The Chinese DF-21 Arsenal: Part 2

Part I available here

Unpacking the DF-21 Inventory

So is the conventional-capable DF-21 inventory structured for the first strike role, the campaign-waging role, or both? The only authoritative open source data that can be used to investigate this question comes from two U.S. Department of Defense (DOD) report series for the general public, which means any answers that we may derive must be hedged on the assumption that DOD’s data is accurate. If this is indeed the case, the data makes a virtually conclusive case in favor of the first strike-only sizing approach—at least as of the most recent data available (mid-2013 timeframe). Table 1 below depicts DOD’s estimates of the DF-21 inventory’s overall size during each of the past nine years.

Table 1: CSS-5 (DF-21) Series Arsenal Growth 2005-2013[i]

Year	Estimated Number of CSS-5 Series Missiles	Estimated Number of CSS-5 TELs
2005	19-23 (MOD 1 and 2 Specified)	34-38
2006	19-50 (MOD 1 and 2 Specified)	34-38
2007	40-50 (MOD 1 and 2 Specified)	34-38
2008	60-80	60
2009	60-80	Total: 70-90 MOD1: <50 MOD2: <50 MOD3: <30 MOD5: Not Yet Deployed
2010	85-95	75-85
2011	75-100	75-100
2012	75-100	75-100
2013	Not Given	MOD1: <50 MOD2: <50 MOD3: <30 MOD5: Unknown

While the DOD data does not break the DF-21 inventory’s composition down by variant, it does provide enough information to help piece together an educated guess. For instance, the 2005-2007 reports clearly state the inventory during those years exclusively consisted of the nuclear-armed ‘original’ DF-21 and its -21A variant. This specificity was dropped beginning in the 2008 report, which suggests the conventionally-armed DF-21C was assessed as first deployed between late spring 2007 and late winter 2008.[ii] It would seem that much of the arsenal’s subsequent growth was due to DF-21C procurement, with some additional growth margin resulting from the DF-21D deployments that apparently began in 2010.[iii]

This growth rate and its implications for the arsenal’s composition could be distorted, though, if the PLA was decommissioning ‘original’ DF-21s while simultaneously deploying new -21As, -21Cs, or -21Ds.[iv] Nevertheless, any such distortion would likely be mild because the overall inventory’s small size in 2005 suggests there were relatively few ‘original’ DF-21s still deployed at the start of the observed period. Also, the 2013 TEL estimates from the National Air and Space Intelligence Center (NASIC) indicates the ‘originals’ have not been fully withdrawn from the arsenal. One other possible distortion source might be if DOD included the PLA’s new DF-16 (CSS-11) in its ‘generic MRBM’ estimates of 2011-2012. This seems unlikely, however, because NASIC’s 2013 report lists the DF-16 Mod 1 as a Short Range Ballistic Missile (SRBM), meaning that the DOD decision to publicly redesignate it as an MRBM was likely made relatively recently.[v]  

Reconciling the NASIC and annual DOD report TEL estimates leads to two major additional observations. First, when one sums NASIC’s 2009 estimates of the maximum number of TELs for each variant, the total is far greater than the 2009 DOD report’s estimate of the entire DF-21 series TEL inventory’s maximum size. Considering there is no credible open source reporting (at least in the English language) that the ‘original’ DF-21 and the -21A use different TELs, either those missiles must use the same TEL or at least one of these two variants fields far fewer than 50 TELs.

Second, assuming the ‘original’ DF-21’s and the -21A’s missile and TEL production runs were sized to do no more than replace the obsolescent DF-3 inventory, and given that the nuclear-capable inventory was approaching 50 missiles in 2007, it would make sense for the -21A’s missile and TEL production runs to have entered their final phase by the time DF-21C Low Rate Initial Production began.[vi] Accordingly, it is not surprising that the estimated number of nuclear-capable TELs did not change between the 2009 and 2013 NASIC reports. In contrast, the fact that no additional DF-21C TELs were apparently deployed after 2009—only two to three years after the missile system’s Initial Operational Capability (IOC)—is extremely surprising.

