Minimizing harm to civilians was central to governmental and public consent for NATO's bombing campaign in the Federal Republic of Yugoslavia-an air war officially justified as humanitarian intervention. The decision to intervene was taken with the awareness that the use of force would be subjected to close scrutiny through the lens of international humanitarian law-and in the court of public opinion.
From the beginning of Operation Allied Force, NATO and allied government and military officials stressed their intent to limit civilian casualties and other harm to the civilian population. The practical fulfilment of this legal obligation and political imperative turned upon a range of decisions relating to targeting, weapons selection, and the means of attack.
Despite precautions, including the use of a higher percentage of precision-guided munitions than in any other major conflict in history, civilian casualties occurred. Human Rights Watch has conducted a thorough investigation of civilian deaths as a result of NATO action. On the basis of this investigation, Human Rights Watch has found that there were ninety separate incidents involving civilian deaths during the seventy-eight day bombing campaign. Some 500 Yugoslav civilians are known to have died in these incidents.
We determined the intended target in sixty-two of the ninety incidents. Military installations account for the greatest number, but nine incidents were a result of attacks on non-military targets that Human Rights Watch believes were illegitimate. (Human Rights Watch is currently preparing a separate report with a full analysis of our legal objections to the choice of certain targets.) These include the headquarters of Serb Radio and Television in Belgrade, the New Belgrade heating plant, and seven bridges that were neither on major transportation routes nor had other military functions.
Thirty-three incidents occurred as a result of attacks on targets in densely populated urban areas (including six in Belgrade). Despite the exclusive use of precision-guided weapons in attacks on the capital, Belgrade experienced as many incidents involving civilian deaths as any other city. In Nis, the use of cluster bombs was a decisive factor in civilian deaths in at least three incidents. Overall, cluster bomb use by the United States and Britain can be confirmed in seven incidents throughout Yugoslavia (another five are possible but unconfirmed); some ninety to 150 civilians died from the use of these weapons.
Thirty-two of the ninety incidents occurred in Kosovo, the majority on mobile targets or military forces in the field. Attacks in Kosovo overall were more deadly-a third of the incidents account for more than half of the deaths. Seven troubling incidents were as a result of attacks on convoys or transportation links. Because pilots' ability to properly identify these mobile targets was so important to avoid civilian casualties, these civilian deaths raise the question whether the fact that pilots were flying at high altitudes may have contributed to these civilian deaths by precluding proper target identification. But insufficient evidence exists to answer that question conclusively at this point.
Another factor in assessing the higher level of civilian deaths in Kosovo is the possible Yugoslav use of civilians for "human shields." There is some evidence that Yugoslav forces used internally displaced civilians as human shields in the village of Korisa on May 13, and may thus share the blame for the eighty-seven deaths there.
In an important development, sensitivity to civilian casualties led to significant changes in weapons use. Widespread reports of civilian casualties from the use of cluster bombs and international criticism of these weapons as potentially indiscriminate in effect led, according to senior U.S. Department of Defense officials interviewed by Human Rights Watch, to an unprecedented (and unannounced) U.S. executive order in the middle of May to cease their further use in the conflict. The White House issued the order only days after civilians were killed by NATO cluster bombs in the city of Nis on May 7. U.S. cluster bomb use did apparently stop at about that time, according to Human Rights Watch observations, although British cluster bomb use continued. Human Rights Watch released its own report on May 11 questioning the civilian effects of cluster bombs and calling for a moratorium on their use.
International Humanitarian Law and Accountability
In its investigation Human Rights Watch has found no evidence of war crimes. The investigation did conclude that NATO violated international humanitarian law.1 Human Rights Watch calls on NATO governments to establish an independent and impartial commission, competent to receive confidential information, that would investigate violations of international humanitarian law and the extent of these violations, and would consider the need to alter targeting and bombing doctrine to ensure compliance with international humanitarian law. Such a commission should issue its findings publicly. Human Rights Watch also calls for NATO to alter its targeting and bombing doctrine in order to bring it into compliance with international humanitarian law.
