Jeffrey LewisDPRK RV Video Analysis

Last year, cameras in Japan noticed an object streaking the night sky — possibly the reentry vehicle from one of North Korea’s July 28 missile test. Did it burn up?  Survive?  If if it did, or did not, what does that mean for North Korea’s ability to deliver a nuclear weapon to targets throughout the United States?

James Acton and I reached out of NHK, which provided original video from three different cameras.  Along with David Wright, we attempted to analyze the video.


Video Analysis of the Reentry of North Korea’s July 28, 2017 Missile Test

James M. Acton, Jeffrey G. Lewis, and David Wright

North Korea has now conducted three test launches of two different ballistic missiles that can strike the continental United States—the Hwasong-14 on July 4 and July 28, 2017 and the Hwasong-15 on November 29, 2017.

North Korea’s ability to deliver a nuclear weapon to intercontinental distances raises the question of whether its reentry vehicles (RVs) would be able to protect the nuclear warhead during reentry. According to published reports, the U.S. intelligence community believes that while the RV on the July 4 test survived down to at least 1 km, the RVs on the July 28 and November 29 tests broke up at a much higher altitudes.

Based on these reports, various officials and independent experts have expressed doubt that North Korea has developed a viable reentry vehicle.  Unfortunately, such optimism may not be justified. Even if none of North Korea’s reentry vehicles performed entirely successfully in its 2017 ICBM tests, it is not possible for several reasons to conclude that North Korea has not or cannot develop a viable RV. In fact, video analysis of the July 28 reentry casts doubt on whether the test was even intended to contribute to North Korea’s reentry vehicle development.

First, the available evidence suggests that at least some of the reentry vehicles tested in 2017 lacked a heavy mock warhead. Reducing the payload would enable the missile to travel higher or further but might also make the RV more likely to tumble during reentry by changing its internal mass distribution. Tumbling would increase the drag and slow the RV relative to an oriented RV. While this would reduce the peak stress and heating on the RV, it could increase other stresses that would cause the RV to fail. It is possible, therefore, that when armed with an actual warhead, the same reentry vehicle could survive.

Second, all three long-range missile tests were lofted: They were fired nearly straight up to great altitudes and reentered at steep angles of attack before landing west of Japan.  The conditions experienced by a reentry vehicle on a lofted trajectory differ in some important respects from those that would be encountered on a standard “minimum-energy trajectory” toward the United States.

To be more precise, lofting reduces the total amount of heat absorbed by an RV, but increases the maximum rate of heating.[1] Calculations for the July 28 Hwasong-14 test and November 29 Hwasong-15 test show—assuming a non-tumbling RV—that the peak atmospheric forces and heating rate on the RV during reentry on the lofted trajectory would be more than twice as great as they would be on a standard trajectory flown by the same missile (of 10,000 km and 13,000 km range, respectively). In addition, for both cases, the total heat absorbed by the RV on the standard trajectory would be only about 20% larger than on the lofted trajectories. This suggests that an RV that failed on a lofted trajectory might nonetheless survive on a minimum energy trajectory.

Third, the Hwasong-15 missile is significantly larger than the Hwasong-14 and has a significantly greater capability to carry large payloads to long distances, which means North Korea does not need to be as careful to minimize the mass of the RV. This could allow it to add additional heat shielding and structural reinforcement to ensure an RV on the Hwasong-15 survives reentry.

Video footage recorded by the Japanese TV station NHK of the night sky on the Japanese coast appears to show the reentry of an object apparently associated with North Korea’s July 28 test. This paper presents an analysis of that video and assessment of what it indicates about the status of North Korea’s reentry vehicle development.



The July 28 Test

In the dead of night on July 28, 2017, North Korea launched a Hwasong-14 intercontinental-range ballistic missile into the Sea of Japan from a location given as Mup’yong-ni, which is in the center of the country and approximately 100 km from the Chinese border.[2] The launch occurred at about 23:40, Tokyo-time.[3] The lofted test took the missile to an altitude of 3,700 km before it landed off the coast of Japan at a distance of about 1,000 km from the launch site.

About forty-eight minutes after launch, at 00:28 on July 29, a video camera located in the town of Muroran in Hokkaido and maintained by Japanese broadcaster NHK, appeared to capture a bright object falling from the sky. Speculation immediately centered on the possibility that the object was the missile’s reentry vehicle. Using videos available online, Michael Elleman at the International Institute of Strategic Studies argued the object was the RV and that it broke up before disappearing at an altitude of three to four km. Meanwhile Ted Postol, Markus Schiller, and Robert Schmucker, asserted that the object “was almost certainly the heavy front-end of the nearly empty upper stage,” not a disintegrating RV.

