Scott LaFoyPossible ICBM Modernization Underway at Sundian

This article was co-authored with Decker Eveleth. Decker Eveleth is a junior at Reed College studying International Comparative Policy Studies. He sometimes moonlights as the disembodied head of an Iranian weapons designer.

China may be modernizing some very old ICBM sites. 

An interesting development has taken place at Sundian, historically a PLARF ICBM launch area, over the past three years. Sometime between February 2016 and November 2017, construction began on four large high-bay structures set primarily over previously identified DF-4 ICBM launch sites. Each site has its own large spoil pile building up nearby, indicating possible digging beneath each high-bay structure.

This would be consistent with possible DF-4 rollout-to-launch tunnel modernization or the coring/excavation work necessary for new silos for the DF-41. The construction sites are similar, though not identical to, the 32 meter by 66 meter coverage structures present during the construction of the Wuzhai ICBM test silo visible in 2017. Additionally, declassified documents suggest that covering active silo excavation projects is a long standing PLA operational procedure.

In this post, we will lay out the case for these sites being possible silos or other ICBM modernization sites, and will present caveats and alternatives towards the conclusion.


The Sundian launch area is a known ICBM-basing position hosting both silo and rollout-to-launch (ROTL, missiles stored in caves that are rolled out and launched in front of cave blast doors) sites for the DF-4 ICBM. These sites were identified and tracked throughout several National Photographic Interpretation Center (NPIC) reports in the 1980s that were declassified and released in the past two decades.

Declassified map overlay in Google Earth

NPIC maps line up with still-trafficked roads and sites. In the above map, we have overlaid a declassified map onto the Sundian valley area. Only Site A and the road to Site A do not fit exactly, though it appears that the Site A portion of the map is notional so that it would fit onto the page of the original report. 

Using commercially available satellite imagery, it has been possible to track down both the silo and the ROTL sites with a relatively high degree of confidence.The geocoordinates for each site are provided below.

DF-4 Rollout-to-Launch Site A. Rollout Cave and Storage Cave Visible (Left) Prior to High-Bay Structure Construction (Right)

Main SiteConstruction SupportSpoil
Site A 33.809893° 111.942382°  33.809904° 111.935813° 33.811101° 111.936180°
Site B 33.764735° 112.207739°None Identified 33.769477° 112.204846°
Site C 33.760101° 112.163774° 33.766477° 112.160139° 33.764868° 112.160736°
Site D 33.786187° 112.024023° 33.781664° 112.035204° 33.782887° 112.035175°

Open-source imagery analyst Sean O’Connor originally highlighted both Site A and the one DF-4 silo in his original PLA Second Artillery Force KMZ in 2009, while we derived sites B, C, and D from declassified documents. It should be noted that the existing DF-4 silo (located at  33.793292° 111.989834°) is not being modernized alongside these four sites. It is a fifth Sundian site that does not host any additional construction on imagery available at this time.

In terms of open-source media-derived tracking, Sundian is thought to be associated with the launch brigade headquartered in Luanchuan, identified as the 662 Brigade (MUCD 96762), previously the 804 Brigade (MUCD 96263), under Base 66 (formerly Base 54). In his 2018 PLA Rocket Force Leadership and Unit Reference, Mark Stokes indicates that this brigade possibly handles both the DF-4 and DF-5 classes of missiles, with the unit possibly converting to a new, undisclosed missile type as of 2016. 

Due to the DF-4’s age and the expectation that it will soon be retired, it would be expected that DF-4 launch sites would either be modernized to accommodate life-extension programs, or that the DF-4 would be retired and replaced with a new system such as the DF-41 in a silo basing mode. The 2018 and 2019 Department of Defense China Military Power Report indicated that the DF-41 had a possible silo launching mode under consideration by the PLA Rocket Forces.

Additionally, it should be noted that we do not expect any of these sites are currently related to DF-5 expansion. Sundian itself has only ever been associated with the DF-4 in open-source literature, and there is no current indication that the PLARF is planning on expanding or shuffling the basing locations of its 20 known DF-5 missiles.

The Timeline

In approximately July or August 2017, foliage clearance and ground cover removal was first observed at these sites . By winter 2018, all four sites had large, high-bay structures built over them. Sites A, B, and C were built directly over known rollout-to-launch sites, while Site D was built behind what was previously identified only as a “Complex Support Facility,” though at least one cave entry point is visible nearby. 