Evidence that DF-21C TEL production ended before 2009 does not necessarily mean DF-21C missile production did as well. In theory, the PLA could have produced many more -21Cs than TELs so that the latter could be reloaded at least once during a conflict. At least through 2012, though, it does not appear the PLA did this. The supporting evidence lies within ratios of maximum number of missiles to minimum number of TELs derived from the DOD reports’ annual estimate ranges.

Table 2: DF-21 Missile-to-TEL Ratio

Year	Maximum Number of Missiles to Minimum Number of TELs
2005	1:1.48
2006	1.47:1
2007	1.47:1
2008	1.33:1
2009	1.14:1
2010	1.27:1
2011	1.33:1
2012	1.33:1
2013	Not Derivable

Ever since initial DF-21C deployments were first documented in the 2008 DOD report, the highest possible number of DF-21 series missiles to the lowest possible number of DF-21 series TELs per the reports’ estimate ranges has never exceeded 1.33:1, with an average notional maximum of 1.29 missiles per TEL over the period. Even when the inventory consisted solely of nuclear-capable missiles in the years covered prior to 2008, the ratio never surpassed 1.5 missiles per TEL (and was inverted in one year). If there were more conventionally-armed missiles per TEL than nuclear-armed missiles per TEL, the 2008-2012 average ratio should notably exceed the 2005-2007 average ratio, the ratio should trend significantly upward between 2008-2012, or the 2012 ratio should exceed both the 2006 and 2007 ratios (with the 2005 ratio discarded because DOD assessed there were fewer missiles than TELs that year). None of the three above statements are borne out in Table 2, and in fact the ratio has only varied within an extremely narrow range over the observed period. This leads to a conclusion that the missile-to-TEL ratio is nearly constant across all DF-21 variants, and reinforces an impression that conventionally-armed DF-21 variant production did not continue at a high rate (and perhaps stopped in the DF-21C’s case) after 2009.[vii]

With the ratio’s stability in mind, a dissection of the PLA Second Artillery Corps’ apparent DF-21 series order of battle helps generate a rough decomposition of the inventory by variant. Mark Stokes, one of the most prominent American experts on the Second Artillery Corps ,estimates that as of 2012 there was one brigade still armed with the ‘original’ DF-21, up to six brigades armed with the DF-21A, up to two brigades armed with the DF-21C, and one brigade likely armed with the DF-21D. Assuming his Chinese-language sources are correct in that conventionally-capable brigades are equipped with a minimum of 12 TELs each, this would translate into a 2012 force structure of 24-30 DF-21C TELs and 12 DF-21D TELs.[viii] Calculating the nuclear-capable force’s size from here depends upon how one chooses to interpret the 2013 NASIC report’s estimates. Given that the NASIC TEL counts by variant did not change between 2009 and 2013, if one assumes the same TEL is used by both the ‘original’ DF-21 and the -21A and that therefore the NASIC estimates for those two variants represents double-counting, this would translate into a 2012 nuclear-capable force of slightly fewer than 50 TELs and a total DF-21 series force structure of up to 92 TELs. However, if one assumes that the ‘original’ DF-21 and the -21A use different TELs (or use the same TEL but NASIC nevertheless chose to report them separately), then the DOD report’s 2012 ceiling of 100 total TELs would suggest a force structure that year of slightly fewer than 50 DF-21A TELs, perhaps 8-10 ‘original’ DF-21 TELs, and roughly the same numbers of -21C and -21D TELs listed earlier.

Applying the 2005-2012 average maximum missile-to-minimum TEL ratio to the above counting cases, one might believe there could be a maximum of 65-78 nuclear-capable DF-21/21As, 31-39 DF-21Cs, and 16 DF-21Ds. Since the sums of these derived missile counts are much greater than the 2012 DOD report’s maximum for the entire DF-21 series, though, the actual missile-to-TEL ratio must either be much closer to 1:1 for each variant or there must otherwise be many fewer missiles of at least one if not more variants than the average maximum ratio would predict. In the force-sizing case most generous to the DF-21C within this framework, a 2012 inventory of 48-50 nuclear-capable missiles still would only allow for a maximum of 40 DF-21C missiles if one assumes an inventory of 10-12 DF-21D missiles.[ix] Regardless of whether this is the case, or whether all the variants’ missile and TEL counts are closer to parity, the fact that DF-21 series deployments appear to have plateaued after 2010 leads to the inescapable conclusion that DF-21C procurement slowed or stopped and that DF-21D procurement did not shift to a high rate at least through 2012. Furthermore, the 2013 NASIC TEL counts combined with the missile-to-TEL ratio’s stability over the observed period strongly suggests that the PLA’s conventionally-capable DF-21 inventory grew little, if at all, through mid-2013. While I will caveat my analysis once more by observing it is predicated upon the DOD and NASIC reports’ accuracy, it would seem that my conclusions are reasonable unless major revisions of DF-21 series historical inventory data or major post-2012 changes to that inventory are reported by authoritative U.S. Government sources in the future.[x]

