OIM - Dynamic Estimates of Displacement in Disaster Regions de 2025

DYNAMIC ESTIMATES OF DISPLACEMENT IN DISASTER REGIONS

Geographic Data Science Lab

DYNAMIC ESTIMATES

OF DISPLACEMENT IN DISASTER REGIONS A policy-driven framework triangulating dataThe opinions expressed in this publication are those of the authors and do not necessarily reflect the views of the International Organization for Migration (IOM). The designations employed and presentation of material throughout the report do not imply the expression of any opinion whatsoever on the part of IOM concerning the legal status of any country, territory, city or area, or of its authorities, or concerning its frontiers or boundaries. IOM is committed to the principle that humane and orderly migration benefits migrants and society. As an intergovernmental organization, IOM acts with its partners in the international community to assist in meeting the operational challenges of migration; advance understanding of migration issues; encourage social and economic development through migration; and uphold the human dignity and well-being of migrants.

Publisher: International Organization for Migration 17 route des Morillons P.O. Box 17 1211 Geneva 19

Switzerland T el.: +41 22 717 9111

Fax: +41 22 798 6150

Email: hq@iom.int Website: www.iom.int This publication was issued without formal editing by IOM.

Cover image: Lea Riggi (2025) Visual credits: Brian McDonald (2022) Required citation: Pietrostefani, E., M. Mason, R. Iradukunda, H. Tran-Jones, I. Loktieva and F. Rowe (2025).

Dynamic Estimates of Displacement in Disaster Regions: A Policy-Driven Framework Triangulating Data. IOM, Geneva. ISBN 978-92-9278-079-1 (PDF) DOI: https://doi.org/10.48550/arXiv.2511.01955 © IOM 2025 Some rights reserved. This work is made available under the Creative Commons Attribution-NonCommercial-NoDerivs 3.0 IGO License (CC BY-NC-ND 3.0 IGO).

DOI: https://doi.org/10.48550/arXiv.2511.01955

© IOM 2025 Some rights reserved. This work is made available under the Creative Commons Attribution-NonCommercial-NoDerivs 3.0 IGO License (CC BY-NC-ND 3.0 IGO). For further specifications please see the Copyright and T erms of Use. This publication should not be used, published or redistributed for purposes primarily intended for or directed towards commercial advantage or monetary compensation, with the exception of educational purposes e.g. to be included in textbooks.

Permissions: Requests for commercial use or further rights and licensing should be submitted to publications@iom.int. https://creativecommons.org/licenses/by-nc-nd/3.0/igo/legalcode

PUB2025/081/RDYNAMIC ESTIMATES OF

DISPLACEMENT IN DISASTER REGIONS A policy-driven framework triangulating data Geographic Data Science LabDYNAMIC ESTIMATES OF DISPLACEMENT IN DISASTER REGIONS FOREWORD The global challenge of internal displacement, exacerbated by conflict, climate-induced natural hazards and disasters, requires innovative and collaborative approaches to ensure effective responses. At the end of 2024, an estimated 123.2 million people were forcibly displaced (UNHCR, 2025a). This staggering figure includes 3.8 million people uprooted within Ukraine due to war, 20.8 million internally displaced across the Horn of Africa by a combination of drought and violence and millions more affected annually in South and South-East Asia, where disasters such as cyclones triggered at least 1.8 million displacements in 2024 alone (ibid.). These estimates are a stark reminder of the need for timely, accurate and spatially detailed data and accessible ways to process it to inform humanitarian action and policy. This publication represents a step forward in addressing the need for more integrated and adaptive displacement data systems. By triangulating traditional data sources with cutting-edge digital trace data – such as mobile phone GPS and social media data – it highlights key considerations for effectively combining these approaches in humanitarian contexts. The report explores these

