AOAV uses a methodology adapted from an incident-based methodology used by Landmine Action and Medact in 2009 which in turn was based on the Robin Coupland and Nathan Taback model.[i] Data on explosive violence incidents is gathered from English-language media reports on the following factors: the date, time, and location of the incident; the number and circumstances of people killed and injured; the weapon type; the reported user and target; the detonation method and whether displacement or damage to the location was reported. AOAV does not attempt to comprehensively capture all incidents of explosive violence around the world but to serve as a useful indicator of the scale and pattern of harm. No claims are made that this data captures every incident or casualty of explosive violence.

Selecting incidents
An RSS reader is used to scan Google News for key terms which relate to explosive weapon use: air strike* artillery* bomb* bombing* cluster bomb* cluster munitions* explosion* explosive* grenade* IED* mine* missile* mortar* rocket* shell.*

At least one casualty from an explosive weapon must be reported in order for an incident to be recorded. Incidents with no clear date or which merely give a location as a country are excluded, as are incidents which occur over a period of more than 24 hours (e.g. 150 people killed by shelling over the last week). Casualty numbers must be clearly stated; reports which only describe ‘several’ or ‘numerous’ cannot be recorded. When there are multiple sources for the same incident, those which provide the most detail or the most recent casualty information are selected.

AOAV uses a wide range of English-language news sources, many of which are translated by the publisher. In total there were 423 different sources used in 2021, with the ten most used being The New York Times (used as a either the first or second source for 127 incidents in 2021), Xinhua (99), Reuters (93), ANI (86), Anadolu Agency (83), Al Jazeera (72), Ukrinform (48), DAWN (42), Associate Press (39), and Myanmar Now (31).

Recording guidelines

Civilian / armed actor or security personnel: Casualties are assumed to be civilians unless otherwise stated or there is an implicit suggestion that those impacted were combatants. The vast majority of news reports state the target status and if there is any doubt, alternative news sources are identified. Casualties are recorded as ‘armed actors’ if they are reported as being members of the military, members of non-state armed groups, or security personnel who are likely to be armed, for example; police, security guards, intelligence officers, and paramilitary forces. 

Intended target: The target for an attack is only recorded if one of the three conditions below are met:

  • The target is declared by the user.
  • It is clearly reported in the source.
  • The specific contextual conditions of use clearly indicate a target (e.g. if an IED is attached to the car of a police officer or soldier, ‘State armed’ is recorded as the target).

Populated area: Incidents are designated as occurring in populated areas likely to contain concentrations of civilians if: a) It is stated in the source (e.g. a busy street, a crowded market); b) If an incident occurs in or near a pre-defined location which is likely to contain concentrations of civilians e.g. commercial premises, entertainment venues, hospitals, hotels, encampments (containing IDPs, refugees, nomads), markets, places of worship, public gatherings, public buildings, public transport, schools, town centres, urban residential neighbourhoods, villages/ compounds. This definition of a populated area is based on Protocol III of the 1980 Convention on Certain Conventional Weapons (CCW) which defines concentrations of civilians as: “any concentrations of civilians, be it permanent or temporary, such as in inhabited parts of cities, or inhabited towns or villages, or as in camps or columns of refugees or evacuees, or groups of nomads.”[ii]

User status: Responsibility for the use of explosive weapons is assigned where any of the following conditions are met:

  • The group or actor responsible has claimed responsibility.
  • The user of the explosive weapon is clearly stated in the report.
  • If the user of the explosive weapon has employed technology clearly associated only with that user in the context in question.

If none of these conditions are met then the user is recorded as unknown. Users are recorded as ‘state and non-state’ when both users are identified but it is not possible to establish which one was responsible for the particular incident.

This methodology is subject to a number of limitations and biases, many relating to the nature of the source material on which it is dependent and the lack of a mechanism to follow up reports with in-depth investigation. It is recognised that there are very different levels of reporting across regions and countries so that under-reporting is likely in some contexts. In addition, only English-language media reports are used, which does not provide a comprehensive picture of definitive explosive weapon use around the world.

The methodology is designed to capture distinct incidents of explosive violence with a clear date and location. In some contexts of explosive violence, particularly during intense armed conflict, casualties cannot be assigned to specific incidents but a total number is reported as the result of a period of days. These casualties cannot be included in the dataset.

As the methodology relies on reports which are filed shortly after an incident took place, there is no mechanism for assessing whether people reported as wounded in the immediate aftermath of an incident subsequently died from their injuries. This is another factor that should be assessed when considering the likelihood that the actual numbers of fatalities of explosive violence are higher than the numbers recorded by AOAV. There is no systematic base-line for determining what constitutes an injury, and AOAV is therefore subject to the assessment of the news source.

On a number of occasions firearms were also reported as having been used alongside explosive weapons. While AOAV always tries to determine the casualties specifically caused by explosive weapons, in these incidents new sources are not always able to clarify which casualties were caused by which weapon type, particularly in incidents that involved large numbers of casualties. It is therefore possible that some casualties in these incidents may not have been caused by explosive weapons.[iii]

AOAV is focused on capturing the harm caused by explosive weapons at the time of use. Explosive weapons that fail to explode as intended can linger in the form of explosive remnants of war (ERW) for years, if not decades, to come. In 2021 AOAV recorded 481 civilian casualties from unexploded or abandoned ordnance. These casualties occurred in 19 different countries and territories. The actual number of casualties from ERW is likely to be far higher.[iv]

Poorly secured or stockpiled explosive weapons can also cause unintended harm to civilians. AOAV recorded three stockpile explosions around the world in 2021.[v]

Media reports used by AOAV are a valuable resource for better understanding the scale and pattern of explosive violence use. However, these reports are less helpful for capturing other types of harm known to be characteristic of explosive weapons in populated areas. Damage to infrastructure, the risk of ERW, long-term health effects, and displacement are all aspects of the pattern of harm caused by explosive weapons which are not fully represented in the data set. However, reporting on these effects is often limited, with news sources focusing on the immediate aftermath of an incident. For instance, only 266 incidents out of 2,500 reported damage to a location in 2021. Effects which are the result of cumulative levels of explosive violence, for instance communities displaced by heavy shelling or continued insecurity, cannot be fully represented by this research.

[i]For more information see

[ii] “Protocol on Prohibitions or Restrictions on the Use of Incendiary Weapons (Protocol III),” to the UN Convention on Certain Conventional Weapons, Geneva, 10 October 1980, (accessed 18 March 2013).

[iii] For example, in street fighting in the Lebanese capital of Beirut on 21 May 2012 rocket-propelled grenades and machine guns were involved in attacks that injured six people. Hussein Malla, “Gunbattle in Beirut amid fears of Syria spillover,” The Associated Press, posted by The Guardian, 20 May 2012, (accessed 18 March 2013).

[iv] For example, The Landmine and Cluster Munition Monitor, which focuses on the impacts of landmines, cluster munitions, and ERW, recorded 3,628 mine/ERW casualties in 2012, 78% of whom were civilians. ICBL-CMC, “Landmine Monitor 2013: Casualties and Victim Assistance,” 2013, (accessed 10 April 2014).

[v] Since 1987 more than 460 unplanned explosions at munitions sites (UEMS) have been recorded. In 2013 Small Arms Survey had recorded 13 UEMS incidents. Small Arms Survey, “UEMS Incidents by Year (1987-2013),” (accessed 10 April 2014).