With respect to NATO violations of international humanitarian law, Human Rights Watch was concerned about a number of cases in which NATO forces:
· conducted air attacks using cluster bombs near populated areas;
· attacked targets of questionable military legitimacy, including Serb Radio and Television, heating plants, and bridges;
· did not take adequate precautions in warning civilians of attacks;
· took insufficient precautions identifying the presence of civilians when attacking convoys and mobile targets; and
· caused excessive civilian casualties by not taking sufficient measures to verify that military targets did not have concentrations of civilians (such as at Korisa).
One disturbing aspect of the matter of civilian deaths is how starkly the number of incidents and deaths contrasts with official U.S. and Yugoslav statements. U.S. officials, including Secretary of Defense William Cohen, Deputy Secretary of Defense John Hamre, and Gen. Wesley Clark, have testified before Congress and stated publicly that there were only twenty to thirty incidents of "collateral damage" in the entire war. The number of incidents Human Rights Watch has been able to authenticate is three to four times this number. The seemingly cavalier U.S. statements regarding the civilian toll suggest a resistance to acknowledging the actual civilian effects and an indifference to evaluating their causes.
The confirmed number of deaths is considerably smaller than Yugoslav public estimates. The post-conflict casualty reports of the Yugoslav government vary but coincide in estimating a death toll of at least some 1,200 and as many as 5,000 civilians. At the lower end, this is more than twice the civilian death toll of around 500 that Human Rights Watch has been able to verify. In one major incident-Dubrava prison in Kosovo-the Yugoslav government attributed ninety-five civilian deaths to NATO bombing. Human Rights Watch research in Kosovo determined that an estimated nineteen prisoners were killed by NATO bombs on May 21 (three prisoners and a guard were killed in an earlier attack on May 19), but at least seventy-six prisoners were summarily executed by prison guards and security forces subsequent to the NATO attack. The countervailing claims about the civilian death toll underscore the need for full accountability by NATO for its military operations.
The Objective of This Report
This report has the limited goal of assessing the number of civilian deaths from NATO attacks, as a step toward assessing NATO forces' compliance with their obligation to make protection of civilians an integral part of any use ofmilitary force. The benchmarks to be used for judging NATO's attacks are those of international humanitarian law, also known as the laws of war.
In concentrating on civilian deaths, this report addresses only peripherally the damage to civilian property and infrastructure upon which civilian welfare depends, an issue to be addressed in a later report. Nor does this report address other broad issues which are important for an assessment of the war. These include the obligations of the international community to act effectively to prevent crimes against humanity and war crimes; the legality under international law of NATO's launching the operation; the constraints arising from issues of sovereignty; and the modalities of international consensus and decision-making. The report also does not address the war crimes and crimes against humanity committed by Serbian and Yugoslav forces against ethnic Albanians. These gross violations of international humanitarian law, as well as abuses committed by the Kosovo Liberation Army (KLA), have been documented in numerous Human Rights Watch reports in 1998 and 1999, and continue to be the focus of investigations.
Compiling and Evaluating the Evidence
A fundamental challenge in the analysis of the war over Kosovo is to distinguish the facts of civilian deaths from the propaganda. In order to investigate civilian deaths resulting from NATO bombing, a Human Rights Watch team conducted a twenty-day bomb damage assessment mission in Serbia (including Vojvodina) and Montenegro in August 1999. The team visited ninety-one cities, towns, and villages, and inspected forty-two of the ninety sites of incidents in which civilian deaths occurred. Human Rights Watch researchers also conducted ongoing investigations inside Kosovo beginning June 12, the day NATO entered the province. While most of this research was on war crimes committed by Serbian and Yugoslav forces against ethnic Albanians, several cases relevant to this report were investigated, including the case of Dubrava prison, and incidents involving refugee convoys. Many of the remaining sites in Kosovo at which NATO attacks resulted in civilian deaths have been visited by independent observers whose findings are on the public record.
The Human Rights Watch team in Serbia and Montenegro met with officials from a dozen ministries in Belgrade, and in other locations met with regional, municipality, factory, and utility representatives. Taking eyewitness testimony and inspecting bomb damage, they were able to verify individual events and assess the veracity of wartime and post-war reporting. Human Rights Watch also met with or requested information from a range of officials of NATO countries, in particular the United States, although little new official information on the bombing incidents apart from official press statements has so far been released.
