Recently, authors writing in these pages (see here and here) have criticized the progress made by the United States, and by implication its allies, on protecting civilians during combat operations. To fully and fairly assess them, such charges merit in-depth contextual examination. For instance, sense of scale can be skewed by zeroing in on operations gone wrong, to the neglect of those in which efforts to avoid, or at least reduce, civilian casualties have proved successful. And discussion of the subject often inadequately captures the complexity of some battlefields, where opponents fail to distinguish themselves, operate among the civilian population, use human shields, or even engage in activities meant to cause civiliancasualties that can be blamed on their enemy, an especially reprehensible form of so-called “lawfare” (see, e.g., here and here).
Yet claims of insufficient progress must be taken seriously. As individuals who have served in military operations, we will be the first to admit that perspective bias can manifest on both sides of the issue. Warfare, and indeed the law of armed conflict, involves a delicate balancing of humanitarian concerns and military exigencies. How individuals and organizations make that balance inevitably reflects their perspective. We can learn from each other.
In this article, we will describe some of the forms of guidance used to impose requirements concerning civilian casualties and damage to civilian objects (“collateral damage”), explain key processes that are relied upon to minimize such harm, and highlight some of the technologies that facilitate doing so. Our purpose is to add a degree of operational context to the discussion, not debate or excuse.
We must caution that the policies and procedures we describe derive primarily from counter-insurgency operations. Many would be unworkable during an armed conflict with a near-peer adversary. In such a conflict, the practical ability of U.S. and allied forces to minimize harm to civilians would be impacted by the scale of operations, the operational tempo, and the enemy’s ability to degrade the information underpinning our decision-making. Policies and procedures are seldom one size fits all.
However, the law of armed conflict (LOAC) would, of course, continue to apply, irrespective of the nature of the conflict. That is where we shall begin.
The Law
Targeting is governed by a number of customary LOAC rules, most of which also appear for States Party (not the United States, but key allies like Australia, Canada, and the United Kingdom in Additional Protocol I to Geneva Conventions of 1949. That instrument governs only international armed conflict, but the main targeting rules contained therein reflect customary international law in both international and non-international armed conflict.
Most fundamentally, civilians and civilian objects are protected from both direct and indiscriminate attacks. “Direct” refers to intentionally aiming at civilians or civilian objects, whereas “indiscriminate” denotes attacks that are not directed at a specific target at all, whether civilian or military. As a practical matter, the former is often more operationally relevant because it raises the issue of the degree to which uncertainty may be tolerated when identifying a prospective target as a military objective, and thus lawfully subject to attack (so long as the other rules discussed here are observed). That is an especially complicated matter when the enemy fails to distinguish itself from the civilian population or operates in proximity to civilians.
The most significant practical protection for civilians and civilian objects, however, comes with the requirement to take precautions in attack, which drives many operational choices. Beyond the legal requirement that “constant care” be taken to spare civilians and civilian objects during military operations (see discussion in DoD Law of War Manual, § 5.2.3.5), attackers must take all feasible measures to verify that the target is a lawful military objective. They must also consider feasible target, tactics, and weapons options that achieve the desired effects of the operation (the “military advantage” in LOAC terms) while limiting collateral damage. And attackers must precede their attacks with warnings if the circumstances permit.
It is essential to understand that the precaution rules are subject to operational realities. For instance, using a particular platform (e.g., an RQ-4 Global Hawk “drone”) to contribute to the verification of a target may not be feasible if there is a greater need for the platform’s use elsewhere. Or striking a different target to achieve the desired effect, but the attack on which risks less harm to civilians, may not be feasible because of significantly increased risk to the forces used to attack it. And many warnings will not be viable when attacking individuals because they would be alerted and might seize the opportunity to enhance their defenses, take shelter, or flee. That said, precautions in attack rules require attackers to do the best they reasonably can in the attendant circumstances to limit civilian harm. This could involve, for instance, tasking a drone in advance of the attack to monitor a potential target for the presence of civilians in and around that target; tasking a second drone to scan-out and look for transients at the time of attack while the primary drone focuses on precise weapons delivery; and selecting a weapon with less blast effect where the integrity of civilian structures in vicinity of the target area is at particular risk.
