There has been much controversy over the numbers of people killed in Darfur. It is increasingly possible to move beyond extrapolation to an actual count of incidents and fatalities. Earlier this week, the Genocide Intervention Network (GI-Net) published its analysis of fatalities due to violence in Darfur, based on available reports for 1 January-8 September 2008. In collaboration with GI-Net, I have begun the exercise of cross-checking all these reports with the UNAMID incident dataset, and extending the analysis up to the end of 2008.

Overall, there is a very good fit between the two datasets, and we are confident that all major incidents have been included. The level of monitoring in most areas of Darfur is now sufficiently good that few incidents escape notice. What remains is detail.

It is important to note that all of what follows are preliminary numbers, certain to change as a number of incidents are investigated more fully. It will take some weeks before we have firm numbers.

It is also important to notice that this analysis is only for violent deaths, and excludes any excess mortality on account of hunger and disease, and all other forms of violence and distress such as sexual violence, forced displacement and loss of ways of life. Violent death is not the total of all harms inflicted upon the people of Darfur.

Nonetheless, I decided to post this very preliminary analysis for two reasons. First, it is important to emphasize that we know a great deal about violence in Darfur. Second, this posting services as an appeal to all who have data on specific incidents to make it available so that gaps in the database can be filled in.

GI-Net gives a total figure of 1,211 violent deaths between 1 January-8 September, for an average of 150 per month. The UNAMID dataset has a total figure of 1,551 reported violent deaths for the calendar year, for an average of 130 per month.

GI-Net’s figures indicate between 359-720 civilian deaths. The range is due to the number of people whose identity was not specified. (The higher figure counts all those unidentified as civilians.) It indicates 279 combatants killed (on all sides).

The UNAMID dataset indicates 496 civilians killed, of whom 60 were IDPs (38 killed in the 25 August incident at Kalma). Its figures indicate 416 combatants killed. For 107, identity was not indicated, and for a further 640 killed in inter-tribal fighting, the distinction between combatant and civilian was not clarified.

A particular issue in determining the numbers of civilian deaths is what status to accord the people killed in inter-tribal fighting. Based on the UNAMID dataset, I have coded these separately: 476 civilians and combatants killed in inter-tribal fighting among Arab tribes and a further 164 in other intra-tribal fights (e.g. Fellata and Gimir). We are planning to re-examine the GI-Net dataset to ascertain how to match up the datasets on this issue.

As all the militia involved in inter-tribal fighting were armed by the government, there is a rationale for placing ultimate responsibility for these killings on the government. But the story is more complicated. Often this fighting takes the form of one tribal militia raiding the villages or cattle camps of its rival, killing civilians and combatants, with a reprisal raid taking place over the following days. These deaths could be coded as either two incidents of militia killing civilians or as a generalized “˜tribal fight.’ My preference is to keep these incidents separate from those in which government-armed militia attack IDPs or non-Arab villagers.

GI-Net counts 43 fatalities from aerial bombardment, UNAMID counts 46 (29 civilians, 6 combatants and 11 not specified).

The UNAMID dataset indicates 90 fatalities in bandit or criminal attacks (16 of them among attackers). Many of the militia attacks also involved looting, especially of livestock, but this number is only for those specifically identified as criminal incidents.

The UNAMID database includes a number of cases in which the Sudan army fought against the Arab militia (38 deaths), different police or security agencies fought against one another or against the army or the SLA-Minawi group (46 killed), or the SLA-Minawi and the SLA-Free Will fought each other (66 fatalities).

The exact numbers and details need to be further analyzed and several incidents need to be followed up in more detail. There are some discrepancies between the two datasets. For example GI-Net includes the fatalities in Omdurman in May and in al Atrun in August, both of which are outside UNAMID’s operational area and so are not included in its figures. On a couple of occasions, notably for the SAF-JEM battles north of el Geneina in February 2008, UNAMID reports “˜high’ casualties (but which, in the absence of a definite figure, have not yet been entered into the data) while GI-Net has specific figures. On another occasion, the UNAMID incident file reports a high figure for claimed civilian casualties attributable to the SLA-Minawi, but no other sources report the incident””we are checking to see if this actually happened. On a third occasion, the UNAMID incident report has details of 62 civilians injured, while (based on a published report), GI-Net has included 62 fatalities. As these cases are cleared up, the true numbers will become more evident.

There are some reasons to suspect under-reporting. For example, there are only a few cases in which the datasets include follow-up reporting to specify the number of wounded people who later died. There are undercounts of fatalities among the armed movements. There are also reasons to suspect over-counting, illustrated in cases which an initial report of fatalities is followed up by an investigation which reveals a lower number. My best guess is that the UNAMID database numbers may rise to about 1,700.

Watch this space.