During the war, the research team compiled a master chronological database from military sources and from Yugoslav media and Internet reports, collating these with press and governmental reporting from the NATO countries. Research also drew upon a variety of bomb damage assessments undertaken by Yugoslav government agencies which, in some cases, have produced meticulous documentation on incidents. In order to assess sometimes contradictory renditions, we reviewed these data sets against other information from Yugoslav sources, while comparing this with information from NATO states, in particular the United States and the United Kingdom.
In the end, Human Rights Watch confirmed ninety incidents involving civilian deaths (see Appendix A). The field mission visited forty-two of the ninety confirmed incident locations and collected primary source information on thirty other incidents. Sufficient corroborating information existed on twenty-two others to recognize their credibility (including five about which NATO has officially confirmed that it attacked nearby targets at the same time). Eight incidents were eliminated altogether because they could not be verified or because the reported civilian deaths were actually deemed to be paramilitary troops or army soldiers (see Appendix C).
NATO has offered explanations for what went wrong or merely confirmed attacks in eighteen incidents. After May 7, when NATO began to publicly release a daily list of fixed targets, it confirmed attacking nearby targets in thirty-one of forty-three incidents that occurred between May 7 and the end of the war. NATO is on record as disputing three of the ninety confirmed incidents; Human Rights Watch was able to verify the authenticity of two of these (the other was in Kosovo) through on-the-ground inspections. Still, with the exception of the highly publicized incidents in which NATOhas been forced to offer explanations of what happened (for example, the attacks on the Chinese Embassy, the Djakovica-Decane convoys, and the Grdelica gorge), no information has been released on individual targeting missions, strike aircraft, or pilots.
The Civilian Deaths
This report documents civilian deaths in Operation Allied Force. Some 500 Yugoslav civilians were killed in ninety separate incidents over seventy-eight days of bombing, although it must be acknowledged that this evidence may be incomplete. In sixty-nine of ninety incidents, the precise number of victims and the names of the victims are known (see Appendix B). In another seven incidents, the number of victims is known and some of the names have been confirmed. In eleven incidents, the number of victims is known but the names are unknown. In three incidents, the names and precise numbers of victims are unknown.
Human Rights Watch concludes on the basis of evidence available on these ninety incidents that as few as 488 and as many as 527 Yugoslav civilians were killed as a result of NATO bombing. Between 62 and 66 percent of the total registered civilian deaths occurred in just twelve incidents. These twelve incidents accounted for 303 to 352 civilian deaths. These were the only incidents among the ninety documented in which ten or more civilian deaths were confirmed.
Available data on each incident are presented in Appendices A and B. They include descriptions of the physical destruction observed at the forty-two sites visited by Human Rights Watch, accounts by witnesses interviewed at each site and elsewhere in regard to particular incidents, documentation on individual incidents, and other available information compiled from public and private Yugoslav and NATO sources. In each incident report the emphasis is upon the evidence of civilian deaths, although any available evidence concerning the apparent target, the means of the attack, and the resulting physical damage is also presented.
Information drawn from the ninety incident reports allows a general picture to be drawn of the civilian deaths by the time, place, and circumstances in which they occurred. The deaths resulted from attacks on a range of targets, under different circumstances, and using a variety of munitions. Fifty-five of the incidents occurred in Serbia (including five in Vojvodina), three in Montenegro, and thirty-two in Kosovo. But between 278 and 317 of the dead-between 56 and 60 percent of the total number of deaths-were in Kosovo. In Serbia, 201 civilians were killed (five in Vojvodina) and eight died in Montenegro. A third of the incidents-a total of thirty-three-occurred as a result of attacks on targets in densely populated urban areas.
