Tracking the DF Express: A Practical Guide to Evaluating Chinese Media and Public Data for Studying Nuclear Forces
Observers of Chinese nuclear politics and force posture are old friends with information challenges. Open-source analysts of China’s nuclear force drew heavily on published Chinese-language writings, footage, and interviews by official Chinese media or private Chinese citizens, as well as commercial satellite imagery. These powerful open-source tools enable scholars to gain insight into some of the Chinese government’s most closely guarded secrets, such as the construction of 119 nuclear missile silos. Reports from well-regarded institutions, such as the PLA Rocket Force Order of Battle report by the James Martin Center for Nonproliferation Studies, offer open-source research that provides concrete data on the Chinese nuclear force, using thoroughly analyzed imagery and Chinese-language materials. Other studies, such as several reports by the RAND Corporation and the Air University’s China Aerospace Studies Institute (CASI), extensively used Chinese military and technical writings to identify patterns in the People’s Liberation Army (PLA)’s strategic thinking in its own words. Combined with the Federation of American Scientists (FAS)’s annual report on nuclear forces, there is a growing and vibrant open-source intellectual community on the People’s Liberation Army Rocket Force (PLARF).
While researchers continue to dissect new information from China, obtaining reliable data has become increasingly difficult for two reasons. First, the Chinese government has curtailed foreign access to sources like academic databases that were previously fair game for scholarly use, complicating the already dense “information fog” over China’s political and military apparatus. Second, unverified, digitally altered, and AI-generated disinformation and misinformation are commonplace on popular social media platforms like X (formerly Twitter). Combined with the multitude of Chinese social media and video websites, weeding out the noise and distraction has become an increasingly challenging task for new researchers in this field.
This essay provides introductory guidance on the usefulness of Chinese social media and video platforms for observers and researchers of China’s nuclear force. This guide may assist researchers in identifying what to look for and on which platform, especially for those who are not advanced or native speakers of Mandarin. In the sections below, I compare a set of popular Chinese social media platforms and discuss the usefulness of each with respect to open-source study of the Chinese nuclear force. I also present a brief glossary of nicknames and vernacular terms related to nuclear matters in Mandarin, along with their translations. I conclude with a brief discussion of the use of AI-enabled translation tools for open-source research on the PLA.
Chinese media and OSINT: What’s good for what?
Sina Weibo (新浪微博)
Weibo is useful for providing timely, authoritative, and chronologically documented information on training exercises, operations, and policy changes that are of propaganda or morale-promoting value. The equivalent of X in China, Weibo is the biggest Chinese-language social media with over 500 million monthly active users as of 2024. It is likely the most influential social media platform in China, as indicated by the vast number of users and a highly agile and effective censorship system. Due to Weibo’s ability to rapidly disseminate information, all major state and military organs, including the PLA Daily, the Ministry of Defense, individual service branches, and all five PLA Theater Commands (战区), maintain official accounts on Weibo (Figure 1). These accounts are directly managed by dedicated propaganda or public affairs teams and authorized to post military content, which sometimes features approved footage and photos of training exercises. Details revealed in the footage or pictures may help researchers identify the unit leadership and the weapon systems used during the exercise. Additionally, Weibo is often the first platform to announce state media PLA news. The People’s Daily, CCTV, and the Xinhua Agency regularly post links to news articles and updates on Weibo to facilitate dissemination.
For researchers, Weibo contains a reasonably reliable search system. Researchers may use the Weibo search bar to look for mentions of “strategic deterrence,” “nuclear force,” or names of nuclear missiles and use the filter function to screen for content released by official accounts. For well-publicized events like a missile exercise, the topic may be included in the trending (热搜) section for real-time updates. However, a significant limitation of Weibo is that scholars must distinguish whether the content posted by the official accounts directly reflects the Party’s policy or simply shows a lower-level interpretation of the policy by individual units. These official accounts are likely managed by young, tech-savvy officers or civilian employees trained in public affairs.
Sometimes, these individuals might improvise and go beyond what they were prescribed. Some more active accounts, such as the Eastern Theater Command official account, have interacted with random Weibo users in the past and have eagerly implied their belligerent stance toward Taiwan. This led many Chinese netizens to interpret the official account’s posts as a sign of imminent military action, which thankfully was not the intention. Additionally, state-run accounts have also taken down content (primarily propaganda material) for reasons other than revealing unapproved or sensitive information. Again, since the accounts are likely managed by younger personnel at the lower level, contents could be removed when it was later found to be too politically sensitive or too unpersuasive. In 2019, the People’s Liberation Army Army (PLAA) official account posted a propaganda article on Weibo aimed at inspiring nationalistic fervor. It quoted a passionate patriotic poem written by Wang Jingwei (汪精卫), whom the Chinese government considered a “traitor (汉奸)” for cooperating with the Imperial Japanese invaders, most likely because the editor had known about the poem but not its authorship. The comment section quickly pointed out this “political mistake (政治错误)”, leading to the content’s prompt removal. As such, researchers should be aware that removed content does not necessarily suggest valuable information worth hiding.