We requested, and received, the original video from NHK (which we call video #1).  NHK also provided other videos that have not been made public—one from a second camera at a different location in Muroran that captured the falling object (video #2), and another from a camera in the town of Esashi (also in Hokkaido) that captured flashes from the object just out of view.  NHK also provided daylight images from all three cameras, as well as the location of those cameras.



What We Know About North Korea’s Reentry Vehicle

We know less than one might expect about the shape and mass of the reentry vehicles that North Korea tested on its two flight tests of the Hwasong-14 ICBM.[4]  This uncertainty motivated our interest in understanding the behavior of the object that was observed falling through the night sky.

During both launches, the reentry vehicle itself was covered with a shroud.  On-board camera footage released by North Korea of the July 4 test shows the shroud being jettisoned about 155 seconds into flight.  The reentry vehicle may be partially visible in this footage, but its shape cannot be determined.

North Korea subsequently released images of a thermonuclear device, which it indicated was to arm the Hwasong-14 (Figure 1).  The images show not only the device, but the shroud and an apparent reentry vehicle. Our working assumption is that the RV in the pictures is the same shape as the reentry vehicle tested on July 4 and July 28.  The November test of the Hwasong-15 may or may not have featured a different design. We used these photographs to estimate the size and shape of the RV (Figure 2).

Figure 1: Images showing Kim Jong Un posing with a mockup of a staged-thermonuclear weapon next to a reentry vehicle, a diagram of the RV, and a shroud. (Image credit: KCNA)


Figure 2: The diagram on the wall of Figure 1 showing the shape of the RV that was likely used in the July 28, 2017, test along with the radius (in cm) of its base and nose estimated from the photographs in Figure 1. The image has been rectified.

The key parameter that determines deceleration during reentry is an object’s ballistic coefficient, or weight-to-drag ratio, β, which depends on both its shape and mass. In particular, β = W/(Cd A), where W is the weight of the object, Cd is its drag coefficient, and A is its cross-sectional area relative to the direction of motion through the atmosphere. The larger the ballistic coefficient, the less it decelerates as a result of atmospheric drag, and the lower the altitude of maximum deceleration becomes.

It is possible to estimate the ballistic coefficient of a reentry vehicle given its dimensions, shape, and mass.  Modeling North Korea’s Hwasong-14 reentry vehicle as a cone with a spherical nose and dimensions given in Figure 2, and assuming it reenters nose-first, its drag coefficient can be estimated[5] to be 0.17.  If the reentry vehicle were loaded with a mock warhead, it would almost certainly have a mass of at least 250 kg. A plausible lower bound on its ballistic coefficient then comes out to be 320 lb/ft2 (15 kN/m2).[6] If the total mass was actually 500 kg—which is probably more realistic —the ballistic coefficient would be 640 lb/ft2 (31 kN/m2).



Estimating the location of the impact site.

It is possible to reconstruct the trajectory of the test based on open information.  We know the launch location very precisely from geolocation.  Images of the launch, including those enhanced with forensic software, show that it occurred from a site previously associated with the assembly of transporter-erector-launchers for North Korea’s ICBM program. This site had previously been identified by researchers at 40.6110° N, 126.4258° E.


Figure 3: Geolocation of the July 28 launch places it at 40.6110° N, 126.4258° E, which is a previously identified site associated with the manufacture of launch vehicles.

We also know the distance traveled by the missile as given by both North Korea (998 km) and the US intelligence community (993 km).  We can use the bearing to the object in the video, as measured from the cameras, to estimate its distance from the cameras to be 259 km (Figure 4). (The two cameras are very close together and the bearing from each to the object in the video is essentially identical.) The result accords almost perfectly with a U.S. Department of Defense claim that the RV fell 103 miles (166 km) from the coast of Hokkaido.  The result is also consistent with reports that an Air France flight that passed through the area about ten minutes prior to reentry (using data acquired from Flight Tracker).

For a long-range missile, the RV separates from the final stage of the missile after that stage’s engine stops burning. After separation, the RV and stage will have nearly the same speeds, typically differing by only meters per second. As a result, the two objects will follow very similar trajectories and will begin reentry at nearly the same time and place.

Figure 4: Visualization of the launch site, missile range, and camera location for North Korea’s July 28, 2017 test.



Reconstructing the trajectory from the videos

Our method for reconstructing the trajectory of the object is discussed in mathematical detail in the appendix embedded at the end. Here, we present a brief summary.