Once the precautions assessment has been conducted, an attacker must evaluate proportionality. By the rule of proportionality, an attack is unlawful if “expected to cause incidental loss of civilian life, injury to civilians, damage to civilian objects, or a combination thereof, which would be excessive in relation to the concrete and direct military advantage anticipated.”
Importantly, it is the expected collateral damage and the anticipated military advantage that drive proportionality analysis. Thus, proportionality determinations are made ex ante, not post factum. The determinative factors are what the attacker knew, or should have known, regarding collateral damage and what military advantage the attacker reasonably expected to achieve; proportionality is not assessed based on the collateral damage or military advantage that eventuated, except as they bear on the reasonableness of the attacker’s analysis at the time the attack was planned, approved or executed. In practice, this is a highly subjective determination because it compares dissimilar values – collateral damage and military advantage – that are themselves hard to measure. As a result, while still important, the more protective rule on the battlefield is that requiring precautions.
Compliance with these rules by those conducting attacks is operationalized through guidance imposed by higher echelons of command, processes designed to assess the likelihood of collateral damage and identify ways to avoid causing it, and technologies that offer the attacker precision, greater clarity in the battlespace, and redundant capabilities. We turn to these next.
Means of Providing Guidance on Civilian Protection
U.S. and coalition forces use rules of engagement (ROE) to restrict the application of force during armed conflict, in part to limit civilian casualties to the extent feasible (Operational Law Handbook, ch. 5). ROE are more restrictive than LOAC, for they also incorporate policy and operational constraints and are usually tailored to the specific battlefield environment in which they apply. The rules can restrict such matters as acceptable targets, the geographic range of operations, time of operations, and the use of particular munitions. ROE also sometimes reinforce LOAC provisions that might be of particular significance in certain combat environments, such as the prohibitions on the destruction of religious and cultural property when being applied in situations of internecine conflict. In many circumstances, they tighten them, as in the case of restricting operations against religious facilities being illegitimately used by the adversary to self-defense scenarios only.
Depending on the nation, operation, and composition of a coalition, limits, and restrictions designed to protect the civilian population may appear in other forms of guidance. For instance, a Targeting Directive might contain granular guidance regarding attacks that require the use of multiple weapons to achieve the desired effects. This might be so, for instance, when a close grouping of buildings comprises the object of attack. In such cases, the guidance, in whatever form it might take, would likely require the interval time between each munition striking its desired point of impact (DPI) to be limited to not more than a few seconds. The rationale for such a stipulation would be to avoid the risk of civilians rushing to the initial blast site, only to fall victim to a subsequent attack on the same or a nearby objective.
Of particular note, operational commanders often use tactical directives to provide guidance, as well as their “commander’s intent” for the employment of force and avoidance of collateral damage. For example, in July of 2009, the International Security Assistance Force (ISAF) Headquarters released General Stanley McChrystal’s Tactical Directive, which instructed ISAF forces to “respect and protect the population from coercion and violence.” In particular, he instructed “leaders at all levels to scrutinize and limit the use of force like close air support (CAS) against residential compounds and other locations likely to produce civilian casualties” and authorized the use of air-to-ground munitions and indirect fires against residential compounds only “under very limited and prescribed conditions” (the specific conditions are classified). The Directive also restricted when and how ISAF forces could enter civilian houses and prohibited the use of force against religious sites such as mosques, except in self-defense.
No-strike and restricted target lists also contribute to the avoidance of civilian harm. Developed through an inter-agency, and at times multinational, process, no-strike lists (NSL) (JP 3-60, at II-12) identify objects, termed “no-strike entities,” that are non-military and protected from the effects of military operations. LOAC, ROE, and policy and operational considerations drive the determination of which objects to include on an NSL. They typically include, for example, medical, non-governmental, educational, diplomatic, cultural, religious, and historical entities. It is important to emphasize that they are tailored to the operation and the area in which it is occurring. For instance, an NSL might include a factory, a strike on which would risk environmental or health hazards because of the possible release of chemicals. No-strike entities are identified, analyzed, verified, cataloged, and disseminated through a formal no-strike process to ensure U.S. and coalition units receive real-time updates on no-strike entities in their area of operations. Only in exceptional and very narrowly defined circumstances, such as a need to defend oneself or friendly forces who in an on-going engagement involving the entity, may the NSL be overridden.