It is also important to note that accessing Weibo sometimes involves more than typing in the URL. Aside from the content available on the front page (e.g., the trending section), the rich content of the platform is only accessible after logging in. One does not need a mainland Chinese phone number to create an account on Weibo. A virtual number from a trusted provider is also sufficient. Even without an account, researchers could still access the Weibo homepage of many accounts by searching the account’s name in a search engine (for instance, here is the direct link to the official PLA Eastern Theater Weibo page). However, Sina Weibo’s search bar will not be available for unregistered users.
CCTV.com (央视网)
CCTV.com is a webpage that gives scholars access to the state media’s TV programs without a registered account. In addition to live-streamed news stories, the webpage also serves as a large but incomplete archive of past TV programs. CCTV.com has high-definition PLA video footage and interviews, which may be particularly useful for open-source analysis. Some of the CASI reports made clever use of video footage released by Chinese state media to identify key information regarding training exercises and unit deployment, particularly the CCTV-7 channel dedicated to military content. Other open-source intelligence analysts were able to map out key personnel, location, and weapon system information from CCTV news broadcasts and military TV programs. The search bar supports keyword searches and includes government-sponsored TV programs from various channels. The search also returns results from CCTV webpages and the Xinhua Agency. This is the most useful for finding information related to specific missile systems. For instance, among the top results for “DF-26” include footage of a DF-26 missile from a documentary (Figure 2). The search result for “dual-capability” also returned a video by a Chinese military commentator who states that China’s hypersonic vehicle is dual-capable (Figure 3). For open-source analysis, having the ability to revisit footage that might contain useful information on the PLARF is a major advantage of this platform. At the same time, the search function is limited to the titles of the program, not necessarily the content. Furthermore, many of the videos available on CCTV.com are commentaries from Chinese military experts. While the commentary from the Chinese experts may be useful, the visual component may not always be the latest developments. Because of the length requirement of the TV program, the visual element may only have looped videos of known weapon testing or parade footage. Researchers may consider comparing the footage from different programs to remove the repeated material.
Bilibili (哔哩哔哩), Douyin (抖音), Kuaishou (快手)
Bilibili, Douyin, and Kuaishou are among the most used entertainment video-sharing and short-video platforms in China. Bilibili is a video service primarily for animation, comics, and games (ACG) content. It has a “bullet comment (弹幕)” function that allows users to inject text over the video content in real time. The platform attracts over 100 million daily users as of 2024. Douyin (the Chinese mainland version of TikTok) and Kuaishou are short-video platforms with a significantly larger user base than Bilibili, with Douyin content reaching over 1 billion active users monthly. Unsurprisingly, the PLA, the individual Theater Commands, and the Chinese government and Party organs maintain an active presence on these platforms for propaganda, news updates, and publicity programs, often repeating the same message sent across other outlets.
However, despite these platforms’ popularity, they are not great resources for open-source nuclear force research for two reasons. First, there is overwhelming noise from private click-farming content creators who would grossly overstate or fabricate China’s military capabilities to profit from nationalistic sentiments. A researcher may find many videos speculating about the capabilities of the H-20 bomber with no credible source to back up the claims. Private content creators typically have no privileged access to information. In the rare cases where some villagers filmed a rocket booster falling from the sky (some Chinese rocket boosters in the past landed in populated villages), the video is often quickly censored on these closely watched platforms. Second, official government accounts on these platforms almost always repeat information that has been released on Weibo and other traditional news platforms. Some Party organs, such as the Communist Youth League (共青团), which pushes propaganda to the younger generation, would convey the same approved message using CGI videos to boost nationalist sentiment, but the content itself is no more useful than reading the official news release. Overall, there is little added advantage of using the entertainment-based services for potentially useful open-source information.