Human Rights Watch was able to determine the intended target in sixty-two of the ninety incidents (68 percent). Of these, the greater number of incidents were caused as a result of attacks on military barracks, headquarters, and depots; thirteen were a result of attacks on bridges (and one tunnel); six resulted from attacks on telecommunications and air defense facilities; five each resulted from attacks on industrial facilities, oil installations, and airfields; and seven were as a result of attacks on convoys or on what were perceived to be military forces in the field. These latter incidents were the most deadly, though two of the ten worst incidents occurred as a result of attacks on bridges.
Almost half of the incidents (forty-three) resulted from attacks during daylight hours, when civilians could have been expected to be on the roads and bridges or in public buildings which may have been targeted. Overall, forty incidents occurred in April, forty-five occurred in May, four in June, and one in March. May 29 saw the most incidents (with five), followed by four on April 14, May 30, and May 31. The pace of the air war peaked at the end of May.
Human Rights Watch was able to determine the weapons involved in the cause of the civilian deaths in only twenty-eight of the ninety incidents. Of these, twenty-one are incidents about which it can be confirmed that precision-guided munitions (PGMs) were used (though there could be others). This includes all of the attacks on bridges or targets in and around the Belgrade area. Cluster bomb use can be positively determined in seven incidents (another five are possible but unconfirmed). In almost all of the other instances, it is impossible to establish the weapon used.
Other than a factual statistical analysis of attacks, insufficient evidence exists to determine the cause of civilian deaths. U.S. Deputy Defense Secretary John Hamre has provided the only analysis regarding the "30 instances of unintended damage" that the Pentagon seems to acknowledge. Of those, he says one third occurred when the target was hit but innocent civilians were killed at the same time. Of the remaining twenty, three were said to be caused by human error when the pilot identified the wrong target, and two were caused by technical malfunction. In the other fourteen instances, the Pentagon has not yet announced whether human error or mechanical failure was responsible.
The Standards Applied
The conduct of warfare is restricted by international humanitarian law-the laws of war. International humanitarian law applies expressly and uniquely to armed conflict situations, with distinct provisions to regulate international and non-international (internal) armed conflicts. In evaluating NATO's use of military force in the Federal Republic of Yugoslavia, the laws of war provide the most relevant standards. With the initiation of the NATO bombing on March 24, 1999, the conflict in Kosovo and all of the Federal Republic of Yugoslavia became an international armed conflict to which the full body of international humanitarian law applied.
Protocol I additional to the Geneva Conventions of 1949 provides the basis for the evaluation here of NATO's bombing. This Protocol has been ratified by most NATO members, and the U.S. government has declared that it accepts all of the relevant standards. The basic principle of Protocol I, and of the laws of war generally, is that the civilian population and individual civilians shall enjoy general protection against dangers arising from military operations. This turns in large part on the requirement that attackers must distinguish between civilians and combatants and between military objectives and civilian objects. They must take all feasible precautions to avoid or minimize harm to civilians, and to this end may not attack civilians exclusively, or combatants and civilians indiscriminately.
Damage to civilian objects and civilian casualties that are incidental to lawful attacks on military objectives are known in military terminology as "collateral damage." The legality of an attack turns upon various factors. Firstly, the attackers must do everything feasible to verify that they are aiming at something specific-they cannot lash out blindly. Secondly, the attackers must establish that the objective to be attacked is a legitimate military objective. And thirdly, the attackers must establish whether an attack would endanger civilians and civilian objects, and must weigh this risk against the military advantage to be gained. Attacks which may be expected to cause incidental loss of life or injuries to civilians, or to cause damage to civilian objectives are indiscriminate if this harm to civilians is "excessive in relation to the concrete and direct military advantage anticipated" (Protocol I, article 57 (2)). The International Committee of the Red Cross (ICRC), the principal authority on the interpretation of international humanitarian law, has cautioned that the argument of proportionality can never justify very high civilian casualties and damage whatever the military advantage envisioned.
In researching each of the incidents involving civilian deaths we have sought to gather the facts that can enable analysts to assess the legitimacy of the real or perceived military objectives targeted; the care taken and procedures and criteria employed to confirm the military nature of the targets; the proportionality of the civilian deaths and the means employed in the attack to the express military objectives, where these were known; the correlation of civilian deaths to the location and nature of the targets selected; the timing of target selection as a factor in its appropriateness and the minimization of civilian harm; the methods and conditions under which distinct weapons systems were employed; and the potentially indiscriminate nature of some weapons systems in general and under certain conditions.