Restricted target lists (RTL) (JP 3-60, at II-13) are likewise developed to mitigate civilian harm. They differ from no-strike lists in the sense that restricted targets are valid military objectives, but restrictions are placed on attacking them for various reasons, as in restricting damage to a bridge that needs to be crossed by friendly forces later in the conflict. Such restrictions can also be driven by a desire to avoid collateral damage that is not prohibited by LOAC rules in the attendant circumstances. For instance, attacks on specified restricted targets might be limited to nighttime, when civilians are less likely to be in the area, or the RTL may mandate the use of particular weapons, like precision-guided munitions, when the targets in question are located in population centers – even in situations where the restriction will negatively affect the military advantage likely to be gained by attacking the target. An RTL will set forth any restrictions on effects caused (for example, a prohibition on the total destruction of civilian airfields used by the enemy), the rationale for restrictions, and the approval authority for engaging the target or causing any specified effects such as incapacitation or destruction.
Processes and Policies that Afford Protection to Civilians
Among the various processes employed to avoid collateral damage during attacks, the Collateral Damage Estimation Methodology (CDEM) takes center stage. Developed by the U.S. military, CDEM is used to identify any “collateral concerns” (i.e., civilians, civilian objects, or other protected objects) within the collateral effects radius (CER) of the warhead, weapon, or weapon class under consideration for use in the attack. If collateral concerns are present, progressively more demanding mitigation techniques are used to remove such concerns.
For example, to eliminate collateral concerns, CDEM may dictate that an aircraft: (1) use a precision-guided munition (PGM); (2) delay “fuzing” on that PGM, which means the munition detonates within the targeted building (as opposed to on impact with the roof) in order to allow the interior walls of the target to help contain fragmentation and blast effect; and/or (3) adopt a specific attack-heading (direction from which the attack will occur) in order to cause the direction of more blast or fragmentation to travel away from the civilians or civilian objects. Shielding and aim-point offset are further techniques that can remove the risk of harm. Shielding is a technique that utilizes structures located between the intended aimpoint and collateral concern(s) to “shield” the latter from the effects of a warhead. It may also be possible to strike other than the optimal (in terms of achieving the desired destruction) aimpoint(s) on a structure to reduce the projected blast and fragmentation that will impact an adjacent collateral concern (known as “aim-point offset”).
Suppose collateral concerns within the CER still exist after mitigation techniques have been exhausted. In that case, CDEM provides a numeric estimate of the number of civilians who may be injured or killed if the attack goes forward. This estimate will then be used to determine if the attack can proceed under any circumstances at all (e.g., whether it is prohibited because it will be disproportionate under LOAC) and, if it can, who may approve the attack.
Historically, the primary decision aid for determining who may approve a strike has been the noncombatant and civilian casualty cutoff value (NCV) that is set forth in the applicable ROE. To illustrate, if the CDE for a proposed attack indicates an NCV value of 2 (i.e., two civilians may be killed) and this does not cross the line of “excessive” collateral damage relative to the anticipated military advantage (proportionality), the ROE may require approval from a 1-star commander (instead of the lower-ranking dedicated targets team). The term “NCV” has recently been removed from U.S. doctrine, but target engagement authority is still tied to tolerance levels for civilian casualties (as built into the relevant ROE).
The numerical estimate CDEM generates derives from empirical data, probability, historical observations, and complex modeling. It takes into account standard baseline factors such as population density, the structural integrity of buildings, the proposed time of attack, and other variables. Where current intelligence can provide a more refined estimate, that data must be employed in the casualty estimate. To illustrate, if a pattern-of-life (POL) analysis of a proposed target indicates that a civilian structure nearby will have more civilians present during the day, this will supersede the standard baseline for expected civilians within such a building provided by CDEM. In practice, targeting, intelligence, and operations personnel mesh their expertise and experience, alongside current multi-source intelligence, to tailor casualty estimation to the operational environment’s specifics. Updates to the CDEM also derive from “lessons learned” analysis of prior operations.