We first calibrated the images obtained by the cameras. We took ten points visible in the daytime shots and, using Google Earth, calculated their distances and bearings from the camera. For video #2, we chose points at sea-level because their altitudes are known accurately (Figure 5). This was not possible for video #1, which shows the town of Muroran and only a small patch of fairly distant ocean, so we instead used the elevation readings on Google Earth to estimate altitudes. Because these readings are only approximate we regard the results obtained from video #1 as significantly less reliable than those obtained from video #2.

Figure 5: Daytime shot from video #2 showing the points used for calibration (numbered black circles) and the location of the object at various points during its trajectory (lettered red circles).


NHK provided altitude and tilt measurements for the cameras, but these seem to differ substantially from our best estimates using ground-truth photographs. For that reason, we left these quantities as free parameters to be determined by the calibration process. In particular, we estimated these quantities by finding the parameters that minimized the distances between where our model predicts the calibration points should appear on screen and where they actually do. Table 1 shows the results.


350 500 650 750 900
Oriented Stage + RV 45 65 85 100 120
Tumbling Stage +RV 15 20 25 30 40

Table 1: Estimated altitude and tilt angles of cameras #1 and #2 obtained by the calibration process.

With these results, it is possible to estimate the altitude of objects at a known distance from the camera. As a test, we used our method to estimate the height of Daikoku Island in video #2 as 41.4 m. Given that its actual height (or rather, its height estimated with an alternative, reliable method) is 40.1 m, the error on our estimate is about 3%—which is pretty pleasing. We also estimated the height of the southern tower of Hachuko Bridge, which is visible in video #1, to be 182.3 m. Given the actual height is 139.5 m, there was a disappointing error in this estimate of almost 25%; as we said above, we take the results from video #1 with a grain of salt.

Using frame-by-frame video analysis software, we then measured the position on the screen of the glowing object at six roughly evenly spaced points in its trajectory (Figure 5). As described further in the appendix (embedded below), with this data, the estimated impact location of the object, and its reentry angle, we calculated the altitude and speed of the object at a series of points in its trajectory.  We focus on data from video #2 since we believe it is a more reliable basis for analysis.

In video #2, the object first appears at an altitude of about 48 km. It becomes brightest—and illuminates thin clouds that it is passing behind—at an altitude of about 41 km. As shown in Figure 6, by the time it reaches 32 km the object appears to have fragmented into two pieces (exactly where this fragmentation starts is difficult to assess but may be around 34 km). The object disappears at around 30 km—apparently because it fades as it slows down and cools, although it is possible it disappears behind thick clouds at that point.

Figure 6: Screenshot from video #2, showing the object at an altitude of about 32 km, by which point it appears to have fragmented.

We used this data to produce a plot of altitude against speed (Figure 7). As it happens, the results for the two videos are quite similar (which is slightly surprising given the problem associated with interpreting video #1). Over the course of its trajectory in video #2, the object’s speed drops from about 6.2 km/s to 5.1 km/s.

Figure 7: Speed and altitude of the reentering object as estimated from video #2 (red dots) and video #1 (blue dots). The black lines show the reentry speed versus altitude for objects of different ballistic coefficient β (in lb/ft2) on the lofted trajectory that was flown by the Hwasong-14 missile during its July 28, 2017, test.



Interpreting the results

As a reentering object slows due to atmospheric drag, most of the lost kinetic energy is converted to heating of the air around the object, some fraction of which is transferred back to the object itself. The intense heating during periods of high deceleration results in the glow seen in the video. Knowing the altitude and speed profile of the object when it is glowing therefore gives information about its ballistic coefficient.

There seems little doubt that the object seen in Japan was associated with North Korea’ ballistic missile launch of July 28. The timing, in particular, is just right—the object was detected at 00:28 (Japan time), which is exactly consistent with the United States’ tracking the missile for 47 minutes after its launch at 23:41 on the previous evening.

However, the object slows down too rapidly in the atmosphere to be a properly oriented (i.e., nose first) reentry vehicle carrying a mock warhead. Figure 7 shows the speed-altitude curves for objects with various ballistic coefficients reentering Earth’s atmosphere after travelling on the lofted trajectory used by North Korea for its July 28 test, along with the data for the object seen in the videos. This figure shows that the object experienced significant deceleration quite high in the atmosphere. In fact, its behavior indicates that its ballistic coefficient is about 20 lb/ft2 (1 kN/m2)—an order of magnitude smaller than our lower bound for a properly oriented reentry vehicle carrying a realistically heavy mock warhead.

To understand the implications of Figure 7, we estimated the ballistic coefficient of the RV, the upper stage of the missile, and a combined object consisting of the stage with the RV still attached, for different assumptions about the masses of the objects. These results are summarized in Tables 2, 3, and 4.


