The potential of spatial methods for conflict and atrocity early warning/early action systems
In a world afflicted with complex conflicts, grievous atrocities, and the compounding effects of the climate crisis, the pursuit of peace and protection for vulnerable populations has never been more critical. As of June 2023, mass atrocity crimes are occurring in 13 countries with five additional countries at imminent risk of mass atrocity crimes occurring if preventative action is not taken (GlobalR2P, 2023). Between 2000 and 2020, at least 37 countries “experienced mass atrocities or had serious concerns raised that they could take place” (UK House of Commons International Development Committee, 2022). Efforts to anticipate, prevent, and intervene early on evolving conflict and atrocities are varied in content, scope, and sophistication and leverage a diverse set of tools and interventions that integrate local knowledge and data. Conflict and atrocity early warning/early action systems (CEWS), originating in the 1950s within military intelligence communities (Wulf and Debiel, 2009), are examples of such development.
CEWS are mechanisms that “are intended to identify and analyze conflict trends, alert to conflict risk, inform decision-making and initiate timely responses to prevent violent conflict” (Peace Infrastructure). CEWS have undergone substantial changes over time, adapting in response to failures and shortcomings, as well as incorporating novel technological solutions and innovative strategies. By 1977, the Defence Advanced Research Project Agency (DARPA) in partnership with academic collaborators, had developed an integrated crisis early warning system (ICEWS) that systematically collected, analyzed, and modeled events data (O’Brien, 2010). Though promising, these types of systems were still within the strategic intelligence framework, emphasized interstate conflict as opposed to state vs non-state conflict, and placed heavy emphasis on military indicators to threats and forecast events (Gentry and Gordon, 2019).
The horrific atrocities that caused the death of millions in Rwanda and Bosnia led to the creation of more robust humanitarian early warning systems under the collaborative efforts of the United Nations and other non-governmental organizations. Established in 1996, ReliefWeb was established as a humanitarian early warning information system and now, administered by the United Nations Office for the Coordination of Humanitarian Affairs, continues to provide timely and reliable information for disaster preparedness, prevention, and response. Subsequently, contemporary operational CEWS have evolved to include diverse systems that intend to predict, detect, and document threats on a global, country-wide, or regional scale, while others are community-centric (considered to be ‘third generation’ CEWS; Rupesinghe 2009) and designed to capture events on a very localized scale (Muggah and Whitlock, 2022).
Most recently, there has been a call to bridge the ‘gap between early warning and early action’ (Wulf and Debiel, 2009). Guidance co-published by the University College of London and the World Bank in 2023 advocates for early warning systems that ‘extend beyond warning dissemination to include damage prevention, mitigation actions, and response conditions to ensure capacity to act’ and facilitate early and anticipatory action (Yore et al., 2023). Despite evident progress in both the academic and programmatic CEWS, we need to have more actionable information systems that strengthen EW/EA organizations’ ability to act.
Our gap analysis
In an effort to understand the breadth of data and data analytics used for conflict and atrocity CEWS, the Signal Program team conducted a literature scoping review and a series of key informant interviews with practitioners working at 23 different EW/EA organizations. The team drew several conclusions from this research:
- A wide array of conceptual models, variables, datasets, and methodologies are being utilized to explore and analyze conflict and atrocities. This was notable in both the peer-reviewed literature and key informant interviews. While this variability drives novel investigation, it also leads to minimal analytic standardization of conflict prediction and prevention. This leaves us with the exigent question: what elements are critical to the development of a highly effective and efficient early warning system?
- There is a tremendous gap between the conceptual models, data, and analytic methods used for CEWS models that are developed and published in the academic literature and those utilized by local and regional EW/EA organizations. While complex predictive models predominate literature, post-hoc descriptive events analysis are the most prevalent in local EW/EA processes.
- Most published research considers conflict dynamics on a global or regional scale, often comparing risk and severity across countries. Very few studies consider a more granular, sub-country analysis. This stands in stark contrast to the key informants we interviewed, who typically collect data on a very small scale (in individual communities, cities, or provinces).
- No published articles commented upon the inclusion of their algorithms into programmatic decision-making or the impact of their CEWS on conflict-affected populations.
- All of the publications included in our scoping literature review were associated with institutions and/or organizations in high-income countries while their content was predominantly regarding conflict in low and middle-income countries.
- Locally operated CEWS normatively consists of data collection, verification, analysis, and information dissemination, which ultimately initiates a response. The data collection phase of the CEWS heavily relies on input from communities and trained volunteers, who collect data on conflict events (event details, involved actors, number of casualties etc). The collected data is subsequently analyzed by organizations overseeing the EWS. Analysis methods range from ad hoc analysis to the use of descriptive analyses and basic spatial analysis, which are then shared with decision-makers weekly or monthly either within the organization or to relevant agencies who may be mobilized in a response.