Combining Sources
While Chinese video and social media platforms could assist OSINT research on China’s nuclear forces, researchers could also combine the visual element of weapon systems with textual data gathered from authoritative Chinese platforms like China Military Online (中国军网), PLA Procurement website (军队采购网), and Qi Cha Cha (企查查). The visual data can help identify many technical aspects of the PLA’s nuclear weapons, but the textual information can greatly inform the acquisition, production, and deployment of the weapon and support systems. Provinces with robust military and heavy industries, such as Heilongjiang, Liaoning, Shandong, and Shaanxi, sometimes release contracting and procurement information locally on provincial and municipal government websites. The information found on local government websites is admittedly sporadic, making broad, systematic collection difficult. Still, such information could serve as valuable first-hand sources for OSINT researchers. For more technical analysis of weapon systems, the China National Intellectual Property Administration (中国国家知识产权局) has a patent database that could be useful to track the development and ownership of certain enabling technologies for nuclear systems. This may be further enhanced by using the China National Knowledge Infrastructure (中国知网CNKI) to locate academic articles on the relevant technology, though access to CNKI articles usually requires an institutional subscription through a university library. Note that many of the Chinese government and military websites do not support a secure HTTPS connection. Some, like Qi Cha Cha, may require the user to access its content from a Chinese mainland IP address. Researchers should deploy cybersecurity tools to take full advantage of these resources.
Nicknames and Vernacular
In addition to the advantages and limitations of different social media platforms, researchers should be aware of the common nicknames and vernacular related to the Chinese nuclear force. Much like how the F-16 is commonly called “viper” instead of the official name “fighting falcon,” there are also nicknames for weapons and systems in the PLA. The table below summarizes several common terms and explains their meaning and primary usage.
A Note on Using AI Translation Services
There is little doubt that AI-enabled translation services like DeepL offer convenient and mostly accurate translations of Chinese texts. However, users should exercise caution when asking the AI to translate lengthy or complicated Chinese texts. Since the Chinese written language system is not space-delimited and often contains a mix of recently invented slang words, formal, and classical Chinese (文言文) phrases and quotes, the translation software sometimes cannot adjust properly to the context in which the classical phrases are used. This could easily lead to misinterpretation. For instance, translating and searching for the phrase “nuclear force (核力量)” may return results that contain the phrase “hardcore power (硬核力量)”, which is unrelated to nuclear weapons. In another example, a PLA Daily article uses the phrase “北约军费连增虚实几何” as the title, which mixes the classical grammar with modern Mandarin. DeepL would translate the word “几何” as “geometry” because it is the most used meaning in modern Mandarin, whereas the correct interpretation is “to what extent” or “by how much” in this context (Figure 4).
In a similar instance, DeepL entirely fails to translate the part that contains classical grammar and offers an incorrect translation (Figure 5). This is most likely because the software treats the original Chinese phrase as a statement, whereas the classical grammar indicates a question.
Therefore, it is prudent to cross-reference and look up phrases individually when using AI-enabled translation tools. Inserting complete paragraphs is likely less accurate than looking up difficult phrases or individual vocabulary.
Conclusion
This paper provides a preliminary guide on using Chinese social media and video platforms for nuclear-related open-source research by reviewing the usefulness and credibility of the content released by various official and privately owned platforms (Table 2). In sum, there is no singular most useful platform for information on the Chinese nuclear force, but some may help piece together interesting findings upon cautious review and cross-reference. While advanced Chinese language proficiency and cultural familiarity remain irreplaceable skills that can greatly enhance the accuracy and speed of open-source analysis, they are neither necessary nor sufficient for successful open-source analysis on China’s nuclear forces. Researchers can still make effective and efficient use of publicly available information by applying analytical due diligence and having context-specific awareness of Chinese sources.
This publication was made possible by a generous grant from the Carnegie Corporation of New York. The statements made and views expressed are solely the responsibility of the author.
Training the next generation of nuclear analysts, one satellite image at a time
The world faces a troubling paradox: as nuclear risks grow in magnitude and complexity, the transparency needed to address them is simultaneously diminishing. Nearly every nuclear-armed state is increasingly shielding critical aspects of its nuclear program from scrutiny; nowhere is this trend more apparent than in some of the most historically transparent states.
As emerging open-source technologies continue to become cheaper, faster, and more accessible, it will be critical – if not necessary – for researchers, journalists, academics, activists, policy advocates, and government officials to arm themselves with the ability to use and accurately interpret these sources, in order to develop a shared understanding of the threat landscape and to avoid misinterpretation and misrepresentation.
However, there are few processes in place to enable the transfer of knowledge to the next generation of open-source nuclear weapons analysts, and the learning curve for new entrants to this field is high; it can take years for a new nuclear researcher to become a highly competent open-source analyst.
This is because nuclear-related open-source analysis is as much art as science: context, experience, and broad knowledge of nuclear policy, operations, and theory are all critical components for performing open-source research at a high level. One must be able to recognize specific types of vehicles from space, spot the differences between nuclear and non-nuclear storage sites, and critically evaluate their sources of information. Moreover, researchers have to understand ethical considerations – they must avoid the temptation to be first at the expense of being accurate, for example – and learn how to communicate their findings and analysis to relevant audiences.