M = 100 kg 250 kg 500 kg
Oriented RV 130 320 640
Tumbling RV 20 45 90

Table 2: Values of the ballistic coefficient (in lb/ft2) for the RV for different values of mass. A real RV with a warhead is expected to have a mass of at least 250 to 500 kg. An object with 100 kg mass is probably unrealistically light even for an empty RV. These values assume a drag coefficient of 0.17 for the oriented case and unity for the tumbling case. The RV is assumed to have a length of 2.4 m and a base diameter of 1.1 m, and the area used to calculate the ballistic coefficient in the tumbling case is the average of the base and side areas.


M = 250 kg 400 kg 500 kg
Tumbling stage 15 20 25

Table 3: Values of the ballistic coefficient (in lb/ft2) for the upper stage for different values of mass. The first two columns assume a mass of 250 kg for the empty stage and 150 kg of residual fuel (5% of the total fuel loading); the final column shows the result for a somewhat heavier stage with fuel. These values assume the stage has a base diameter of 1.4 m and a length of 2.6 m, use an average of the base and side areas to calculate the ballistic coefficient, and assume a drag coefficient of 1.5.[7]


M=350 500 650 750 900
Oriented Stage + RV 45 65 85 100 120
Tumbling Stage +RV 15 20 25 30 40

Table 4: Values of the ballistic coefficient (in lb/ft2) for a combined object of the RV and upper stage for different values of mass. The 500 kg object could be a 250 kg stage with 150 kg of residual fuel plus a mock RV with a mass of 100 kg. The 650 kg object could be a 250 kg stage with 150 kg of residual fuel plus a 250 kg RV. The 900 kg object could be a 250 kg stage with 150 kg of residual fuel plus an RV with a mass of 500 kg. These values assume a drag coefficient of unity for the oriented case and 1.5 for the tumbling case.[8]

Since Figure 7 shows that the object appears to have a ballistic coefficient in the range of 15 to 25 lb/ft2, the values in these tables suggest that the object in the video is unlikely to have been a realistic reentry vehicle.  The more likely possibilities are that it is either a very lightweight (~100 kg) tumbling RV (without a mock warhead and likely not a real RV), a tumbling upper stage with residual fuel,[9] or a combination of the two.

The fact that only one object is seen glowing in the video further restricts the explanations regarding the reentering object.  We have analyzed various possibilities:

  1. The stage and a lightweight RV reentered together as a single object. In this case, the total mass of the object is too small to include a realistically heavy warhead.
  2. The missile was launched without a payload, so there was no RV to reenter.
  3. The stage and RV separated, but one of these objects broke up above 50 km altitude before it could be seen glowing. This seems unlikely because at altitudes above those at which the object was observed, the aerodynamic forces should not be strong enough to break up either a rocket stage designed to withstand the forces associated with launch or a warhead designed to withstand reentry.
  4. Both objects glowed on reentry, but at different enough times that only one was seen in this video clip.

This explanation assumes the RV separated from the stage after the missile’s boost phase and the two objects reentered at different enough times or locations that only one of them was captured in the video. However, assuming a reasonable value for the relative speed of the two objects following separation—5 m/s—then on reentry from the lofted test trajectory of July 28 the RV would reach an altitude of 40 km only about four seconds before the stage did and at a range about 1.5 km longer (these results vary little if the RV is oriented or tumbling during reentry). Both objects, therefore, should be in the video camera’s field of view at essentially the same time.

Both videos show the sky for five to six seconds before and after the object is visible, with no other objects appearing. Even if something had been visible somewhat outside the time window of the clip, it seems likely it would have been noticed by the people who saw the object reentering or by the people who edited this clip out of the longer video. So this explanation seems unlikely.

  1. The stage and RV separated and the RV subsequently remained oriented nose-first during reentry and was heavy enough that it had a large ballistic coefficient, and therefore started to glow below 30 km altitude, at which point it was obscured by clouds and not seen in the video.

Figure 8 shows that an object with a ballistic coefficient larger than about 300 lb/ft2 might not begin to lose energy fast enough to begin to glow until it was below 30 km altitude. In this case, the object seen in the video would be the tumbling upper stage and the RV would have survived to lower altitudes, although the video gives no information about how low.

Is there evidence for thick clouds that could have obscured the object below an altitude of 30 km? The object in the video appears to break into two large fragments at about 32 km altitude. The dimmer one quickly slows and disappears, while the brighter one continues briefly before disappearing at about 30 km altitude. This object may simply have slowed enough that it stopped glowing at this point, possibly because the event at 32 km further reduced its ballistic coefficient. It is also possible that at 30 km the object continues to glow but passes behind clouds. We note that at an altitude of about 41 km the object passes behind clouds that transmit most of the light. So the clouds would have to be thick enough to stop any light from passing through them for several seconds of travel. We cannot rule out this possibility but it seems somewhat unlikely.

