Yesterday, Marcy Wheeler wrote a post pointing out that two of the six Americans confirmed killed by drone strikes were servicemen, a pair of Marines killed during a close air support mission (the other four are Anwar al Awlaki, targeted directly for assassination in a drone strike, and three other suspected Al Qaeda militants killed as “collateral” in drone attacks). There’s obviously some conflation here; the Marines died in a CAS mission during a firefight in Afghanistan, while the others were targeted more or less directly in the strategic drone campaign against Al Qaeda. Nevertheless, Marcy argues that the death of the Marines reveals a larger problem about drone targeting; drone strikes are launched using insufficient information, thus leading to lots of accidental deaths (whether American military or foreign civilian.)
I think that this is mostly wrong, with a few elements that may be right. The first problem is, I think, a misunderstanding of what modern close air support looks like. On twitter yesterday, Marcy expressed the view that if an F-16 had launched the airstrike, there would have been an additional layer of intelligence and accountability. But, from the point of view of modern CAS, this is simply wrong. Much CAS in Afghanistan is delivered from medium altitude by fighter-bombers such as the F-16 and the F/A-18. These aircraft spend very little time over the target, and have very little ability to determine with any precision the events on the ground. The weapons they release (often 2000# bombs) are targeted based on information from ground troops and (if available) live footage taken from drones. Close air support of this nature has, however, been part of the Afghanistan War since 2001, when special forces operators directed most of the targeting in support of Northern Alliance forces.
CAS is also delivered by aircraft such as the A-10 and the AC-130, which fly lower, slower, and have more time over the target. However, a A-10 pilot still has less information about the course of a firefight than virtually any drone operator; drone pilots fly slower and can stay on station longer, and are less concerned about the possibility of getting shot down. An AC-130 is a different story, because orbiting the battlefield is part of its job, but AC-130s are relatively few and don’t deliver much of the CAS in Afghanistan.
Moreover, the ordnance carried by a Predator drone does a lot less damage than the ordnance carried by an F-16 or an A-10. This isn’t always good; sometimes a 2000# bomb is an effective way to suppress or destroy an enemy position, or to kill a concentration of enemy fighters. In friendly fire terms, however, the small weapons payload of the Predator is a distinct plus; NATO soldiers only die when they’re directly targeted by the Predator, as opposed to simply being near an F-16 strike.
The ideal CAS platform would be something like a Super Tucano, which has numerous weapon hardpoints, gunfire capability, and a low enough speed to loiter over the battlefield until the pilot figures out what’s going on. For a variety of reasons (few of them relating to the question at hand), the Air Force has nary an interest in buying the Super Tucano or an aircraft like it.
The future of close air support is, as Drunken Predator suggests, fighter-bombers layered upon drones. This isn’t ideal; there are some cases when having an A-10 would be of great help, and many cases in which a Super Tucano could handle CAS very effectively. However, this is almost certainly a better situation than held at the beginning of the Afghan War, when fighter-bombers (and sometimes strategic bombers, such as the B-52 and the B-1B) delivered weapons without the assistance of near-ubiquitous drone footage. In this system, drones collect intel, combined with ground troops, and deliver some of the weapons to targets, with the manned aircraft launching heavy ordnance based on drone and ground intel.
It should go without saying that friendly fire incidents happen in all of these scenarios. Close air support is necessary because enemy forces sometimes have positions which are either hard to attack directly from the ground, or from which they can pin down friendly troops. In the latter situation especially, decisions on where to target bombs and missiles often have to be made in a very short amount of time with limited amounts of intel. Marcy quotes a report indicating that the spatial separation of different parts of the CAS team played a role in mistaken killing of the Marines, but doesn’t put this in any comparative context. The history of CAS in World War II, Korea, Vietnam, Gulf War I, and the wars of the War on Terror is replete with instances of CAS gone horribly awry, including low altitude strafing of friendly infantry and vehicles. Pilots, SOF operators, and infantrymen all make dreadful errors about the precise location of friendly and enemy forces; in previous conflicts, the lack of communication between ground and air forces has been a tremendous problem. The deaths of two Marines in a drone CAS strike in Afghanistan doesn’t tell us very much about the proclivity of the current system to create friendly fire casualties; for that, we’d need much more robust data comparing the frequency of such casualties in situations with and without drones. I don’t have that data handy, but I think there’s very good reason to think (based on the immediate availability of intel and the size of ordnance fired by drones) that the presence of drones tends to cut down on friendly fire casualties.
As should be obvious, this has nothing whatsoever to do with the legitimacy, legality, accuracy, or good sense of the campaign to target suspected terrorists with drones in Pakistan, Yemen, Somalia, et al. The issues raised by the use of drones in support of conventional military operations as opposed to the use of drones in what amounts to a strategic bombing campaign-light aren’t completly separable, but they’re distinct enough that great care should be taken before conflating the two.