To help address these challenges, the FAS Global Risk team’s Nuclear Information Project, with generous support from the Carnegie Corporation of New York, hosted its inaugural Open Source Nuclear Analysis Bootcamp from 6-12 July 2025.
The bootcamp brought more than two dozen next-generation open-source practitioners from across the United States to Washington DC, where they participated in interactive modules, group discussions, and hands-on sleuthing.
On Day 1, FAS trainers introduced participants to the basic principles of open source analysis, nuclear transparency, primary source assessment and evaluation, and cognitive biases and distortions. In addition, our partners from the Stanley Center for Peace and Security introduced ethical frameworks for practitioners to apply to their investigations: a theme that continued throughout the week.
The bootcamp brought more than two dozen next-generation open-source practitioners from across the United States to Washington DC, where they participated in interactive modules, group discussions, and hands-on sleuthing.
On Day 1, FAS trainers introduced participants to the basic principles of open source analysis, nuclear transparency, primary source assessment and evaluation, and cognitive biases and distortions. In addition, our partners from the Stanley Center for Peace and Security introduced ethical frameworks for practitioners to apply to their investigations: a theme that continued throughout the week.
Throughout the bootcamp, participants were introduced to tools and techniques such as Google Earth Pro, QGIS, Blender, Google Dorking, Freedom of Information Act Requests, FlightRadar24, InVID-WeVerify plugins, and more.
At the end of the week, participants put their newfound knowledge to the test by working collaboratively to solve open-source challenge questions: spotting signatures at China’s nuclear test site; assessing the operational status of Zaporizhzhia Nuclear Power Plant; assessing the performance of a North Korean missile test; modeling a North Korea solid motor test stand; and calculating the plutonium output from a nuclear reactor.
Following the overwhelmingly positive feedback from participants and trainers, the FAS Global Risk team is excited to begin preparing for the second iteration of our open-source bootcamp. Given the significant demand for continued opportunities to learn and collaborate on open-source nuclear analysis, FAS remains committed to fostering and developing an OSINT community of practice, and we look forward to welcoming future participants to our programming over the coming months and years.
Geolocating China’s Unprecedented Missile Launch: The Potential What, Where, How, and Why
On September 25, 2024, the Chinese Ministry of National Defence announced that the People’s Liberation Army Rocket Force (PLARF) had test-launched an intercontinental ballistic missile (ICBM) into the South Pacific. The announcement stated that this was a “routine arrangement in [their] annual training plan.” However, the ICBM was launched from Hainan Island, an unusual location for this kind of missile. In addition, the reentry vehicle impacted in the South Pacific, an estimated 11,700 km away, marking the first time China had targeted the Pacific in a test since 1980 when it tested its first ICBM (the DF-5) at the Jiuquan Satellite Launch Center.
This map, created by Dr. Marco Langbroek (@Marco_Langbroek on X), shows hazard areas from Navigational Warnings and a NOTAM with the reconstructed ballistic flight path. (source)
Given the unusual nature of this test launch and the lack of official information about the status of China’s nuclear forces, this event is an opportunity to further examine China’s nuclear posture and activities, including the type of missile, how it fits into China’s nuclear modernization, and where it was launched from.
What missile is it?
When news of the launch broke, navigational warnings and trajectory calculations indicated the missile was launched from northeast Hainan Island, a Chinese province in the South China Sea. This is not where China normally test-launches its ICBMs, and there is no ICBM brigade permanently deployed on the island. The location and the range of the missile indicated that it was a road-mobile missile launcher, either a DF-41 or DF-31A/AG type. In the days after the launch, several images surfaced with clues about the type of missile and its potential launch location.
An image of the missile lauch released by the Chinese People’s Liberation Army on September 25, 2024.
The first image, released by news outlets on September 25, showed features that made it clear that this was, in fact, a DF-31AG missile. The DF-31AG is a modernized version of China’s first solid-fuel road-mobile ICBM, the DF-31, which debuted in 2006. Since 2007, China has been supplementing and now completely replaced the initial DF-31 versions with the longer-range DF-31A. The DF-31A launcher had limited maneuverability, so in 2017, China first displayed the enhanced DF-31AG launcher. The DF-31AG will likely completely replace the DF-31A in the next few years.
The DF-31AG launcher is thought to carry the same missile as the DF-31A, but the 21-meter-long eight-axle HTF5980 transporter erector launcher has improved maneuverability and is thought to require less support. The single erector arm seen in the above image matches other images of the DF-31AG. The image seems to show that the launcher was partially covered by some sort of camouflage during the launch.