This publication represents a step forward in addressing the need for more integrated and adaptive displacement data systems. By triangulating traditional data sources with cutting-edge digital trace data – such as mobile phone GPS and social media data – it highlights key considerations for effectively combining these approaches in humanitarian contexts. The report explores these insights, through focused case studies, offer practical guidance for integrating diverse data streams to support more timely and informed interventions. The push for data innovation is particularly timely as funding cuts have reduced the humanitarian sector traditional data capabilities. We demonstrate the power of triangulating data using two study cases. First, we use the first year of the war in Ukraine, which escalated in February 2022, and show the potential of data triangulation to generate actionable insights in complex and rapidly evolving settings that demand timely data, policy decisions and humanitarian responses. Second, we focus on the Pakistan floods of August 2022 to test the broader applicability of our approach – shifting the context from conflict to a climate-induced disaster setting. T ogether, the Ukraine and Pakistan case studies illustrate the versatility and effectiveness of data triangulation in supporting timely, data-driven responses across different types of crises. This project has been made possible through the collaborative efforts of the Geographic Data Science Lab at the University of Liverpool and the International Organization for Migration’s (IOM) Displacement Tracking Matrix. This work has been developed in close partnership with operational agencies, ensuring its relevance to real-world needs and its potential to inform evidence-based decision-making. The University of Liverpool and IOM Displacement Tracking Matrix recognize the value of data triangulation in enabling partners, including local authorities, to enhance coordination and programming through robust, evidence-based strategies. We welcome constructive conversations on how this approach can be further refined and applied, and we look forward to collaborating with stakeholders to promote sustainable, rights-based solutions for displaced populations worldwide. Professor Tim Jones Vice-Chancellor University of Liverpool Laura Nistri Global Displacement Tracking

conversations on how this approach can be further refined and applied, and we look forward to collaborating with stakeholders to promote sustainable, rights-based solutions for displaced populations worldwide. Professor Tim Jones Vice-Chancellor University of Liverpool Laura Nistri Global Displacement Tracking Matrix (DTM) Coordinator IOM iiDYNAMIC ESTIMATES OF DISPLACEMENT IN DISASTER REGIONS iii This report is the result of a collaborative effort between the Geographic Data Science Lab at the University of Liverpool and the International Organization for Migration’s Displacement Tracking Matrix (DTM). The authors wish to express their sincere gratitude to colleagues at IOM Ukraine’s Data and Analytics Unit and IOM DTM London. Special thanks to Douglas Leasure, Andrea Aparicio Castro and Edith Darin of the University of Oxford for their valuable contributions to the project and its outcomes. Additionally, sincere thanks are extended to those who participated in the project’s three workshops and hackathon, the names of which are listed below. The content of this report would not have been possible without the generous contribution of participants; their comments and insights aided greatly in shaping the findings and conclusions of this report. Further thanks are also extended to organizations that helped provide data and technical support during the project’s hackathon, namely Snowflake and Direct Relief, both of whom the event and its success would not have been possible without. We gratefully acknowledge Ellen Van de Weghe (IOM DTM Pakistan) and Fawad Qureshi (Snowflake) for their valuable contributions as judges during the hackathon. This report was made possible through the support of the Policy Support Fund at the University of Liverpool, funded by Research England. The research aligns with the strategic objectives of UK Research and Innovation (UKRI) by enhancing social and economic resilience through the development of tools that help communities and policymakers more effectively manage displacement challenges. It also contributes to national security and risk awareness by improving the capacity to anticipate and respond to displacement crises. Furthermore, the

the strategic objectives of UK Research and Innovation (UKRI) by enhancing social and economic resilience through the development of tools that help communities and policymakers more effectively manage displacement challenges. It also contributes to national security and risk awareness by improving the capacity to anticipate and respond to displacement crises. Furthermore, the project promotes the rights, dignity, and long-term well-being of displaced populations by advancing data-driven strategies that support sustainable integration or return.

CREDITS AND ACKNOWLEDGEMENTS

Workshop attendees: Andrea Aparicio

Mohamed Bakr Luong Bang Tran Adam Bekele Jos Berens Dominik Bursy Carmen Cabrera Dan Caspersz Alexander Chilton Flora Chu Franziska Clevers Laura Coskun Rachel Cribbin Edith Darin Adham Enaya Marianthe Evangelidis Gabriele Filomena Emma Goatman Joseph Goodall Ali Guenduez Shannon Hayes T ony Hoad Linh Hoang Thuy Kate Hodkinson Mike Johnson Dirk Jung Damien Jusselme Varun Khandelwal Cara Kielwein Doug Leasure Nando Lewis Lisa Lim Ah Ken William Lumala Alice Marshall Alex McCarthy Brian McDonald Andrea Nasuto Euan Newlands Abdul Samad Omari Oluwatosin Orenaike Benjamin Pfau Elena Philipova Joshua Phillippo-Holmes Fawad Qureshi Lorenzo Sileci Joseph Slowey Yaroslav Smirnov Ellen Van de Weghe Vivianne Van der Vorst Ana Varela Varela Belinda Volans Huan Wang Nick Ward Scott White