The Four High-Bay Structures

Each structure is between fifty to sixty meters long and thirty to forty meters wide. While an exact height measure was not taken, each structure appears to be at least 1.5 to 2 stories high with a triangular/gable roof. The shadows on October 2018 imagery implies that the structures are pavilion-style structures that lack fully enclosing walls, though February 2019 imagery indicates that the structures are at least partially enclosed. It is not clear if this is due to the structures being finished, changes in the phase of construction, or weather-based enclosures due to harsher temperatures.

Based on spoil, construction status, and road orientation, Sites A, B, and C appear to have large vehicles or heavy machinery traffic, consistent with the visible construction and spoil removal. Site D does not have clear evidence of heavy vehicle traffic. 

The Four Spoil Piles and Three Construction Support Areas

All four sites are located in relatively secluded valleys and have visible spoil piles within their respective valleys, indicating nearby soil and rock removal. Sites A, C, and D have in-valley concrete batching plants and have additional temporary support buildings for what appears to be construction related vehicles. Site B’s valley does not clearly have a concrete plant, but is much more heavily wooded and visually obscured. All construction support areas and spoil have appeared since February 2016, alongside the main construction sites.

The Four Valley Spurs

All four structures are similar to the structure placed over silo coring and construction work for the new silo Catherine Dill identified in Wuzhai, which is used for flight testing new missiles or missile launch systems. Wuzhai also had its own on-site concrete batching plant to support construction operations located at 38.888177° 111.605411°.

However these four sites are on flat ground instead of on mountainsides, which deviates from previously known DF-4 silo configurations, but matches certain DF-5 silos and the possible DF-41 silo identified by Hans Kristensen at the Federation of American Scientists at Jilantai training base. The Jilantai site is a novel (for China, at least) totally flat-ground silo. Based on the limited imagery provided, the silo does not appear to have the side flame-ducting previously observed in PLARF DF-5 silos, and as seems to be visible at the Wuzhai site. As Kristensen points out, this possibly indicates that the silo is meant for a cold or mortar launched missile (which does not need the same flame ductwork), which would be consistent with current analysis of the DF-41. This is still only a possibility, as it is not confirmed that the DF-41 retains its cold/mortar launch mode when moved into a silo configuration.

Hans Kristensen, Federation of American Scientists, 2019 . Imagery Credit to Digital Globe via TerraServer (RIP) and Pixel World

If there is coring or excavation work going on for a silo, it should eventually be accompanied by similar signatures to the Jilantai site. A long road so that any missile-loader can properly align with the silo, a concrete skirt for the missile loader to actually sit on while loading the missile, and obviously the silo itself. 

Each site, being situated in a fairly steep valley, would need additional clearance work to accommodate a Jilantai-style line-up road, although a line-up road may not be necessary if these silos utilized whatever crane-loader was used to load the DF-5 valley-floor silos as very briefly mentioned here. Additionally, the new test silo at Wuzhai has what appears to be an 80 meter line-up road, which would be much more easily supported by all four Sundian sites.

 Latitude, LongitudeFirst observable ground clearingFirst appearance of structureStructure removedDuration (months)
Site A33.809893° 111.942382°July/August 2017March 2018Still Present on Imagery22+
Site B33.764735° 112.207739°July/August 2017August 2018Still Present on Imagery17+
Site C33.760101° 112.163774°July/August 2017November 2018Still Present on Imagery14+
Site D33.786187° 112.024023°July/August 2017November 2018Still Present on Imagery14+
Wuzhai38.88874° 111.59757°April 2014January 2015December 201734
Jilantai39.74600° 105.50900°June 2018September 2018May 201920

Dates based on a survey of Planet Labs, CNES/Airbus, and Maxar imagery

In addition to the sites’ visual signatures, the timetables for the construction matches what we know of PLA silo construction operating procedures as outlined both in declassified literature and as observed at the Wuzhai and Jilantai silos. According to a declassified NPIC report detailing DF-5 silo construction timetables, a peak-roofed shelter similar to the ones observed in Sundian covers the silo when excavation begins. The silo is excavated in the early stage, before the circular structural forms for the silo are lowered into the silo in the mid-stage. Construction of these critical components takes about ten months plus excavation time. The reports indicated that silos did not go into operation until after two more years of work, a timeline that roughly matches the silo construction procedure at Wuzhai. 