U.S. CDEM has been approved for use during NATO combat operations and has also been influential in helping non-NATO allies like Australia to reduce civilian casualties. Of course, the precise application of collateral damage estimates will be affected by issues that remain contentious, such as the conditions under which individuals lose their protection as civilians because they are directly participating in hostilities or a member of an organized armed group, and whether there is a rule requiring doubt as to civilian status to be resolved in favor of that status. Nevertheless, according to Human Rights Watch (p. 18), CDEM “is the military’s best means of minimizing civilian casualties”.
CDEM is designed to be simple and repeatable, which means it can be utilized where there is little time for deliberation. However, when a full CDE process is not possible, “field CDE” may be conducted by those supporting forces in the heat of battle, such as fast jet pilots, “drone” operators, or “Joint Terminal Attack Controllers” (JTAC). JTACs are highly-skilled personnel who direct combat aircraft engaged in kinetic air operations from a forward position on the ground. They support both conventional infantry and special forces.
It is, of course, easier to mitigate the harm to civilians caused by an attack on a pre-planned target than one that becomes necessary when a firefight is taking place between opposing forces (“troops in contact,” or TIC) and civilians remain in the vicinity. This is why the failure to distance civilians from military objectives or broader areas of operations is such a vexing violation of the LOAC requirement to take “passive precautions” (i.e., the requirement that the party in control of an area take steps to protect civilians from dangers resulting from military operations).
In some TIC situations, it may be possible for troops to withdraw from the area (especially in protracted counter-insurgencies). But where units are pinned down by enemy fire or taking casualties, that possibility does not present itself. Further, an entrenched enemy in an urban setting cannot be defeated when a withdrawal policy is in effect (unless one would prefer to starve out the enemy by siege, with the dire consequences that may entail for civilians). In such cases, “field CDE’ is useful to avoid, or at least abate, civilian harm. That said, the uncomfortable truth is that civilians are always at greatest risk where the time for deliberation is condensed.
Another method used to minimize civilian harm by U.S. and coalition forces is employment of the concept of positive identification (PID) of a target during the target vetting process. U.S. Joint Doctrine describes (p. 208) PID as “identification derived from observation and analysis of target characteristics including visual recognition, electronic support systems, noncooperative target recognition techniques, identification friend or foe systems, or other physics-based identification techniques.” The requirement for PID is satisfied when there is a “reasonable certainty” that the proposed target is a legitimate military objective (see, e.g., Operation Iraqi Freedom Combined Forces Land Component ROE Card.) Of course, certainty is a somewhat ambiguous standard, and reasonableness is highly contextual, but what is clear is that PID is a relatively high threshold. That said, PID is sometimes straightforward to establish, most notably when enemy forces are attacking you.
Specific requirements may be imposed for PID in various types of operations. This is usually done in the ROE, although it may appear in other forms of guidance. As an illustration, the ROE might require that in the absence of “eyes on target” (e.g., ground forces at an observation post witnessing enemy movements in real-time), two forms of verification, such as a cell phone intercept and prior intelligence indicating an enemy meeting will be held in the building to be attacked, are generally required for PID.
Even when the PID standard is satisfied, the realities of combat operations can prevent target engagement. For instance, a unit must establish PID and maintain PID “chain of custody” from the point of the initial identification through the target’s engagement. Thus, if a drone observing an individual known to be a member of the enemy forces loses track of that individual, PID has to be reestablished before the attack may proceed. Similarly, in air combat, the Advanced Medium-Range Air-to-Air Missile (AMRAAM) uses radar to search for and identify enemy aircraft (i.e., establish PID). The ROE may provide that the pilot must confirm the radar’s positive identification before deciding to engage the enemy aircraft.
