Accuracy of Freefall Aerial Bombing
It is said that freefall bombing is inaccurate, but that’s a very inaccurate thing to say – all pun intended.
This article estimates with two methods how accurate modern freefall aerial bombing is. The methods don’t meet scientific standards, as the intent is rather to find a good rule of thumb for, for example, defense planning. We come to a conclusion that for bombs dropped from a non-harassed modern bomber at high altitude, for example over 5000 meters, a reasonable rule of thumb for CEP is 50 meters. The article consists of a literature study and an OSINT piece verifying the former.
My interest in the subject started when I gathered a crude dataset of different air launched weapons with, for example, ranges, carrying aircraft, weather requirements and accuracy. Most of the data is listed right on Wikipedia, but the accuracy of modern freefall bombing efforts was very elusive. It was unclear whether this is due to secrecy or the complex dependencies of accuracy to multiple variables.
If you drop a stone, it will hit the point directly beneath it. But if you drop a shaped bomb from a freely moving airplane in conditions with varying wind and air pressure, calculating the point of impact becomes increasingly hard. If the wind and pressure conditions change during the flight path, calculating the trajectory beforehand becomes downright impossible. In reality there are even more sources for variance. Manually choosing the time of release is error prone, as is flying the aircraft at a constant level path and even the bombs may not be uniform or released at exactly the same moment. Sometimes dispersion is also sought after. It’s better to have eight bombs hit different parts of an area target than a single point.
To give an understanding of how much these variables affect the accuracy, let’s stop for two data points. During World War Two it was estimated that a three-degree change in heading at release lead to a 200-meter deviation at impact and a flight speed deviation from calculated of just a couple of kilometers per hour led the bombs astray for tens of meters.
Despite the complexities, I believed it has to be possible to have at least a statistical estimate of bombing accuracy or alternatively an accuracy function with a couple of the most important explaining variables, for example, drop altitude. These didn’t seem too secret, so I suspected that by asking on Twitter I would get at least mediocre sources. Naturally, I got more than I bargained for. To verify these further, I suspected I’d need to do some calculations of my own. Enter video footage from Syria, where Russia has used massive amounts of freefall bombs. I focused on a case study of two popular show reels of UAV recorded videos. The first video has likely one Su-25 run and one Su-24 run and the second video was presumably several Tu-22M3 only missions. The latter video was especially interesting as, since the plane type is a bomber, the strikes are certainly all from a relatively high altitude. My assumption is from 5.000 to 8.000 meters. By happenstance, Tu-22 is also the plane type I was most interested in to begin with.
During the past years USA has replaced almost all free fall bombs with JDAMs. The JDAM is a kit that is installed onto a conventional bomb. It makes the combination many times as expensive, but doesn’t differ in explosive or fragmentation potential. Conventional bombs cost a couple of thousands and a JDAM kit about 26.000 dollars. Thus, the JDAM has to be better in some other way. While there are a few possibilities, it’s relatively safe to assume the JDAM is more precise. With GPS the JDAM achieves a 5-meter CEP and without it a 30-meter. Thus, we get a lower boundary for the CEP of freefall bombing. If freefall bombing would be as accurate as a JDAM, JDAMs wouldn’t be used.
The Russian solution to the same need is the SVP-24, which is not an addition to the bomb, but rather a bombing computer added to the airplane. Thus, bombing with SVP-24 fulfills the definition of freefall bombing. Some Russian sources claim that they can achieve GPS guided JDAM level accuracies e.g. 3 to 7 meters with the SVP-24 in ideal conditions. They further claim, that even in battle conditions the accuracy would be on the level of 20 to 25 meters. While it is unclear if these accuracies mean the CEP, weaker accuracy measures are seldom used. Thus, these are very challenging claims to achieve. Personally, I find them hard to believe, but at least they add to our understanding and confidence of the maximum freefall bombing accuracy estimate.
Interestingly the same source estimates that bombing without such a computer has accuracies between 150 to 400 meters. Here, the high end, for a change, feels intuitively too large, as it corresponds to the maximums estimated in World War Two.
So, let’s look at that more historic data and begin from World War Two. The earliest estimate gives us a figure that only forty percent of the bombs hit a circle with radius of about 450 meters. This was before 1944 when the CEP was even introduced as the standard way of measuring accuracy in the US. When the new measure was introduced, the CEP accuracy had already improved to around 300 meters from the altitude of 5.000 meters using the B-17 and B-24 bombers.