In assessing specific attacks, with a view to general observations on the conduct of the air war, the primary issue is whether due care was taken for the protection of civilians. Was the prospect of civilian deaths sufficiently taken into account in the targeting, the weaponry employed, and the means and conditions under which weapons were employed? This involves a review of the selection of targets, and the procedures through which these are determined, matters beyond the scope of the present report. So too is the larger question of whether the military objectives identified and targeted by NATO forces were wholly within what is permissible under humanitarian law. A separate report is being prepared by Human Rights Watch that will provide a full legal analysis of this aspect of the conduct of Operation Allied Force, as well as documentation concerning another 150 incidents in which civilians were injured in NATO attacks and scores of incidents in which there was damage to civilian property. The present report addresses the air war only throughits cost in civilian lives, as an indicator to be taken into account in assessing the larger picture of compliance with international humanitarian law.
Conclusions and Recommendations
Yugoslav civilian deaths in Operation Allied Force occurred under all circumstances, day and night, during good and bad weather, from the use of "smart" and "dumb" bombs, in attacks on almost every type of target. The number of incidents increased (and peaked) in the last three days of May. During this period, the intensity of the attacks also peaked. This was also a time when the percentage of precision-guided munitions being used by NATO aircraft was declining (due to inventory shortages and cost considerations). Most of the increased bombing effort, particularly in the large number of dumb bombs being dropped by B-1 and B-52 heavy bombers, was taking place in western and southern Kosovo. Attacks at a greater intensity in this area, which was largely depopulated, did not result in any increase in civilian deaths.
Throughout the air war, then, the incidents of civilian deaths per number of strikes seem to have remained fairly constant. Human Rights Watch therefore concludes that civilian deaths in Operation Allied Force were not necessarily related to the pace or intensity of the war, but occurred as a result of decisions regarding target and weapons selection, or were caused by technical malfunction or human error. This suggests that affirmative measures-restrictions on certain daylight attacks, prohibitions on the use of cluster bombs in populated areas, greater care in attacking mobile targets, better target selection-could indeed have been taken to further reduce the level of civilian harm during these military operations.
Five of the ten worst incidents involving civilian deaths (see Table 1 following Appendix C) were attacks on presumed Yugoslav military convoys or transportation routes, four in Kosovo. NATO Gen. Wesley Clark stated after the war that NATO often observed military vehicles moving on roads in Kosovo "intermixed with civilian convoys," particularly during bad weather. This does not exempt NATO from the obligation to take fundamental precautions to focus their effort on military objectives. In fact, after the first two incidents, on April 12 and 14, the civilian deaths led to changes in rules of engagement. While pilots had previously been required to visually identify the military nature of traffic before attacking, after the initial incidents new guidance directed that if military vehicles were intermingled with civilian vehicles, they were not to be attacked.
Similarly, after a mid-day attack on the bridge in the town of Varvarin on May 30 which resulted in civilian deaths (incident no. 81), NATO again provided excuses for the incident but then changed the rules of engagement for attacks on bridges. NATO Spokesman Jamie Shea publicly stated that the alliance had bombed a "legitimate designated military target" and stated that "we take the same precautions at midday as we do at midnight." Yet after the incident at Varvarin, according to Lt. Gen. Michael Short, the air war commander, pilots were directed not to attack bridges during daylight hours, on weekends, on market days, or on holidays. There is no evidence that the daylight timing of the attack at Varvarin (or on many other fixed targets) was critical to the destruction of the target-the attack was not directed specifically against military traffic. Around-the-clock bombing in these and other cases rather seems to have been part of a psychological warfare strategy of harassment undertaken without regard to the greater risk to the civilian population.