At its core, PID is about verifying that the intended target is a military objective, as is required by LOAC. Therefore, absent PID, the target cannot be attacked as a matter of law. Due to operational needs and policy concerns about civilians and civilian objects, the United States has extended the concept in the last decade beyond identifying a target to include assessing the presence of any civilians. It has also sometimes set thresholds of civilian casualties beyond which attack is prohibited irrespective of the anticipated military advantage (at least absent high-level approval). For instance, in 2013, the Obama Administration adopted a policy requiring “near certainty” that no civilians will be killed or injured during counterterrorism strikes outside “areas of active hostilities” (as described in the policy guidance). The Trump Administration maintained this standard.
Technologies that can be Leveraged to Enhance Civilian Protection
Finally, the U.S. military and its closest allies are in a privileged position when it comes to the technology at their disposal to avoid harming civilians or civilian objects. High-resolution earth-observation satellites provide imagery for applications ranging from planning to post-attack battle damage assessment. Unmanned aerial systems combine robust sensors for target detection and identification, have the ability to pass target information to strike aircraft, and can assist other attack platforms by laser designating (i.e., ‘buddy lase’) a target, which will guide the latter’s weapons to its desired point of impact (commonly, ‘DPI’).
However, when it comes to civilian casualty reduction, the evolution of air-delivered munitions is perhaps the most significant development. In 1991, when Air Force leaders reviewed Operation Desert Storm, they saw the need for an inexpensive PGM that could be used in any weather. The Joint Direct Attack Munition (JDAM) provided this capability as a “bolt-on” guidance kit that converts unguided gravity bombs into all-weather PGMs. The affordability, versatility, and accuracy of the JDAM have led to their extensive use since Operation Allied Force in 1999 (the NATO air campaign against the then Federal Republic of Yugoslavia). JDAMs have continued to significantly increase in accuracy since then. Indeed, the Air Force surged production of weapons like the JDAM, Small Diameter Bomb 1, and AGM-114 Hellfire during the campaign against Daesh (or ISIL) in Iraq and Syria due to the desire to reduce civilian harm, and provided munitions to allied air forces due to their depleted stocks from operations in Afghanistan and Libya.
The United States also has recently developed a modified Hellfire missile (i.e., the AGM-114R-9X or “Flying Ginsu”) for pinpoint airstrikes with no explosion, “drastically reducing” the chances of civilian casualties. Indeed, multiple U.S. officials have suggested that this weapon’s use should have been publicly disclosed years ago to illustrate a willingness to minimize civilian harm in densely populated urban centers in Iraq, Afghanistan, Syria, Somalia, or Yemen.
Precision technologies have opened up new possibilities for avoiding the harm to civilians and their property that is the inevitable result of armed conflict. In that sense, advances in precision represent an immensely positive trend. However, the allure of precision must be guarded against. This includes exaggerated expectations as to what precision can deliver, leading to unreasonable counterfactual demands on the military. For example, the “flying Ginsu” just mentioned may be an effective option for surgically targeting a high-value individual in a moving target, but its utility in more conventional operations is highly circumscribed.
Concluding Thoughts
Again, it is not our purpose to debate those who argue that U.S. and allied forces are failing to adequately address the tragic reality of civilian casualties and damage to civilian objects; we only seek to add operational perspective to the discussion. We also acknowledge that the mere existence of the guidance, processes, and technologies to avoid civilian harm described above does not, in itself, solve the tragedy of collateral damage. They have to be employed in good faith and with a pervasive commitment to avoiding collateral damage whenever possible. Considering differing perspectives on the challenge of civilian harm during combat forces us to reflect on the scope for, and means of, improvement. No process, system, holistic approach, or mind-set is beyond reproach.
That said, criticism is most impactful when evidence-based (e.g., the investigation described here). To suggest, for example, that the U.S. and allied militaries are somehow dismissive of, or irresponsible with respect to, civilian casualties undermines thoughtful and effective analysis. As should be apparent from the discussion above, civilian casualties do not necessarily equate to a failure to comply with LOAC or apply best practices in civilian casualty minimization on the battlefield. Rather, our personal experience is that U.S. and allied forces take exceptional measures to reduce the pain inflicted by war. Because we are not always successful, the task at hand is, and will remain, to apply what we learn from any failures in future conflicts.
The views presented are those of the authors in their personal capacity and do not necessarily represent the views of the any government departments or institutions with which they are affiliated.