One major invention behind accuracy improvement was likely the Norden bombsight. We can find a lot of data for it starting from testing in the 1930s. From an altitude of 1.200 meters a CEP of 11 meters was achieved in training. From higher up, they managed to achieve a 23-meter CEP. And when the set-up was moved to actual war, Air Corps achieved a 120-meter CEP from the altitude of 4.600 meters. Still zooming out and taking into account the whole attack, the bombs ended up on average 300 to 400 meters from the intended targets varying especially by unit and bombing altitude. Ending up on average 300 meters from the target is the practically the same thing as a 300-meter CEP – perfectly in line with the earlier measure for the B-17 and B-24.
During the war this inaccuracy made dive bombing an interesting choice for all belligerents. The Germans trained their crews for a CEP of 25 meters compared to 50 to 75 meters for level bombing. Both of these measures were expected to at least double in the heat of the battle. American efforts for dive bombing were on a similar level. As these figures for level bombing can roughly be stated as a CEP of 100 to 225 meters, they are a lot better than the ones presented for the Norden bombsight in the previous sources. This is likely due to combat being more challenging than what the bombing schools estimated.
The next data point that we have is a US estimate on the capabilities the Soviets could develop by the mid-60s. It’s a pretty safe assumption these are close to or slightly better compared to their own capabilities during the time of the writing. The best thing about these estimates is that they are presented as a function of bombing altitude making us able to draw that function.
To summarize, the visual bombing improves with lower altitude a lot more than radar directed bombing and the best estimated accuracy is about 122 meters from an altitude of 3.000 meters. This gives us a total estimated range for CEP of 120 to 900 meters with CEP more than doubling when the altitude doubles.
Fast forward another decade to Vietnam and the bombs dropped by the F-105s achieved a CEP of 111 meters. This was when the airplanes were not shot at. The CEP increased to 136 meters under anti-aircraft artillery fire. Another report gives the A-1 a 90-meter and the F-4 a 150-meter CEP, when bombing from 600 meters of altitude. The difference is attributed to the faster speed of the F-4. The accuracy reverts back to World War Two levels of 300 meters during the night time or during adverse weather conditions. Yet another source states both this huge variance and my research problem painstakingly clearly by saying that the daily accuracy average ranges from 30 to 300 meters depending on tactics, target and weather. With radar bombing they managed to control some of the variance and get the accuracy to about 150 meters. The surprising thing is that this was considered as good as dive bombing accuracy, although the figures from World War Two for dive bombing already looked better.
Finally, the book “The Precision Revolution” gives a direct estimate of 61 meters for the CEP of US freefall bombing in 1990. Haven’t personally read the book, but Tuomo Rusila pointed this out in the previously mentioned Twitter discussion. It is good to keep in mind how dominant the US was in Iraq. However, it is quite probable, that the technology and techniques did improve dramatically from the Vietnam era. The rate of improvement has probably slowed down since the 90’s. Both due to diminishing marginal returns, but also due to the diminishing importance of freefall bombing.
In conclusion, it’s difficult to believe any modern freefall bombing would achieve a lower than 25-meter CEP and on the other hand it seems quite proven that a 60-meter CEP can be achieved. Everything is naturally highly dependent on the conditions, skill and technology used.
Next, we’ll compare these figures to what Russia has documented for us in Syria.
For the first strike, that I’m presuming to be done by a Su-25, we can identify three points of impact. The distances between the points range from 102 to 257 pixels in my original. At the same time, what I believe is a truck, is about 18 pixels long. If the truck is 8 meters in reality, the distances between the impacts are 45 and 114 meters. Calculating an exact CEP is not very fruitful with only three impact points. This is the only one of the strikes that I have currently geolocated and it’s at 36.407257°, 37.153259°. Looking at the distances Google gives, we get a rough validation for my estimates and subsequently proof that the truck is quite close to 8 meters.
In the second strike we can identify six points of impact. Using the road width as a reference point with presumably approximately 6 meters of width, we get a 68-meter distance between the furthest separated points of impact. However, looking at the location of the buildings, it’s likely that the aim point is close to the center of the frame or to the left from it, so all of the impacts are to the right and up from the aim point. I’m assuming this strike was carried out by a Su-24.
Let’s move on to the next video.