As for the inventory’s current size and growth trends, cursory statements by authoritative DOD sources during winter 2014 suggests little changed since the 2013 NASIC report. In February 2014, LGEN Michael Flynn, Director of the Defense Intelligence Agency (DIA), testified to the Senate Armed Services Committee that the PLA possessed “a limited but growing number of conventionally armed, medium‐range ballistic missiles, including the DF‐16, which will improve China’s ability to strike regional targets” and that it also “continues to deploy growing numbers of the DF‐21D anti‐ship ballistic missile.”[xi] No mention is made of the DF-21C. When LGEN Flynn’s statement is combined with NASIC’s January 2014 testimony to the U.S.-China Economic and Security Review Commission that “the Second Artillery is expanding its conventional MRBM force,” it becomes reasonable to assume that DF-16 and DF-21D are the two conventionally-armed MRBMs in production, but that the production rates for both were not high at least as of early 2014.[xii] Of course, we will not know if this is true unless and until DOD releases a new quantitative inventory estimate for each PLA MRBM variant, and not just of their MRBM TELs.

Tomorrow, reconciling the conventionally-armed inventory’s size and roles

---

[i] All data in this table except the 2009 and 2013 estimated number of DF-21 TELs by variant is compiled from the 2005, 2006, 2007, 2008, 2009, 2010, 2011, 2012, and 2013 Office of the Secretary of Defense Reports to Congress on the Military Power of the People’s Republic of China. 2005 was the first year this report provided estimates regarding the PLA ballistic missile arsenal’s size and composition. The 2009 and 2013 estimated number of DF-21 TELs is provided in the NASIC “Ballistic and Cruise Missile Threat” for both years. Unsurprisingly, none of the above reports describe the DOD’s DF-21 inventory estimation methodologies.

[ii] Specifically, the DF-21C IOC window likely occurred between May 2007 and March 2008: the publication months of those years’ DOD reports.

[iii] DOD assessed DF-21D IOC as occurring in late 2010. See Andrew Erickson. “Admiral Willard, COMPACOM, Tells Asahi Shimbun’s Yoichi Kato That China’s Anti-Ship Ballistic Missile (ASBM) Has Reached Equivalent of “Initial Operational Capability” (IOC)-Full Transcript Just Released.” 27 December 2010, accessed 8/21/14, http://www.andrewerickson.com/2010/12/admiral-willard-compacom-tells-asahi-shimbun%E2%80%99s-yoichi-kato-that-china%E2%80%99s-anti-ship-ballistic-missile-asbm-has-reached-equivalent-of-%E2%80%9Cinitial-operational-capability%E2%80%9D/.

[iv] See Mark Stokes. “The Second Artillery Force and the Future of Long-Range Precision Strike.” in Ashley J. Tellis and Travis Tanner, eds. Strategic Asia 2012-2013: China’s Military Challenge. (Washington, D.C.: National Bureau of Asian Research, 2012), 144.