- Existing CEWS are making incredible contributions to EW/EA efforts. However, our interviews revealed several challenges with existing CEWS. Namely, poor data quality, quantity, and completeness can result in incomplete or skewed analyses and the understanding of evolving conflict patterns. Furthermore, this data is often manually analyzed, and there is a need and an appetite to build in-house capacity for more complex and spatially-explicit analytics.
The power of spatial methods to strengthen atrocity prevention efforts
The precipitation and escalation of conflict and the propensity for mass atrocities are spatially heterogeneous. Variables that influence one communities’ conflict and atrocity risk will differ from another. From descriptive mapping to more complex, spatially-explicit predictive models, spatial methods enable practitioners to map, analyze, and examine relationships of variables over space and time. Within the conflict and atrocity prevention space, spatial methods can begin to bridge the gap between higher-level analytics proposed in academic literature and what is utilized in operational CEWS. It can easily shed light on hotspots and trends of violence and displacement, provide insights into the diverse variables that influence conflict and the occurrence of mass atrocities, potentially anticipate mass atrocities in a spatially-heterogeneous way that informs local interventions, and provide monitoring and evaluation tools for interventions.
While numerous forms of conflict incident monitoring already exist, the incorporation of spatial analytics can amplify the understanding of incident trends in any given geographic location. With widely available systems like Google Maps, BingMaps, Apple Maps, Earth Observation, and OpenStreetMap, we are able to accurately geo-locate and monitor the extent or the spread of incidents or incident patterns. Spatial analysis may help identify the nature and location of a conflict event, communities and demographics affected, and determine its impact or severity. Spatially-explicit data regarding past and present social, political, conflictual, demographic, and/or climate-related information can be used to predict the spread or mitigation of atrocities.
Farmer herder conflict
An example of how spatial methods can influence conflict and atrocity prevention is evident when we consider the periodic violence that erupts in Nigeria between farmers and the nomadic Fula or Fulani herdsmen that has claimed the lives of thousands of people (Gaffey, 2018). This type of violence can be avoided if the migratory patterns of the Fulani are tracked using GIS technology. GIS can also be used alongside another technology developed in Kenya called Chipsafer which prevents cattle rustling by remotely tracking and geolocating individual livestock. If farmers are caught in the Fulani migratory paths, one can reasonably assume that it might lead to cattle rustling, which is often a trigger for the violence. Using spatial methods, we can create virtual boundaries to ensure that the Fulani herdsmen are safely diverted from their normal path and to enable agencies like the regional government or law enforcement agencies to escort herders away or protect the farmers who sit in the middle of a potential migration path (Palihapitiya, 2019).
Near-real-time conflict monitoring
Another example, below, shows a cluster of events from Kaduna, Nigeria that were tracked by the Interfaith Mediation Center (IMC) on Google Maps and imported into an early warning platform called Waayama (Figure 1). The map below shows high densities of reported events accompanied by a detailed breakdown of incidents in the dashboard, below. The dashboard captures incident type, location, time, and more to facilitate data analysis. This example from Nigeria demonstrates the value of spatial methods particularly when used alongside a relational database (Palihapitiya, 2019). The data collected from the field can be verified and mapped through this system in near-real time. The maps and dashboard can be shared with key early response stakeholders so that they may also monitor and take early action to prevent atrocities.
Figure 1. Geolocations of recent activity in Nigeria. Source: Waayama. (2013). Waayama early warning system Nigeria.
Going further
Together, these preliminary examples start to evidence the added value of spatial methods in conflict and atrocity prediction, prevention, and early response. But spatial methods, when incorporated in relevant, sustainable, and ethical ways into CEWS and conflict and atrocity prevention workflows have so much more potential to improve situational awareness, create more precise and impactful interventions, promote organizational efficiency, and advance communication and collaboration.
Cluster and outlier analysis (also known as Anselin’s local Moran’s I) has the ability to identify statistically significant clusters of violence and, perhaps more compelling, outliers wherein certain communities portray lower or higher levels of violence. Identification of ‘hot spots’ (which is technically a different spatial analytical method) can help direct resources and interventions, whereas examination of outliers (either of high or low level of violence in comparison to neighbors) can provide insight into the variables that may make the community more or less resilient to conflict. Space-time evaluation can show trends in conflict in, as it is aptly named, space and time. And geographically-weighted regression models can help reveal location-specific variables that are associated with the precipitation, increase, or de-escalation of violence.
Going even further, spatially-explicit predictive models, such as the one proposed by Witmer et al. (2017), have the ability to anticipate the effect of variables such as population dynamics, socio-political indicators, and climate change on conflict evolution (Figure 2). This spatial forecasting of conflict is compelling in that it can inform anticipatory funding, peace-building efforts, and policy, but the data required is vast, validation is requisite, and thus its utility in CEWS and near-real-time action remains limited.