The DF-31AG uses a cold-launch method, meaning the missile is ejected from the canister using compressed gas or steam before the first-stage engine ignites. Unfortunately, this also means it is harder to geolocate the site of the launch because there are unlikely to be burn marks that would normally remain visible on the ground after hot-launching a missile.
How did the missile get there?
The nearest deployment of DF-31AG missiles is with the 632 Brigade located in Shaoyang in mainland China, around 800 km away. There is no confirmation that the missile came from this particular brigade, but the distance gives some perspective as to the process and amount of time it takes to bring a DF-31AG to Hainan Island.
To transport the mobile ICBM to Hainan, it was likely placed onto a railcar and brought to a port such as Yuehai Railway Beigang Wharf before being loaded onto a ship and transferred to Haikou port or a similar location at Hainan Island. From there, the missile was likely driven, along with the accompanying support vehicles, to a sheltered and protected area near the final launch location.
It remains unknown whether the missile was launched directly from the launch position itself, remotely from a local command post, or remotely from a centralized authority.
Where was the missile launched from?
To find the precise location where the DF-31AG was launched on the island, we had to rely on the few photos and videos available to us (mostly captured by locals). To do this, we collaborated with analyst Ise Midori (@isekaimint on X), who carried out a complex analysis of the various launch videos to pinpoint the approximate launch location.
In the above image of the launch, one of the first noticeable features is the devastated vegetation, which matches what we would expect to see after typhoon Yagi impacted northeast Hainan in early September. There is also a small body of water barely visible at the bottom right of the image, which provides a clue when searching for the launch location.
After analyzing the available images, photos, and videos, Midori determined the general area where the launch likely occurred to be in Wenchang, Hainan. While we are unable to determine the exact location with high confidence due to a lack of clearly identifiable signatures, we expect it to be within the area of interest indicated below, potentially at the highway intersection.
Meanwhile, the image below began circulating on social media shortly after the launch. The image reportedly captures the DF-31AG as it was driving to its launch position, although the cloud coverage does not match that from the photo of the launch and could have been taken hours beforehand.
An image of the DF-31AG missile driving on Hainan Island that circulated on social media shortly after the launch.
After observing the road markings and vegetation in the image, satellite imagery from Planet Labs PBC revealed a unique road that matched these signatures. This road is also only 1.9 km away from the launch location area of interest, increasing confidence that the launch occurred at or near this area.
Notably, both the launch area of interest and the location of the DF-31AG on the road are within the boundaries of what seems to be a new military training zone constructed in recent years. This also helps increase confidence in the launch area of interest and highlights this area as important for future observation.
Why here, and why now?
While China has not test-launched an ICBM into the Pacific in over four decades (it normally test launches the missiles in a very high apogee within its borders), it is not unusual for China – or other countries for that matter – to test-launch their nuclear-capable systems. It is interesting, however, to consider why China may have chosen to launch from Hainan Island instead of somewhere that is operationally representative or perhaps easier to travel to on the mainland coast. Nevertheless, the location allows China to fly the missile at full range without dropping missile stages on the ground or overflying other countries. It is unknown whether China will test-launch more ICBMs from Hainan Island in the future.
These types of tests also take months of extensive planning and coordination. Thus, the launch was likely motivated by broader political factors, not in response to particular recent events. Tong Zhao, Senior Fellow at the Carnegie Endowment for International Peace, points out that this test was crucial for the PLARF to reestablish its internal and external credibility following corruption scandals and unprecedented leadership shifts. Additionally, reports of issues with the quality of certain missiles likely prompted a desire to reestablish recognition of operational competence.
Further, because the PLARF conducted the test launch as part of a “military drill” rather than a technological development program, it likely aims to convey military prowess and combat readiness. Conducting an ICBM test over the ocean also likely reflects China’s ambition to solidify its international status as a major nuclear power since the United States also regularly tests its ICBMs over the open ocean.
Notably, the Pentagon confirmed they received advanced notice of China’s test launch, which potentially sets a precedent for pre-launch notification and could leave room for further communication on risk reduction measures. Moving forward, it will be interesting to see if China begins to routinely conduct these kinds of tests beyond its borders and if it continues to provide pre-launch notification to relevant states. The new DF-41 has yet to be test-flown at full range in a realistic trajectory.
This research was carried out with generous contributions from the Carnegie Corporation of New York, the New-Land Foundation, Ploughshares, the Prospect Hill Foundation, Longview Philanthropy, and individual donors.