[v] See the 2013 NASIC “Ballistic and Cruise Missile Threat,” 13. The first public DOD statement that DF-16 is an MRBM and not an SRBM appears to be LGEN Michael T. Flynn, USA. “Statement Before the Senate Armed Services Committee.” Defense Intelligence Agency, 11 February 2014, 28. For an evidence-based non-DOD estimate that suggests the DF-16 has a range exceeding 1000 kilometers, see Mark Stokes. “Expansion of China's Ballistic Missile Infrastructure Opposite Taiwan.” Project 2049 Institute, 18 April 2011, accessed 8/21/14, http://blog.project2049.net/2011/04/expansion-of-chinas-ballistic-missile.html

[vi] It is difficult to find consistent open source estimates of the DF-3 inventory’s size during the 1980s and 1990s. Some sources claim between as few as 50 and as many as 120 were deployed at their peak. See 1. “DF-3A/CSS-2.” Federation of American Scientists , 15 April 2000, accessed 8/21/14, http://www.fas.org/nuke/guide/china/theater/df-3a.htm; 2. “DongFeng 3 (CSS-2) Intermediate-Range Ballistic Missile.” Sinodefense.com, 27 February 2009, accessed 8/21/14, http://www.sinodefence.com/strategic/missile/df3.asp. The 1999 Natural Resources Defense Council’s estimate of Chinese nuclear forces suggested there were 40 DF-3s paired with 48 DF-21-series missiles (identified by NRDC as -21As but most likely also including ‘original’ -21s) in the Second Artillery Corps’ arsenal. See Robert S. Norris and William M. Arkin. “NRDC Nuclear Notebook: Chinese Nuclear Forces, 1999.” Bulletin of the Atomic Scientists 55, No. 3 (May/June 1999), 79-80. Assuming that Chinese theater nuclear targeting requirements were reduced after the U.S. withdrew from its Philippines bases in 1991-1992 and that fewer DF-21s were needed to replace DF-3s because of the former’s greater survivability and operational responsiveness, it seems plausible that  the PLA’s nuclear-armed MRBM inventory requirement would have decreased by a few tens of missiles during the 1990s.

[vii] This apparent conventionally-armed DF-21 production and deployment plateau is also identified in the Center for Strategic and International Studies’ (CSIS) most recent estimates of the Second Artillery Corps’ historical MRBM force size trends from 1985-2013. See Anthony Cordesman, et. al.”Chinese Military Modernization and Force Development: Chinese and Outside Perspectives.” (Washington, D.C.: CSIS, 02 July 2024), 298-299.

[viii] For the number of brigades per variant, see Stokes, “Expansion of China's Ballistic Missile Infrastructure Opposite Taiwan.” For the number of TELs per conventionally-armed brigade, see Stokes, “The Second Artillery Force,” 144. Of note, Stokes states in both of these works that there are eight to ten DF-21 series brigades. My analysis therefore relies on the assumption that Stokes’ detailed 2011 order of battle estimate remained accurate through 2012.

[ix] For a report indicating the DF-21D missile inventory was roughly this size in mid-2011, see Stokes, “Expansion of China’s Ballistic Missile Infrastructure Opposite Taiwan.”

[x] Note that the July 2014 CSIS analysis suggests the PLA had 80 ‘original’ DF-21s and -21As, 36 DF-21Cs, and 6 DF-21Ds deployed in 2013. While the CSIS -21C and -21D inventory estimates are consistent with the estimates in this essay, the estimates for the two nuclear-armed variants are not. The CSIS force structure estimate for nuclear-armed variants in 2005 exceeds the 2005 DOD report’s estimate by 10 missiles. Assuming that the CSIS numbers for -21Cs and -21Ds are relatively accurate, the CSIS estimates for total DF-21 series inventory in 2010 exceed the 2010 DOD report’s estimate by 21-31 missiles (all of the difference being in nuclear-armed variant force structure). See Cordesman, 299. It should be noted that the CSIS estimate of 80 ‘original’ DF-21s and -21As in 2010 and 2013 is consistent with the maximum 2013 inventory size (78 missiles) that was calculated in this essay using the average maximum missile-to-minimum TEL ratio, then discarded due to its inconsistency with the DOD reporting. None of this is to say the CSIS estimates are incorrect. It must be observed, though, that they differ significantly from the authoritative U.S. Government estimates for the same reporting periods. Since my essay’s analysis is derived from the latter, my conclusions are flawed if the U.S. Government’s publicly-reported data was incorrect.

[xi] Flynn, 28.

[xii] Lee Fuell. “Broad Trends in Chinese Air Force and Missile Modernization.” Statement to the U.S.-China Economic and Security Review Commission, 30 January 2014.