Figure 2. Observed and forecasted violence given a specific shared socioeconomic pathway as predicted by a spatially-explicit forecasting model proposed by Witmer et al. (2017).
While the potential for spatial methods to transform EW/EA activities is clear, the creation and adoption of these methods must acknowledge the workflows, capacity, and needs of the end user. The science of spatial analytics is advancing at an exponential pace, but its integration into EW/EA and CEWs, specifically, requires a nuanced understanding of the data available, biases inherent to these data, the spatial literacy of the end-user, and the ethical ramifications of utilizing spatial methods. Spatial optimization or spatially-explicit predictive modeling may be appropriate for some, but it should be end-users who drive the design and implementation of any spatial methods into CEWS.
Call to action
Given the exigent crises already plaguing humanity, the escalating threats of climate change, and rapidly evolving technologies, the need to create more robust CEWS mechanisms and actions is critical. Worldwide, existing EW/EA organizations are engaged in impactful conflict and atrocities prevention work. As EW/EA organizations continue to address the complexities of atrocity prevention in a multitude of regions where democracy, rule of law, development, and other structural conditions conducive to sustainable peace are still being established or reestablished, the Atrocity Prevention Lab will look for ways to support these efforts.
As academic methodologists from a deeply privileged institution in a high-income country, we hope to leverage this platform to not only advocate for spatial methods to create more impactful systems but also to utilize our position of power to center the voices of those who engage in atrocities prevention at the forefront, who are not appropriately recognized and are often left out of decision-making.
We wish to invite these frontline actors and those interested in collaborating with them into this space. We explicitly are inviting those voices who are underrepresented, underheard, and revolutionary. Together, we can save lives.
Citations
Chipsafer. (2023). https://www.chipsafer.com/.
Gaffey, C. (2016, April 20). The Nigerian Conflict You’ve Never Heard of. Newsweek. https://www.newsweek.com/nigerias-herdsmen-and-farmers-are-locked-deadly-underreported-conflict-45029.
Gentry, J. A., & Gordon, J. S. (2019). Concepts of Strategic Warning Intelligence. In Strategic Warning Intelligence: History, Challenges, and Prospects (pp. 11–26). Georgetown University Press. https://doi.org/10.2307/j.ctvb4bsfx.6
House of Commons. (2022). From Srebrenica to a safer tomorrow: Preventing future mass atrocities around the world. Third Report of Session 2022–23. Parliamentary Copyright House of Commons. https://committees.parliament.uk/publications/30270/documents/175201/default/
Muggah, R., & Whitlock, M. (2022). Reflections on the Evolution of Conflict Early Warning. Stability: International Journal of Security and Development, 10(1). https://stabilityjournal.org/articles/10.5334/sta.857
O’Brien, S. P. (2010). Crisis Early Warning and Decision Support: Contemporary Approaches and Thoughts on Future Research. International Studies Review, 12(1), 87–104. http://www.jstor.org/stable/40730711
Palihapitiya. (2019, December 19). Information and communication technologies for peace. In Understanding International Conflict Management (1st ed., Vol. 1, pp. 144–157). Routledge. https://doi.org/10.4324/9780429448164
Peace Infrastructures. (n.d.). Early Warning and Early Response. Peace Infrastructures. https://www.peaceinfrastructures.org/thematic/early-warning-early-response.
R2P Monitor. (2023, June 1). R2P Monitor, Issue 65, 1 June 2023. Global R2P. https://www.globalr2p.org/publications/r2p-monitor-issue-65-1-june-2023/
Relief Web. (2023). https://reliefweb.int/.
Rupesinghe, Kumar & Foundation for Co-Existence (Colombo, Sri Lanka). (2009). Third generation early warning / Kumar Rupesinghe. Colombo : The Foundation for Co-Existence.
Witmer, F. D., Linke, A. M., O’Loughlin, J., Gettelman, A., & Laing, A. (2017). Subnational violent conflict forecasts for sub-Saharan Africa, 2015–65, using climate-sensitive models. Journal of Peace Research, 54(2), 175–192. https://doi.org/10.1177/0022343316682064
Wulf, H., & Debiel, T. (2009). Conflict early warning and response mechanisms: Tools for enhancing the effectiveness of regional organisations? A comparative study of the AU, ECOWAS, IGAD, ASEAN/ARF and PIF.
Yore, R., Fearnley, C., Fordham, M., Kelman, I. (2023). Designing inclusive, accessible early warning systems: Good practices and entry points. Global Facility for Disaster Reduction and Recovery, the World Bank, and University College London. https://reliefweb.int/report/world/designing-inclusive-accessible-early-warning-systems-good-practices-and-entry-points