Figure 8: High rates of heating of the air and the object itself will occur when atmospheric drag causes the kinetic energy of the object to decrease rapidly. This figure plots d(V2)/dt, which is proportional to the rate of kinetic energy loss, against altitude for different values of the ballistic coefficient β. The upper red dot shows the rate of energy loss for a β = 20 lb/ft2 object at 48 km altitude, where we estimate the object begins to glow. The two lower red dots indicate that an object with β > 300 lb/ft2 would not reach that same rate of energy loss until it was at or below about 30 km.


It therefore appears likely that the object in the video is not an RV loaded with a realistically heavy mock warhead. In any event, we conclude that it is impossible to infer information about the status of North Korea’s RV development from these videos.




This analysis leads us to the following conclusions about the reentry vehicle used on the Hwasong-14 test of July 28, 2017:

  • The object seen in the video has a very low ballistic coefficient that is not consistent with either an oriented or tumbling RV carrying a realistically heavy mock warhead; such an RV would be expected to have a mass greater than about 250 to 300 kg.
  • The fact that only one object appears to have been seen glowing above 30 km altitude strongly suggests that object is either the upper stage or the upper stage with a lightweight RV (lighter than about 250 kg) attached to it. This may indicate a failure of the mechanism intended to separate the RV from the stage. However, a very lightweight mock RV may not have been designed to withstand reentry and may not have been intended to separate; in that case, this situation may not indicate a problem with the separation mechanism.
  • We cannot rule out the possibility that an RV carrying a mock warhead separated from the stage and remained oriented during reentry, and therefore did not create enough heat to be seen in the video before it was obscured by clouds below 30 km altitude. (It is also possible, but seems unlikely, that it was visible before the start of the video clip.) In this case, the video does not provide information about the fate of the RV.


Our conclusion is that the videos show an object slowing down and breaking up relatively high in the atmosphere, but that object is not a reentry vehicle carrying a mock warhead. As a result, the videos do not appear to give any information about the status of North Korea’s development of RVs.

Although it is possible that the missile tested on July 28 carried an RV with a mock warhead but that this object was not captured by the videos, a comparison of the July 4 and July 28 tests provides an additional clue that it did not. Our modeling of the missiles used in these two tests suggests that in addition to some differences in the burn times of the missile stages between the two tests, the second launch may have used a lighter payload in order to increase the apogee of the test from 2,800 km on July 4 to 3,700 km on July 28. This may have been done to exaggerate the apparent capability of the missile. If so, that could also suggest that the missile tested on July 28 carried an RV without a mock warhead, or a lightweight mock RV that was not intended to reenter, or no RV at all.


Author Affiliations

James Acton holds the Jessica T. Mathews Chair and is co-director of the Nuclear Policy Program at the Carnegie Endowment for International Peace. Jeffrey Lewis is a scholar at the Middlebury Institute of International Studies.  David Wright is co-Director of the Global Security Program at the Union of Concerned Scientists.  We would like to thank NHK for providing the footage upon which this analysis was based.


[1] Carl Gazley, Jr, “Atmospheric Entry,” in Handbook of Astronautical Engineering, ed. Heinz Hemann Koelle (New York: McGraw-Hill Book Company, 1961), 10-27–10-17.

[2] The location, identified by the US intelligence community as Mup’yong-ni, is also known as Jonchon or Yongnim-rup. See also Allison Puccioni, “IHS Jane’s examines North Korean missile bases,” Jane’s Intelligence Review, February 2015.

[3] According to U.S., Japanese and South Korean sources, the launch occurred 11:41 pm and the flight time was about 47 minutes.

[4] We know next to nothing about the reentry vehicle on the Hwasong-15, as it was covered by a shroud.

[5] Frank J. Regan and Satya M. Anandakrishnan, Dynamics of Atmospheric Reentry (Washington, DC: American Institute of Aeronautics and Astronautics, 1993), 355.

[6] In this paper we give values of the ballistic coefficient in units of lb/ft2, which is conventional in the US engineering community.

[7] S. Hoerner, Fluid-Dynamic Drag, self-published, 1965, Chapter 16, Figures 14, 20.

[8] S. Hoerner, Fluid-Dynamic Drag, self-published, 1965, Chapter 16, Figures 14, 15, 20.

[9] This possibility was also suggested by Postol, Schiller, and Schmucker.

Acton Technical Appendix