Michael ZihanzuDYNAMIC ESTIMATES OF DISPLACEMENT IN DISASTER REGIONS

CONTENTS Foreword Credits and acknowledgements List of figures and tables

1. Data triangulation in crisis response: Bridging traditional and digital sources

2. Key lessons on triangulating data for displacement estimates in disaster contexts 2.1 Displacement data systems: current models and their limitations

CONTENTS Foreword Credits and acknowledgements List of figures and tables

1. Data triangulation in crisis response: Bridging traditional and digital sources

2. Key lessons on triangulating data for displacement estimates in disaster contexts 2.1 Displacement data systems: current models and their limitations 2.2 Turning digital trace data into opportunities 2.3 From insight to action: recommendations for integrating digital trace data in humanitarian response

3. Digital trace data in action: The case of Ukraine 3.1 Evolving methodologies improve accuracy of displacement data 3.2 Digital trace data: enhancing crisis monitoring in real time 3.3 Displacement trends 3.4 Comparing returnees

4. Insights from Pakistan: Expanding the use of triangulated data for humanitarian response 4.1 Pakistan floods 2022: context and datasets 4.2 Key conclusions from the hackathon 4.3 Problem statements 4.4 Hackathon results: problem-solving in practice

Conclusion References ii iii v 1 3 4 9

13 18 18 19 22 28

31 32 32 34 35 42 43

ivDYNAMIC ESTIMATES OF DISPLACEMENT IN DISASTER REGIONS

LIST OF FIGURES AND TABLES

Figure 1. Figure 2. Figure 3. Figure 4. Figure 5. Figure 6. Figure 7. Figure 8. Figure 9. Figure 10. Figure 11. Figure 12. Figure 13.

Table 1. Table 2. Table 3.

Table 4. Table 5. Table 6. Table 7.

An overview of displacement data systems, data-collection methods and limitations High resolution, large-scale coverage and near real-time availability Hard-to-reach location access Actionable insights through geographic data science Recommendations for integrating digital trace data in humanitarian response IDP national comparison IDP oblast comparison

Table 6. Table 7.

An overview of displacement data systems, data-collection methods and limitations High resolution, large-scale coverage and near real-time availability Hard-to-reach location access Actionable insights through geographic data science Recommendations for integrating digital trace data in humanitarian response IDP national comparison IDP oblast comparison T emporal mapping of IDP patterns in select months of 2022 at oblast level Snapshot mapping of GPS MD at raion-level 2022 Returnee oblast comparison February-August 2022 Bivariate map of displacement and flooding levels Maps of decrease in daily Facebook population and map of deprivation levels in Pakistan Screenshot of an interactive dashboard made by Group 4 during the hackathon

Traditional data and digital nontraditional data definitions Data comparison of primary datasets discussed Comparison of digital nontraditional data sources for crisis response and displacement analysis Definitions and use of displacement-related terms across data sources IDP national comparison Key conclusions from the hackathon Problem statements used by groups during the hackathon

4 9 10 11 13 23 24 26 27 29 36 37 40

2 6

12 21 23 33 34 v1 Chapter 1 – Data triangulation in crisis response: Bridging traditional and digital sources

DYNAMIC ESTIMATES OF DISPLACEMENT IN DISASTER REGIONS

DATA TRIANGULATION IN CRISIS RESPONSE:

BRIDGING TRADITIONAL AND DIGITAL SOURCES

CHAPTER 1

Internal displacement remains a critical global issue. An unprecedented 83.4 million people were living in internal displacement at the end of 2024, according to the newly released Global Report on Internal Displacement 2025 (IDMC, 2025a). This scale of displacement underscores the urgent need for innovative, data-driven approaches to track and understand population movements. Traditional data systems provide vital information for those responding to humanitarian crises. However, as human mobility patterns become increasingly complex, the need for reliable, timely and spatially detailed data to inform the development of policy and humanitarian response is becoming increasingly acute. Traditional data

track and understand population movements. Traditional data systems provide vital information for those responding to humanitarian crises. However, as human mobility patterns become increasingly complex, the need for reliable, timely and spatially detailed data to inform the development of policy and humanitarian response is becoming increasingly acute. Traditional data streams are often not well-equipped to meet these needs (IOM, 2018b). This report, a collaboration between the Geographic Data Science Lab at the University of Liverpool and the International Organization for Migration’s (IOM) Displacement Tracking Matrix (DTM), examines how traditional data sources can be effectively integrated with emerging digital trace data – such as mobile phone GPS and social media activity – to enhance the monitoring of displacement in humanitarian settings. By leveraging these diverse data streams, the report demonstrates how triangulation can improve the precision and reliability of displacement estimates. This approach is informed by lessons learned from recent crises, particularly the escalation of the war in Ukraine and the 2022 floods in Pakistan. Building on the innovative application of digital trace data and the development of robust data infrastructure in these contexts, the report outlines a scalable framework for triangulating data across a broader range of crises, including those triggered by natural hazards and public health emergencies. By leveraging real-time, high-resolution data sources we aim to create a more responsive, scalable and accurate system for understanding displacement, allocating resources and evaluating interventions. This report presents the findings of a structured pilot effort designed to test this approach in two defined contexts, leading to a set of key conclusions that inform the future design of data triangulation systems for humanitarian action – followed by detailed technical insights and recommendations. The report is split into three sections. First, we examine the current data landscape around displacement in conflict and disaster contexts. We provide an overview of the gaps in traditional data streams available during displacement events, and the limits these put on effective humanitarian action. Traditional data streams refer to established sources of information used in humanitarian

examine the current data landscape around displacement in conflict and disaster contexts. We provide an overview of the gaps in traditional data streams available during displacement events, and the limits these put on effective humanitarian action. Traditional data streams refer to established sources of information used in humanitarian contexts, such as surveys, administrative records and key informant interviews. We then explore the potential for digital nontraditional data to complement traditional data, provide additional insight and enhance efforts to respond to conflict and disaster events. Digital nontraditional data, or digital trace data, refers to information generated passively or actively through digital platforms and devices, such as mobile phone GPS signals, social media activity, satellite imagery and online transactions.2 Chapter 1 – Data triangulation in crisis response: Bridging traditional and digital sources DYNAMIC ESTIMATES OF DISPLACEMENT IN DISASTER REGIONS Table 1: Traditional data and digital nontraditional data definitions In its second section, the report focuses on the war in Ukraine, which escalated in February 2022. Statistical indicators derived from digital trace data sources – including displacement rates and return rates – are presented. These indicators are benchmarked against IOM data on Ukraine to assess their validity and enhance confidence in the use of digital trace data for monitoring displacement dynamics. This section provides links to technical documentation for all datasets, as well as publicly available code, to ensure the project’s methodology is transparent and replicable in other conflict and disaster contexts. The third section of the report builds on the innovative data triangulation methods used in Ukraine to quantify conflict-related displacement, applying them to the 2022 Pakistan floods and showcasing how these approaches were tested during a hackathon to evaluate their broader applicability across crises emergencies. The report concludes by summarizing the value that digital nontraditional data can add and its significant benefit to humanitarian response in displacement events. This work bridges a critical gap in the humanitarian sector by advancing innovative, data-driven approaches to displacement analysis. Its impact will

The report concludes by summarizing the value that digital nontraditional data can add and its significant benefit to humanitarian response in displacement events. This work bridges a critical gap in the humanitarian sector by advancing innovative, data-driven approaches to displacement analysis. Its impact will extend to strengthening humanitarian response, improving policy formulation and fostering resilient, rights-based solutions for displaced populations worldwide. Traditional data Traditional data streams refer to established sources of information used in humanitarian contexts, such as surveys, administrative records and key informant interviews. Digital nontraditional data Digital nontraditional data, or digital trace data, refers to information generated passively or actively through digital platforms and devices, such as mobile phone GPS signals, social media activity, satellite imagery and online transactions.3 DYNAMIC ESTIMATES OF DISPLACEMENT IN DISASTER REGIONS Chapter 2 – Key lessons on triangulating data for displacement estimates in disaster contexts

KEY LESSONS ON TRIANGULATING DATA FOR

DISPLACEMENT ESTIMATES IN DISASTER CONTEXTS

CHAPTER 2

This section provides an overview of the potential for triangulating multiple data sources to estimate population displacement in disaster settings, whether caused by conflict or climate-induced natural hazards. The insights presented here are informed by a series of expert workshops held in early 2025, which brought together stakeholders from government, academia, international organizations and the private sector.