If the silos at Sundian follow the 34 month timetable set by the silo at Wuzhai, we should expect to see the cover come off Site A in early 2021. However, there are several factors that could impact the exact timeline, such as whether or not the lessons learned at Wuzhai contributed to Jilantai’s shorter 20 month timeline, if these lessons can be applied at Sundian, or if local geologic and logistical conditions affect excavation. Additionally, Jilantai’s status as a training silo instead of an operational position may mean it has reduced construction requirements, allowing for a shorter timeline.

Alternatives and Counterpoints

There are several possible alternative explanations for the construction at Sundian.

This may be the final decommissioning of DF-4 ROTL sites and the conversion of these sites into other underground facilities, such as storage, communications, or mobile missile support forms of storage or support sites.

It is also possible that these are dummy silos. The PLARF is reputed to engage in extensive camouflaging and deception operations, and the construction of inactive silos could contribute to such an operation and increase the cost of identifying potential real DF-41 silos, should they be built elsewhere. 

There has been at one declassified instance of a DF-5 ICBM silo being initially misidentified as a DF-4 rollout-to-launch site, so there is a slight possibility that the purpose of the DF-4 sites shifted but was never declassified or identified by outside analysts. This point is only raised to point out that perhaps the DF-4 ROTLs were misidentified, never declassified, and the current construction is a modernization of hitherto unknown DF-5 sites. This is still a long-shot, from an analytical perspective.

Civilian Highway Construction through the main valley

We also considered the possibility that the structures might be related to the highway construction project currently underway throughout the valley. We assess this to be unlikely for two reasons. If the high-bays are actually garages for construction vehicles or a source of raw stone/material for the nearby concrete plants, it would be an odd place to put four of them in locations not closer to the construction. The highway does clearly require a lot of on-site quarrying and concrete production, but it already maintains near-to-site quarries and concrete production plants. It seems odd, though not impossible, to exploit known strategic facilities for raw materials when quicker sources of materials seem available.

Unanswered Questions

There are several analytic questions that must be answered before these sites could be marked as probable silos as opposed to possible silos. 

Open questions include why Site B does not have visible construction support equipment consistent with the other three sites. It is not clear if the nature of the construction differs, if the construction is on a different timeline than the other three sites, or if it is drawing its construction resources from a significantly less convenient source.

As well, Site D does not have the same historical identification as the other sites. There is not currently available documentation in the CIA’s archives to indicate Site D was previously a launch position, which raises some questions about whether it is a new launch site or the new facility of unknown use. 

The ability of very heavy vehicles, as would be needed for the transport and loading of a DF-41, to transit the roads in these valleys has also not been fully assessed, nor the challenges that nearby streams and water sources may pose to silo construction and long-term operation. 

Answers to these questions would change the calculus for assessing the purpose of this site. The answers to these questions will also be more apparent once the high-bay structures are removed, or if additional roads are constructed to clearly support large, heavy vehicle traffic. 


Based on the visual signatures, the timelines, and the temporal proximity to the construction of two other known silos, these sites seem most consistent with possible silo construction. However the lack of direct observation and some of the outstanding unanswered questions prevent this assessment from rising to the level of probable silo construction.  

These sites have proper site covers that obscure the main construction, spoil piles that imply significant clearance and excavation, on-site concrete batching (for three of the sites) that indicates construction (and not just excavation), and are primarily previously-identified ICBM launch sites (again, for three of the sites). As well, these sites were initiated around the same time as the silo at Jilantai training base and on the heels of the test silo at Wuzhai, at least circumstantially pointing to a possible silo construction program. 

However, so long as the sites themselves are obscured by their high-bay structures, the alternatives cannot be discounted. Additional imagery analysis is needed when these structures are removed to assess whether or not blast doors and concrete loading aprons have been installed, or if these sites have been converted for use for other purposes.

We would expect that these sites, if conforming to previously observed ICBM timelines, should be uncovered by early 2021 and accessible for further analysis. 

References and Further Reading

The Test Silo at Wuzhai

The Training Silo at Jilantai

The Decline and Potential Retirement of the DF-4

A 2017 DF-4 Flight Test

The DF-41’s Silo Launch Mode

Declassified CREST (CIA/NPIC) Documents Consulted