Capability Signals in “Joint Sword B”

The aim of this essay is to propose a set of observable and testable capability signals for the PLA Air Force. Using open-source information, it asks whether the PLA Air Force has the sortie-generation, maintenance and resupply, base rotation, and mission-support capacity to avoid cooling down before an exercise, spike sharply on the exercise day, and avoid a prolonged post-exercise collapse. That is a preliminary line of inquiry for judging whether the PLA possesses a meaningful “cold-start” capability.

This essay uses publicly reported ADIZ flight counts as an outward indicator of activity, because they are among the few OSINT inputs that are available with some consistency. For that reason, every “capability signal” discussed here is an indirect inference. It has to be supported by timing, mission composition, post-exercise recovery curves, and cross-checks against multiple open sources. A single spike day, by itself, is not enough to prove stronger logistics.

Because “Joint Sword B” occurred close to Lai Ching-te’s October 10, 2024 National Day speech, the political context is unavoidable. But this essay does not try to answer the question of motive. The focus here is on how to observe capability.

Among the four clearer PRC exercises directed at Taiwan, a striking comparison emerges. The August 2022 exercise, the April 2023 exercise, and 2024’s “Joint Sword A” all broadly followed the pattern of “quiet before the exercise, surge on exercise day.” Looking at total publicly reported ADIZ counts, the 2022 exercise averaged about 2.47 sorties over the prior 30 days, fell to 1.67 over the prior 3 days, and then rose to 26 during the exercise period. The 2023 exercise averaged about 3.30 over the prior 30 days, fell to 2.33 over the prior 3 days, and then rose to 44.67 during the exercise period. Joint Sword-A averaged about 5.63 over the prior 30 days, fell to 0.33 over the prior 3 days, and then rose to 41 during the exercise period. That rhythm is consistent with a familiar pre-exercise logic seen across many militaries: concentrate maintenance, preserve readiness, and reallocate missions, fuel, and munitions, which can temporarily reduce visible activity before the main event.

The anomaly in “Joint Sword B” is not just that exercise-day counts were high. It is that the earlier quiet pattern broke down. The prior 30-day average was about 9.27. The average for days -14 to -8 was about 4.71. But the average for the prior 7 days rose to 13.14, and the prior 3 days rose further to 15.67. The exercise day on October 14 reached 111. In other words, “Joint Sword B” was not a case of “suppression followed by release.” It looked more like “elevated pre-activity followed by a further surge.” If that pattern is not simply a product of reporting rules or event-definition issues, it deserves to be treated as an anomalous signal in any warning model.

Two Interpretive Hypotheses

This essay argues that “Joint Sword B” should be evaluated by comparing two main hypotheses, rather than by locking onto a single explanation too early.

The first explanation is that logistics and rotation capacity improved. If the PLA could keep ADIZ activity elevated before the exercise, push to a very high peak on exercise day, and then quickly recover to baseline afterward, that would imply maintenance, fuel and munitions support, base dispatch, and cross-unit rotation capacity above prior estimates. Observable indicators supporting this hypothesis would include simultaneous activity from multiple bases rather than a brief overextension by a single base; a complete “air campaign” package including fighters, AEW aircraft, electronic-warfare aircraft, reconnaissance aircraft, ASW aircraft, and tankers; routine patrol activity not crowded out by the exercise; no obvious collapse in counts over days +15 to +30; and public evidence that the elevated pace was not merely a short-duration spike generated at a single point.

The second explanation is that there was an undeclared pre-exercise phase, or that the formal naming date did not mark the real start of the operational rhythm. That is not unusual in military activity. The front end can consist of routine training, combat patrols, force concentration, or unnamed drills, with formal branding applied only later. Indicators supporting this hypothesis would include same-direction, same-platform, and same-mission activity already visible over the prior 7 to 14 days; maritime or aviation notices, NOTAMs, navigation warnings, or public coast guard and naval activity that appeared early; commercial satellite or open flight-path data showing related activity before the official announcement; and a formal naming date that lagged the actual activity timeline.