With respect to target selection itself, one of the worst incidents of civilian deaths, and certainly the worst in Belgrade, was the bombing of state Serb Radio and Television headquarters in Belgrade on April 23 (incident no. 30). There was considerable disagreement between the United States and French governments regarding the legality and legitimacy of the target, and there was a lively public debate regarding selection of Yugoslav civilian radio and television as a target group. There is no evidence that the radio and television headquarters meet the legal test of military necessity in target selection, as it made no direct contribution to the military effort in Kosovo. In this case, the purpose of the attack again seems to have been more psychological harassment of the civilian population than to obtain direct military effect. The risks involved to the civilian population in undertaking this urban attack grossly outweigh any perceived military benefit.
Another issue of intense public interest in the war is NATO's use of cluster bombs. There are seven confirmed and five likely incidents involving civilian deaths from cluster bomb use by the United States and Britain. Altogether, someninety to 150 civilians died from cluster bomb use. The first confirmed incident was on April 10 (incident no. 14) and the last was on May 13 (incident no. 57). After the technical malfunction of a cluster bomb used in an attack on the urban Nis airfield on May 7 (incident no. 48), the White House quietly issued a directive to restrict cluster bomb use (at least by U.S. forces). Cluster bombs should not have been used in attacks in populated areas, let alone urban targets, given the risks. The use prohibition clearly had an impact on the subsequent civilian effects of the war, particularly as bombing with unguided weapons (which would otherwise include cluster bombs) significantly intensified after this period. Nevertheless, the British air force continued to drop cluster bombs (official chronologies show use at least on May 17, May 31, June 3, and June 4), indicating the need for universal, not national, norms regarding cluster bomb use.
What is striking about the Yugoslav conflict, given the level of intense media coverage and public interest it has received in the United States and abroad, is that there is almost a complete lack of any public accountability by any of the national NATO members for missions undertaken in the NATO alliance's name. Little information has been released on nations or aircraft involved in bombing missions, on specific targets, and there is sparse information on weapons used in individual circumstances.
Human Rights Watch calls on NATO and its individual member states to:
· establish an independent and impartial commission, competent to receive confidential information, that would investigate violations of international humanitarian law and the extent of these violations, and would consider the need to alter targeting and bombing doctrine to ensure compliance with international humanitarian law;
· alter NATO's targeting and bombing doctrine to reflect the rules of engagement adopted during Operation Allied Force to increase civilian protection, as an important step toward bringing the doctrine fully into compliance with international humanitarian law;
· conduct an impartial and independent investigation of the nine incidents which were the result of attacks on inappropriate targets that Human Rights Watch believes were illegitimate. (Human Rights Watch will identify other examples of inappropriate targets in a separate report currently in preparation);
· carry out a full review of the compliance with international humanitarian law of the psychological warfare strategy of harassment of the civilian population evident in many of the attacks;
· acknowledge and evaluate all instances of civilian deaths and "collateral" damage-and not just some twenty or thirty select incidents-if there is to be a publicly relevant post-war analysis;
· declassify all NATO and national operations reports that could establish the precise nature of munitions employed in each attack to enable a comprehensive evaluation of the humanitarian dimension of the use of cluster bombs or other weapons, and suspend the use of cluster bombs until such evaluation has occurred;
· release comprehensive information on their operations-including chronologies of attacks, target lists, numbers and types of weapons expended, as well as any analysis or evaluations of the causes of incidents of civilian deaths or damage-that would enable independent observers to carry out a proper analysis of these operations under international humanitarian law; and
· examine targeting emphasis and weapons selection during the war and take whatever corrective measures are needed in the future to further minimize the civilian effects of the use of military force.
1 Rules of international humanitarian law arise from international agreements such as the Geneva Conventions, or develop as international customary law. States have an obligation to ensure compliance with all provisions of international humanitarian law, and to suppress all violations. War crimes constitute some of the most serious violations of international humanitarian law, known as grave breaches. These violations give rise to the specific obligation to search for and punish those responsible, regardless of the nationality of the perpetrator or the place where the crime was committed. Examples of war crimes are wilful killing, torture or inhuman treatment of noncombatants, wilfully causing great suffering or serious injury to body or health of noncombatants, or launching an indiscriminate attack in the knowledge that the attack will cause excessive loss of life or injury to civilians.