The workshops included: Shaping a Policy-driven Framework for Displacement Estimates 13 February 2025, Online

Engaging Government Stakeholders for Practical Insights 25 March 2025, London Presenting the Framework and Exploring Broader Contexts 21–22 May 2025, Berlin Participants included representatives from the International Organization for Migration’s (IOM) Displacement Tracking Matrix (DTM), the IOM Global Migration Data Analysis Centre (GMDAC), the IOM Ukraine Data and Analytics Unit, the UK Foreign, Commonwealth & Development Office (FCDO), the Centre for Humanitarian data at United Nations Office for the Coordination

Displacement Tracking Matrix (DTM), the IOM Global Migration Data Analysis Centre (GMDAC), the IOM Ukraine Data and Analytics Unit, the UK Foreign, Commonwealth & Development Office (FCDO), the Centre for Humanitarian data at United Nations Office for the Coordination of Humanitarian Affairs (OCHA), the London Stock Exchange Group (LSEG), Snowflake and several non-governmental organizations (NGOs), including Direct Relief, Aid Ukraine and Operations for Change. Academic institutions represented included the University of Oxford, the London School of Economics (LSE), the University of Exeter and the University of Liverpool (UoL). The workshops highlighted the critical role of multisource data triangulation in improving the accuracy and responsiveness of displacement estimates. This approach is particularly valuable in rapidly evolving crises, where timely, high-resolution information can significantly enhance the targeting and effectiveness of humanitarian interventions. Digital nontraditional data sources – such as mobile phone records, social media and satellite imagery – offer high spatial and temporal granularity, while traditional survey methods contribute essential contextual understanding and validation. When used in combination, these traditional and nontraditional data sources provide a more comprehensive and reliable foundation for informed decision-making in humanitarian response planning.4 DYNAMIC ESTIMATES OF DISPLACEMENT IN DISASTER REGIONS Chapter 2 – Key lessons on triangulating data for displacement estimates in disaster contexts 2.1 DISPLACEMENT DATA SYSTEMS: CURRENT MODELS AND THEIR LIMITATIONS Figure 1. An overview of displacement data systems, data-collection methods and limitations Internal displacement has reached a record high. According to the Internal Displacement Monitoring Centre (IDMC), 83.4 million people were living in internal displacement worldwide at the end of 2024 – the highest number ever recorded and more than double the figure from six years earlier (IDMC, 2025a). The scale of internal displacement has severe consequences, pushing millions into precarious living conditions and undermining access to essential services such as health care, education and livelihoods. It also

recorded and more than double the figure from six years earlier (IDMC, 2025a). The scale of internal displacement has severe consequences, pushing millions into precarious living conditions and undermining access to essential services such as health care, education and livelihoods. It also poses a significant barrier to achieving sustainable development. As the crisis of internal displacement has grown, the need for high-quality data has become more important than ever. Data systems providing accurate and timely information are crucial for quantifying the scale of displacement, the allocation of resources in disaster response, the monitoring of crisis events over time and for the evaluation of the effectiveness of interventions (see Figure 1).

Source: Elaborated by the authors. “Current data-collection methods” sourced from the International Organization for Migration’s Displacement Tracking Matrix (DTM) available at: https://dtm.iom. int/about/infosheets .5

DYNAMIC ESTIMATES OF DISPLACEMENT IN DISASTER REGIONS Chapter 2 – Key lessons on triangulating data for displacement estimates in disaster contexts Table 2 presents the five primary datasets discussed in this report and highlights their respective strengths and trade-offs. These include both traditional and nontraditional data sources used to estimate internally displaced persons (IDPs) in humanitarian contexts. IOM’s General Population Surveys (IOM RDD) in Ukraine are repeated cross-sectional sample surveys designed to provide reliable, ground-level insights into displacement and mobility trends. Each survey round collects responses from an independent sample using a consistent set of questions, with data gathered through Random Digit Dialling (RDD) and Computer-Assisted T elephone Interviews (CATI)

(IOM, 2024a).