The third possibility is short-term overextension or a stress test. But if the peak was generated by overextension, a recovery penalty should appear afterward. Counts over days +4 to +14, or +15 to +30, should fall below the prior 30-day baseline. The currently available ADIZ data do not strongly support that explanation. For “Joint Sword B,” the average over days +4 to +7 was about 5, which was indeed below the 30-day baseline of 9.27. But days +8 to +14 recovered to 8.71, and days +15 to +30 averaged about 9.63, with no sustained collapse. The 2022, 2023, and “Joint Sword A” cases also do not show a stable pattern of prolonged post-exercise decay. So the overextension explanation remains possible, but it has to be tested against the recovery curve.

The fourth possibility is a data and classification problem. In particular, Taipei later stopped publishing platform-by-platform breakdowns. That makes it impossible, in this essay, to estimate whether the PLA Air Force was actually operating with a complete “air campaign” package in its forward activity.

Primary Observation Indicators

First, timing. Watch the periods -14~-8, -7~-1, -3~-1, and then +4~+14 and +15~+30. The first set helps determine whether there was a genuine “quiet before exercise” pattern or an undeclared pre-exercise phase. The latter set helps determine whether there was a real recovery gap.

Second, the post-exercise recovery curve. If counts remain below the prior 30-day baseline for an extended period, that supports an overextension or recovery-phase explanation. If there is only a brief drop and then a quick return to baseline, the explanatory power of a simple overextension story declines, and rotation, redeployment, or higher sortie-generation capacity becomes more plausible.

Third, completeness of mission composition. A surge in fighters alone means something different from a synchronized appearance of fighters, bombers, AEW aircraft, electronic-warfare aircraft, reconnaissance aircraft, ASW aircraft, and tankers. The latter is much closer to a standard “air campaign” package.

Fourth, whether routine missions were crowded out. If normal ADIZ pressure, reconnaissance, ASW, and warning missions clearly drop around the exercise peak, that is more consistent with resource concentration. If routine missions are not displaced, it suggests greater spare capacity.

Fifth, spatial dispersion and multi-axis activity. A peak in one direction or one airspace may just be localized pressure. Activity across multiple directions, multiple airspaces, or multiple bases is more useful as evidence for the improved-capacity hypothesis. This needs to be checked against public ADIZ location data, official graphics, visible density in commercial satellite imagery, and open maritime and aviation notices.

Sixth, the gap between public naming and actual activity. If same-direction, same-composition, and same-rhythm activity already appeared before the formally named exercise day, then the possibility that public branding lagged the real exercise start needs to be taken seriously. That does not automatically imply better logistics, but it does change how the starting point and pre-exercise window should be interpreted.

Confidence Levels

High confidence that “Joint Sword B” constitutes an anomalous case. The reason is that its prior-7-day, prior-3-day, and exercise-day values all differ sharply from the quiet pre-exercise rhythm visible in the previous three exercise cases.

Moderately high confidence that “Joint Sword B” does not fit a simple “quiet before exercise” model. The public ADIZ counts are enough to support that judgment, though reporting definitions and formal exercise-date definitions still matter.

Moderate confidence that “Joint Sword B” indicates improved logistics capacity. The public ADIZ counts show no prolonged post-exercise collapse, but the available data are still insufficient to judge whether a full “air campaign” package was present.

Low to moderate confidence that “Joint Sword B” was primarily an undeclared pre-exercise phase or a short-term stress test. Both explanations can account for part of the pattern, but ADIZ counts alone are not enough to distinguish between them.

The preliminary judgment here is straightforward. Earlier cases support “quiet before a major exercise” as a useful but non-absolute warning rule. “Joint Sword B,” by contrast, is the clearest counterexample in the current data. Politics may explain motive, but they do not explain capability. On the capability side, the unresolved question is whether this case reflects a changed logistics model, an undeclared exercise phase before formal naming, or a short stress test that did not leave a prolonged recovery gap in public ADIZ counts. Open-source data cannot settle that question on their own, but they do define a clear task list: check pre-exercise mission composition, base origin, support-platform ratios, public maritime and aviation notices, the post-exercise recovery curve, and whether multi-axis activity can be sustained without sacrificing routine missions. If multiple observable axes simultaneously support the pattern of “no quiet period before the peak, no collapse after the peak, a complete support package, and stable multi-base rotation,” then PLA Air Force sortie-generation and sustainment capacity should be revised upward. If the evidence instead concentrates around same-type activity already underway before the official label appeared, then “Joint Sword B” should be treated as a case where public naming lagged the real start of the exercise.