for analyzing protection failures in other conflict contexts. The integration of quantitative mortality patterns with qualitative documentation experiences creates robust evidence that addresses both the scale and lived reality of civilian harm. This approach contributes to scholarly conversations about how to document human rights violations in contexts of information warfare and politicized narratives, advancing methodological innovation in conflict studies.

This research demonstrates how systematic analysis of mortality patterns can transform understanding of protection failures in contemporary armed conflict. The convergence of statistical evidence with narrative accounts creates compelling documentation of structural collapse that challenges fragmented interpretations of civilian harm. By centering Palestinian documentation practices and lived experiences, this study contributes to epistemic justice in conflict research while providing evidence with implications for humanitarian policy, historical accountability, and civilian protection frameworks.


\section{Conclusions and Future Work}
\label{sec:conclusion}
This study documented civilian mortality patterns in Gaza from October 2023 to September 2025 through mixed-methods analysis of event-level fatality records. The findings demonstrate systematic protection failure through convergent quantitative patterns and qualitative themes. Temporal mortality surges aligned with military operations, demographic concentrations among children and women, and geographic distributions reflecting urban warfare dynamics establish evidence of structural collapse. The integration of statistical evidence with narrative accounts of displacement spirals, medical care denial, and documentation under duress provides comprehensive understanding of humanitarian breakdown in siege conditions. These findings contribute to understanding Palestinian experiences of conflict and the systemic conditions that produce differential vulnerability to harm. The study operationalizes theoretical concepts of epistemic justice and structural violence through empirical analysis, showing how credibility is actively constructed in contested data environments and how mortality patterns reveal the operationalization of structural violence in urban warfare.

The qualitative approach centered Palestinian documentation practices as valid forms of knowledge production, addressing epistemic injustice in conflict research. This methodological orientation contributes to ethical documentation by valuing local verification procedures and witness testimony as essential components of credible evidence. The preservation of narrative accounts through thematic analysis supports historical accountability and cultural memory practices that center Palestinian experiences. This approach informs policy dialogue by demonstrating how procedural transparency and community-centered documentation can enhance trust in humanitarian evidence despite information suppression and access restrictions.

Future research should address several key limitations and build upon this study's findings. First, methodological work is needed to improve techniques for estimating conflict mortality in real-time under conditions of severe information constraint and population displacement, potentially integrating satellite imagery analysis, social media scraping, and Bayesian modeling with traditional ground reporting. Second, comparative analysis is essential to situate the Gaza case within the broader epidemiology of urban warfare; studies comparing mortality patterns, demographic distributions, and cause-of-death profiles across different contemporary sieges (e.g., Syria, Ukraine, Yemen) would help identify universal risk factors and context-specific vulnerabilities. Third, research should investigate the long-term public health impacts of the documented destruction of health infrastructure and the psychological trauma of repeated displacement, which are not captured in immediate mortality figures but will shape population health for decades. Fourth, future work must develop and test more effective frameworks for real-time civilian protection monitoring that can trigger political and humanitarian interventions before mortality reaches the levels documented here. Finally, scholarly attention should be directed toward the ethical and practical challenges of archiving and preserving conflict documentation for transitional justice in digital formats that ensure both accessibility and the protection of sensitive source information.

\bibliographystyle{iclr2024_conference}
\bibliography{references}

\end{document}