IOM DTM’s Community Needs Identification (CNI PK ) was established following the widespread flooding in Pakistan in 2022. It provides information on the number of displaced persons, along with the multisectoral needs of communities, at the settlement and village level (IOM, 2023). CNI PK is implemented through a

(CNI PK ) was established following the widespread flooding in Pakistan in 2022. It provides information on the number of displaced persons, along with the multisectoral needs of communities, at the settlement and village level (IOM, 2023). CNI PK is implemented through a key informant survey, in which IOM enumerators interview community leaders or representative groups to gather data and estimate displacement figures. The assessment was conducted in multiple rounds, targeting specific settlements and villages during each phase. In contrast, the digital trace data sources presented in Table 2 – namely GPS Phone Data (GPS MD), Meta’s Marketing Platform API (Meta MAPI) and Meta’s Data for Good products (Meta DfG) – are all derived from private sector platforms. These datasets capture either geo-located observations of mobile devices at the coordinate level (GPS MD) or aggregated information on social media users across spatial units such as grid cells or administrative boundaries. Existing data systems are wide-ranging and rely on a combination of collection techniques to produce comprehensive insights (IOM, 2023b) (see Figure 1). In this report we consider both displacement estimates derived from key informant interviews (CNI PK) and a representative, repeated crosssectional sample survey IOM RDD) ( Table 2). Key informant interviews remain one of the mostcommonly used approaches for data collection in disaster contexts, as demonstrated during the 2022 floods in Pakistan (SNI PK). In this method, trained enumerators gather information from local officials or community leaders, enabling estimates of displaced populations and the identification of multisectoral community needs. This approach allows humanitarian organizations to collect data in a relatively timeand resource-efficient manner, often generating timely operational insights. However, each phase of data collection often targets only selected settlements and villages, which limits the generalizability of findings to affected areas. The method also requires significant human resources and can encounter access constraints, leading to delays between the onset of the disaster,

often generating timely operational insights. However, each phase of data collection often targets only selected settlements and villages, which limits the generalizability of findings to affected areas. The method also requires significant human resources and can encounter access constraints, leading to delays between the onset of the disaster, data collection and subsequent reporting. Data-collection methodologies are generally tailored to specific country or disaster contexts. Where full access to displaced populations is possible, humanitarian organizations may implement or utilize detailed registration techniques to capture survey-like data through interviews with individuals or households (Carletto et al., 2022; Kilic et al., 2017). For example, IOM DTM displays data collected by the Ukrainian Ministry of Social

CURRENT MODELS FOR DATA COLLECTION6

DYNAMIC ESTIMATES OF DISPLACEMENT IN DISASTER REGIONS Chapter 2 – Key lessons on triangulating data for displacement estimates in disaster contexts Table 2. Data comparison of primary datasets discussed Acronym Data Provider Geographic unit Frequency Type Unit Access Processing IOM RDDª General Population Survey Ukraine IOM Admin 1 (oblast) Admin 2 (raion) Initially every 2 months; quarterly from Round 13 onward Sample survey Respondent High cost Manpower CNI PKb Community Needs Identification Pakistan IOM Settlement/ Village (aggregated up to Admin 1 and Admin 2) Depends on funding and government approvals Key informant Settlement/ Village High cost with some access issues Manpower GPS MDc GPS Phone Data Private company GPS Hourly Multiple Apps Mobile device High cost Computational Meta MAPId Meta’s Platform Marketing API Private company Area unit (e.g. Admin 2) Daily Single App User (age and sex) Free Computational Meta DfGe Meta Data for Good Private company Area unit

device High cost Computational Meta MAPId Meta’s Platform Marketing API Private company Area unit (e.g. Admin 2) Daily Single App User (age and sex) Free Computational Meta DfGe Meta Data for Good Private company Area unit (grid or admin-level) Subdaily Single App User Free Computational Notes: a. The IOM RDD General Population Survey in Ukraine is a repeated cross-sectional study that uses a consistent questionnaire administered to new respondents each round. In its first 12 rounds, a random digit dial (RDD) method was used to survey 2,000 adults (18+) livin