In the third strike, we finally see that in reality the CEP doesn’t describe how several freefall bombs behave, if dropped from the same airplane. Naturally, they do not disperse circularly, but elliptically. Before jumping to conclusions, it’s good to note that the second highest impact point is struck the last and noticeably later than the others. This is also why you can’t yet see an explosion in the still frame. Thus, it isn’t clear that the bombs would disperse a little, but are just spread out due to the movement of the dropping airplane and sequential release from the bombing shaft of the Tu-22.
Unfortunately, this image has no terrain features I could recognize and measure against. But since the explosions looks comparably the same size as in the other strikes and the different freefall bombs used by Russia shouldn’t differ much in that sense, I’m inclined to believe this strike has about the same dispersion as the others.
The next two images are from a large strike against an area target. Since the images are from two different segments of the video, it’s possible they are not even from the same day. At least one can’t see some of the smoke from the first segment in the following one. In addition to these images, there were several other runs on the target. While in the case of an area target, dispersion might be sought after, it’s notable how large the dispersion is. The building marked with the red line in both images is in fact quite huge. This can be seen from the following zoom in.
My estimate is that it should be at least 40 meters in length, which makes the distances between the individual explosions in each of the bombings to be at least 100 meters and up to 300 meters.
Fifth strike is again hard to measure due to lack of measurable features. The lines might be trenches making them about 1 meter wide. Again, the dispersion feels to be on a familiar level. The actual target of the strike might be some pillboxes or sandbag fortifications.
In the final strike we have another area target with low accuracy. Again, this may be intentional. However, the building in the middle looks like some kind of an industrial hall, which should be at least 20 meters in length. In the image it is 55 pixels long. As the maximum distance between two impact points is 629 pixels, these are then approximately 229 meters apart. As we have at least eight impact points, it’s finally somewhat meaningful to also approximate the CEP for this particular strike. We can fit half, i.e. four of the impact points, within about 80 meters of an imaginary aim point somewhere in the average of these impacts.
Conclusion and Discussion on Errors and Further Studies
Looks like in the videos the Russians aren’t quite achieving the 61-meter CEP USA claims to have achieved in 1990. The equipment may be worse, the dispersion may be intentional or USA might have inflated their accuracies. However, it’s clear there’s no magical 3-meter accurate SVP-24 in play. I’m still inclined to believe that the Americans’ achievement could be surpassed now almost 30 years after. Also, as it is better to be safe than sorry in the context of defense planning, I’m advocating a 50-meter CEP as a good rule of thumb for freefall aerial bombings.
Whole different question is then how accurate you need to be? Depends not just on the bomb, but also on the target. If you don’t score a direct hit on a tank, it’s going to be very difficult to harm it. If you get within tens of meters from an unarmored fellow standing upright, he’s pretty much dead. Fragments from a modern bomb can fly for hundreds of meters, but it’s always also random whether you get hit by one. Then again, the attacker might deploy a cluster or an incendiary bomb, making the calculation totally different.
There were several sources of error and inaccuracies in this study that would need to be eliminated for a scientific article. The sources in the literature study may be motivated to lie in either direction, but especially the OSINT-part would benefit from more analysis. First of all, the Russians are certainly selecting only successful mission videos to publish. Second, several of the frames are so tight that far away misses are outside the camera angle. Like mentioned, we’d need to know the aim point to estimate not just the relative dispersion but the actual deviation and CEP from the target. I’m pretty sure, we can’t assume the UAV camera crosshairs are pointed at the target point. These errors mostly make the strikes look more accurate in this study than they are in reality.
Third area of errors is the level of effort I personally put into the analysis. In many of the strikes it looks like there are more impacts close to each other that might be discerned with a frame by frame analysis. Taking these into account would generally improve the CEP. Also, I could’ve used real math in finding out the weighted averages of the impact points instead of just measuring the maximum distances between them. Similarly, real OSINT geolocation could’ve been used in finding out the real dimensions of the distance reference features. These errors could change the end result in either direction. Unfortunately, I could find only one of the strikes from the Bellingcat database of geolocated Russian strikes in Syria. Also I didn’t quickly find anything on Google Maps around Palmyra matching the second video.
In conclusion, I suggest using the mentioned 50-meter CEP in your work as a rule of thumb. It’s a conservative estimate from the defender’s viewpoint. However, if you still need more accuracy in accuracy estimates, please go on with the research and let me know of your results, too!